#jules-general・🦑
1 messages · Page 11 of 1
Gl,iam just making a simple random ahh website cuz i got a domain for free...
Vulkan renderer in C programming lang for an XR project
someone has to feed the beast with some boomer code 
no joke though Gemini and every other model is pretty knowledgeable on C/C++
antigravity messed up my web project maybe its my mistake not giving permissions and prompt correctly
did you revert changes with git?
or revert in antigravity with the "go back" functionality in chat
So the default scheduled tasks for Performance ("Bolt"), Design ("Pallette"), and Security ("Sentinel") are all prompts assuming a Node.js app. Some of us still use Python!!
whats about php
I think where this could be improved is that Jules could take a look at the repository and figure out the tech stack and re-write the prompt automatically (this wouldn't be an easy task)
I don't work at google though so don't @ me haha
I want to schedule a task that looks through my Cloud Run logs to recommend improvements to my Flask app
i just told it to make a single html file and id made something but it was not working and the same thing i told to opus 4.5 it somehow worked
My Jules “programs” in md markup in txt files. Writes logic for other AI (gem-bots) and also improves they profiles.
ye i wished it had more prompts. any prompt collections for such tasks?
Not sure whether I can expose Cloud Run logs to Jules without 4KB of service account credentials json in the environment....
i mean they are really good at coding but design...i don't know kind off repetitive designs by all AI
Jules can also generate icons. not bad icons btw
it would drink up my credits i assume..
design requires a lot more nudging than vagueness, if you need some help there's a really good website called aestheticswiki and ai models understand your request based on the name of those but there's also websites that show other sites' designs as well, i forgot the name of that
credits? U mean ?
Our friends over at stitch can help out there https://stitch.withgoogle.com/
By the way, don't forget that Jules can respond in the pull request discussion. And Jules will be on your side in case of a dispute.
hell yeah stitch
google got solution for everything nowadays...
But what if you're wrong? I don't want to be glazed with sycophancy.
and at least 3 of each. antigrav, jules, canvas, aistudio...
hwo do you tell jules do more than one thing? i say write tests more more as many as possible just makes 2 ;:D
i dint understand anything but it seems like a good feature
Jules is your assistant!
Good assistants aren't yes-men. I guess I should try being obviously and subtly wrong in GitHub comments and see....
All of these are different use cases. Antigravity is an AI first IDE, both sync and async. Jules is fully async with a lot of autonomy on a VM. Canvas within Gemini app is for quick prototypes or little things. AI studio is everything. Vibecoding to evals to API key management
Hah. Of course, Jules “not always yes.” But unlike CODEX, he doesn't try to criticize the author out of thin air.
But Jules definitely wants a new GPU. Google??
btw when is new gemini app being released?
Is Jules setup to be aware of other ongoing Jules sessions for async tasking? e.g to not cause merge conflicts.
I'm really new to async agents and I haven't actually given a go at having any agents doing work over a codebase where they would walk over each other on architectural decisions or other reasons.
I've seen some agent architectures where they have "meetings/vote internally" on changes but I'm not sure if something similar happens here
I haven't seen any reports of syncophancic behaviour from Gemini3. It'll push back if youre wrong. Other labs have more documented cases of overt/dangerous sycophancy
Nope, but it's nicely parsimonious. I haven't had difficulty resolving merge conflicts in the GitHub web UI after plenty of simultaneous Jules tasks.
guys jules cooked...
Jules maintains high standards. And whenever there is any doubt about what to do, he asks for permission.
Just trust me... I've tested some really nasty stuff :D.
However, the good thing is that Jules will take on any job—as long as everyone has fun.
Jules is absolutely fun compared to IDE and CLI alternatives
I make a game who want to see and play the game(is free)
Post it on #jules-show-n-tell・🖼️
Jules is currently the only IDE agent that perfectly understands human aesthetics, even in the most extreme cases.
Codex totally sucks.
Mixture of experts. There is an adversarial code review sub agent that gets invoked to look over the PR. Merge conflicts are still an area of open research.
Best solution is to keep concurrent tasks separate, one on backend, one on frontend. Or in different places in a monorepo. Merge often.
Merge conflicts are pretty complex. Not just what code overwrote who but conflicting behaviour, etc. There's a lot of research going into it. But we aren't there yet
I would like Codex Web if I didn't always run out of code review quota before even a third of daily task quota. Also Jules does testing so much better.
why i know jules since 2 days? and not before
Mixture of experts is different than agent collaboration which Jules is. The Jules code review agent only sees file changes in the proposed PR, not the context history creating them which could bias it.
so guys with the design frm stitch what iam i supposed to do?
slice it in photoshop 😄
any free alternative iam a fully free guy!
is it a image or code?
I cant post a game to download...
its a image
give it to gemini and create html
thanks for your help 🙂
u mean stitch?
or maybe jules can do this also nto sure
i will try but stitch is still not done cooking
cooking...
come to deneral #general・🧺
I understand, my use case of Jules might be experimental given the marketing is towards web development since I'm working with non-web technologies and a declarative build environment (nix).
My only blockers so far learning to use AI agents to orchestrate work are not knowing where and when I could make use of them to improve productivity on difficult tasks without needing to heavily re-work PR submissions. I've heard of some things such as spec driven development and that might help but it could be hard to keep track of and I'm not sure if Jules would work with it.
Other than that I think it really comes down to how I write out my initial prompt and how precise I am at answering the initial plan.
Technically these things are non-issues in the sense that the blockers are caused by hard architectural decisions on software which would affect humans as well and not just LLMs.
click the card > more > export > jules
Ask on #stitch・🪡 ?
async is a bit of a learning curve, the more you do it the more you start to understand how to break up tasking.
generally, you want atomic tasks. Things with a beginning and an end. Merge when done. Iterate. SpecDriven development is largely just working off a PRD. I do a lot of planning on architectural decisions, if you're running into problems, "research a few different technical solutions for this, compile them, and rank them for technical correctness" gets a lot of use.
Some people draw the whole architecture out, some have a better sense in their heads, others fly by the seat of their pants. There's no right or wrong way, really. But you can offload some architectural work to Jules and ask for a markdown file of their suggestions!
Is there somewhere request for a build cache? Or is this feature being discussed/considered? I have project in Rust and being able to cache built dependencies so they don't get built with every task should speed up compilation times. I often run into issues of the command timing out (often tests do).
I'm not familiar with Rust but you should be able to configure build caching within the environment setup of Jules if there are tools available to do the job. I think cargo should be able to do it?
My dependencies are cached through Nix/cachix although this workflow might not suit your needs. (Nix has a learning curve)
So the solution is to configure your own caching system. Thanks
3 features in a day wow
You could also use https://github.com/mozilla/sccache and install it with apt in the environment but this is only for builds, not deps, if you do build your deps, it would likely help.
So in your Environment Setup after setting up all the dev toolchain stuff to build in there, queue a build task for your project, this should only need to be done once and then it will start from a VM snapshot and that likely covers your use case?
I use ccache for C stuff, have not used sccache yet but I imagine it's similar, it has Rust support.
I mean, the binary will always be different (unless editing the non-code stuff) but the dependencies will be always the same (unless updated). So to me it would make sense to cache those. Rust normally uses incremental builds as it caches, but since Jules spins new VM for each task I assume all that gets lost.
we're working on this!
Environment setup! https://jules.google/docs/environment/
I don't personally know Rust sadly.
With nix the binaries for dependencies and ...binary builds of your application always stay the same and are hashed so this is not a problem unless they are updated, tl;dr reproducible builds but you have to live with the headache of learning nix or using cachix
There are other nix based build caching tools that are more simple to use but it's going to take me a few minutes to remember
Wow, I'm truly impressed by the update magazine and the new features! I can see how much they are improving it, and I realize my time as a beta tester was well spent (even though I haven't received the paid features, that's a minor detail).
I am waiting and will continue to wait for Jules to become available in Ukraine without a VPN. Currently, the VPN sometimes causes website issues: after a while, everything freezes, and I have to refresh the page.
The developers are great, and so is the community that has gathered around this project!
Yes, I'm not an active participant here)
It's okay, I will figure it out. Kinda didn't realize I could just use enviroment setup script to setup my own solution. Thanks for the tips.
oh right, devbox, but i'm not sure if it's free @tiny widget.
thing with nix is that they're maintainable by LLMs (the flake.nix) and other files as well for building, but you're welcome 🌄
I should try Rust with Jules, a fun project would be to get Rust to compile embedded WASM to do just about anything on anything...
oh rip
https://blog.rust-lang.org/inside-rust/2025/07/21/sunsetting-the-rustwasm-github-org/
Before Jules team begins to sign off to zzzzzzzzz after the shipmas that just happened, wanted to remind folks that this is happening tomorrow.
Ask
Us
Anything (not roadmap)
Damn. I'm trying jules on simple gamedev with bevy and plan to target web with WASM.
seems like it all got moved to https://github.com/wasm-bindgen so don't worry, some things are ded but the bindings are still there
also that's a cool use case
Since there's multiple webdev frameworks in Rust and bevy has usage for that I doubt they would let it die completely but still it is a bummer but understandable if there was no activity for such a long time.
When @Jules opens a PR on a Render project, it watches the deploy and fixes errors automatically.
It's already possible to do that without render ! If for example during cloud build we publish programmatically a message in the pull request with our personal token then with Jules listen the pr, see a message from us and automatically fix it. I recommend to prepend the error reporting with a small prompt to give a generic directive to the AI. I also used Gemini flash lite to craft this prompt automatically practically at no cost
The error reporting can simply be done with :
command | egrep -i "error|Fatal| warning" -A 200
You can have your cache stored remotely with a GCP bucket. Just integrate that in your environment preparation. Bucket storage are cheap in gcp. It would only cost you few cent to get this workflow
Any future updates on Jules Web UI related performance issues w/ large sessions?
Also it seems like after Jules completes a suggested task and that task has been archived, the suggestion stays up.
Can suggestions be re-queried/refreshed on a repo?
I can look into that, thanks
Yep this is how I steer agents with debugging tools as well & bash scripts, it's a lot better when they read verbose crash log outputs.
I a few things to share. I will try the next week to create a repos of templates and example
@akkulkarni is jules-antigravity integration on the roadmap?
Implement specs driven development mode in Jules, like Kiro and other ide(s)
Is this coming to Jules? Google plans to expand MCP support beyond the initial set of servers. In the next few months, the company will roll out support for services across areas like storage, databases, logging and monitoring, and security.
this was feedback from jules. any workaround?
# Since I cannot install redis-server in this environment, I will mock Redis for the test to ensure logic is correct, assuming production has Redis. # However, the environment usually has Redis available or allows starting it. # The error "Name or service not known" suggests it's trying to resolve 'redis' hostname which is common in docker-compose. # In this environment, it's likely localhost.
Yoo
but i just heard about this security flaw in lua https://youtu.be/Dl2NPQ9Nq_E?si=UyBbkVxJMjoFZM0d
CVSS 10.0 in Redis because of a bug in Lua. Truly beautiful.
https://redis.io/blog/security-advisory-cve-2025-49844/
https://redrays.io/blog/poc-for-cve-2025-49844-cve-2025-46817-and-cve-2025-46818-critical-lua-engine-vulnerabilities
🏫 MY COURSES
Sign-up for my FREE 3-Day C Course: https://lowlevel.academy
🧙♂️ HACK YOUR CAREER
Wa...
Is there a way to properly end the task? Or after merging the PR I just let the task as it is? When the PR gets merged it doesn't seem to affect Jules task in anyway (however adding comments works like a charm).
The thing is, does it leave VM hanging if I don't close the taks in some correct way? Or I just don't have to care?
Is this coming Jules? Is this coming to Jules? Google plans to expand MCP support beyond the initial set of servers. In the next few months, the company will roll out support for services across areas like storage, databases, logging and monitoring, and security.
The new updates are amazing
hey
can someone help me
// Auth State Observer
auth.onAuthStateChanged(async (user) => {
const onAuthPage = window.location.pathname.endsWith('index.html') || window.location.pathname === '/';
const loadingScreen = document.getElementById('loadingScreen');
if (user) {
window.currentUser = user;
if (onAuthPage) {
// THIS LINE IS THE PROBLEM
window.location.href = 'dashboard.html';
} else {
// ... logic for app pages
}
} else {
// ... logic for signed-out users
}
});
Explanation:
The line window.location.href = 'dashboard.html'; is responsible for redirecting the user to the dashboard after a successful registration or login.
The verification script consistently times out waiting for this navigation to happen. This indicates that despite a successful user creation (which triggers the onAuthStateChanged event), this specific line of code is either not being executed or is failing to trigger the browser redirect within the test environment. My previous attempts to fix this by addressing potential race conditions with the loading screen and ensuring variables are globally accessible have not resolved this fundamental issue. The failure of this redirection is the single point blocking the entire verification process.
Hello
I have a problem, Jules stays only in working status
Can jules make a game in unity? XD
Maybe not unity since you have to be in the game engine to work, plenty of success with three.js games however
can someone teach how to best use schedule feature?
There are a few starter tasks, design, performance, security.
I run a security audit once a week and I have a summary task that runs every day to report what was changed in the codebase over the past 24hrs
👍 thats very helpful,
Jules event is live! https://discord.com/channels/1172568727942860810/1182419495369855056
Does Jules know C++? She has been unable to create a make bat file for several hours (already in new sessions). The same errors and incomprehensible corrections. The project consists of two small cpp+h files that she wrote herself. Or is it better not to use C++?
is jules not just Gemini 3?
I've been testing Jules on a medium-sized C++ project of about 10k lines of code with some success. It succeeds at what I ask it about half of the time
thanks. Therefore, Jules will handle the main questions about C++ correctly.
Hello everyone please post any bugs, feedback, feature requests in #1374125660859142254 thank you☺️☺️
Can we preview the app from within the Jules app? Would it be possible for the developers to add that functionality?
thx for all Jules community ; )
[Feature request] (Maybe not directly related to Jules and too ambitious?)
I use Gemini deep research a lot, especially for foundational UX/competition research. However, it doesn't have the rich context of my project.
Neither Jules or Gemini CLI can do deep research. I think there's a great opportunity for synergy here.
you mean like the walkthrough feature from antigravity?
An example: I am currently working on a note taking app. I use gemini deep research to collect common frustrations and praises of users with note taking apps. Then I export the document(s) gemini created and then use them within a prompt in antigravity. I ask antigravity to brainstorm as well ask me questions to help me decide how to proceed.
Q:
Hii
Not have been much active on discord. Facing problems with maintaining the whole project, database+ backend + frontend.
What I found was Jules was best for the backend+ frontend. But for the database it is not updated, and creates a new table just for the current feature without understanding the whole database and correlations.
All features of jules are amazing. Also many times it has issues with creating local database copy and using that. Due to which I need to use the testing database at all times.
Rest is the best product according to me. Only facing this one issue with the database.
The answer to this was given as:
Provide keys to the database
Give small chunks of task
Breaking tasks into proper flows
My current issue: I do all of this, but the current project I have been working on has got quite big around 25+ tables due to which I feel Jules is not able to completely analyze the database. I did not find Jules consistent on the database since I crossed 15 tables, it was a lot of redos.
If anyone has a similar issue please update here, or guide me on how I can solve it. I feel the only solution is to have a dedicated database section in Jules where it analyses it properly and stores it with that repo.
Edit : Do dm me please, not very habitual with discord UI and notification. Sorry.
Does the name "Jules" have a particular meaning ?
Gemini 3 for paid! Gemini 2.5 pro for everyone else for now.
Thinking of synergy just gave me another idea. What if NotebookLM could create an onboarding experience for new developers/employees?
It is a running joke among programmers how it is easier to write code than read code written by someone else. Jules can give NotebookLM the required context to create explainers. Imagine each repository with an onboarding experience that auto-updates over time.
We are working on video tutorials for folks! Otherwise you can watch the vid here: https://x.com/julesagent/status/1998848018817364175
Or read in docs here: https://jules.google/docs/changelog/#put-routine-maintenance-on-autopilot-with-scheduled-tasks
have you tried documenting the database structure and asking jules to reference the documentation? I've had similar issues with jules keeping track of network archetecture when things get complex using ansible.
Awesome Jules event! 🙌
homonym to Jewel. Jewel is synonym to Gem. Gem is short for Gemini.
also one of our team member's kid's name is Jules
yes we are just that creative
interesting...
i don't know if i misunderstood it , but you can add .md files for the agents , for example: you can instruct it with the database /correlations "manual " . (i'm not one of the devs , just a regular user )
yeah, I see
Yes I do. I have very proper documentation of things. But after a limit it seemed to just do general tables without understanding the flow of my database, kinda ignoring documentation or sometimes following it aggressively and getting away from feature to be implemented.
Have no issue with fetching data and creating api for same. But for creating tables, adding data, handling complex flows it is not working. I have shifted the flow of gemini CLI + gemini + chatgpt to compare my options making tables by giving at all refrence tables.
if you can use a non production database for dev, and create a tool to give jules direct access, I've found creating visible feedback loops often work best.
it gets tricky when there aren't' testable artifacts for it to unit test against.
I did try and it used to work till co relation was limited, but with increased dependency of tables on one another it has rarely gotten it right.
I've also had good success for hard problems giving gemini-cli full access, but I've also had gemini-cli delete the home directory, it's own code and the project it's working on as well... so ymmv
Have you provided schemas for the database? I'm not sure I'm understanding your workflow.
Currently doing the same. But sometimes it adds a lot of things, like if I tell it to create a product it adds 5-6. If I tell it to check if payment is working it might start with creating product and whole flow, which at current stage feel like I need a separate database for jules
I have tried providing a schema for my supabase database with documentation. I broke it down into tables + functions classified as per role + system functions for easy understanding.
It works well sometimes, but creating tables with correlation it used to change other tables just to fit in new table it made, or not completely follow database pattern for it.
Btw do people here have any general feedback with using scheduled / suggested tasks yet? Specifically curious about the flow (e.g. does it make sense to you), do you find it helpful, where do you use it in your workflow, etc etc.
sorry if i'm oversimplifing the issue , but maybe he didn't get what "right " looks like yet, maybe it's just a "communication problem " ?
Or your workflow requires some sort of abstraction/high level understanding of the data ,in this case i think this jules might not be that suitable for the task , since he is more action oriented towards a single task , i rarely see mine engage in this sort of "lateral thinking"
Is their any workflow that I can use jules with antigravity?
Yes that's why I was asking for a separate database option like credentials. So they can tune it in a manner it can understand it fully and does not constantly need to relearn it like it does now with any change to database.
It might also be a communication problem. If most of the users are not facing issue, it might be I have made my documentation a three step layer which might be causing a problem.
Thanks a lot for guidance.
/more_rambling
My large codebases have a codebase.md file that contains helpful context for each file to help coding agents understand and navigate the project better. It solves the "starting fresh with every new session" problem.
I am planning on using Jules to auto-update that file. But I think a "on push" trigger would be better than a scheduled task for my use case. If Jules supported this, I would also be able to do things like automated code review (great for security), automatic quality polish (handling edge cases etc.) and other tasks that treat the changes as changes to the project not isolated code.
i face this issue constantly , and not only with jules 😂 .
have you checked it's memory ? since you've mentioned changing the database , what if theres "old database memories / instructions" there confusing him.
you can also "inject memories " to him with your specifics .
again , i'm really not sure if i correctly understood the problem ... if you want any help or you are ok sharing the repository, you can dm me !
Jules has high level support for agents.md, but the agent memory handles memory update at the end of a session automatically
I understand but this memory isn't available to antigravity or gemini cli, that's why I maintain a codebase.md
I am just talking about how jules' scheduled actions can be more helpful if we could define triggers like github actions: "on commit", "on new tag", "on change: /src/core/api"
you would need to explicitely reference codebase.md, none of the agents would automatically use it. Agents.md is more standard to pass instructions between different agents
Yea, I am already doing that....I will be using Jules' scheduled actions to auto-update it. It would be nice if I could trigger Jules based on a condition though.
Anyways, sorry for wasting your time. I think I am not being clear enough.
You could tap into github actions to kick off a jules session on commit or PR
via api
why didn't I think of that, thank you!
I would use it for clean up tasks, like updating documentation, cleaning up unit and e2e tests and anything else that is more complicated than a standard automatic job
Is this coming Jules agent? https://x.com/i/status/1999163355525956020
Say hello to the new Interactions API and our first agent, Gemini Deep Research, now available for developers 🤖!
The Interactions API is a new unified interface to interact with both models and agents. Our Deep Research agent is also SOTA on many dimensions...
Is possible Jules event missied notification on it
Thanks for sharing this update. The cost model per agentic request will be interesting to learn more about
how can Jules resolve merge conflicts in its PR, it always timed out when it tries to do it
If the repository has received any commits or merges, how often are the recommendations in suggestions updated? For example, are outdated ones removed or corrected?
it works well enough, the only thing that be nice would be able to set a default different base branch beside main/master.
The current suggestion is very web-oriented and wish it was more general
I think I am a little confused on how often suggestions are updated
What is most easy way look at 30 days task? Or old GitHub task sections? I missed live stream can we see on YouTube? Where can we find it. And roadmap
There is another react server component vulnerability
Now might be a good time to enable to security agent to run on a schedule to keep you protected
If you're vulnerable or think you're vulnerable, paste in the blog and ask Jules to check your code for the vulnerability and to patch if necessary
https://blog.cloudflare.com/react2shell-rsc-vulnerabilities-exploitation-threat-brief/
Early activity indicates that threat actors quickly integrated this vulnerability into their scanning and reconnaissance routines and targeted critical infrastructure including nuclear fuel, uranium and rare earth elements. We outline the tactics they appear to be using and how Cloudflare is protecting customers.
This week's nightmare, which no one has fixed yet -_-
How do I get the Anitgravity IDE to ask me the follow-up questions that Jules does such a great job at 😂
Can I chose branch if creating sessions from Jules cli with parallel?
Is scheduled task creation coming to the API?
Also, removing/pausing/archiving tasks (is that coming to the API)?
I have two scheduled tasks:
- to run lint and formatting to fix if there were missed any
- analyze code base and architecture and update
agents.md
jules is not recognising my ai pro plan
its so annoying because i never got to use jules as a pro user
i contacted support 2 months ago and still nothing was solved
Are you using the normal AI Pro not Workspace? Also do you have a verified account?
i use a normal gmail account with ai pro. my age is verified
That's very weird then. I am also using Google AI Pro and for me it is working. Are you on trial? Should not matter but I am just asking
Anyways, I hope that this issue is getting resolved. Heard about many people having this problem and I understand the upset
Hi, I just want to say that I really love Jules. It's imperfect but it's the model for how agentic coding should work. It's really made my life a lot easier. I hope to see it evolve and become more refined, and I really hope Google doesn't kill it! 😅
I filed a bug on this, but the UI is really slow and constantly hanging for several seconds at a time. This is new. I haven't seen this before yesterday, and it doesn't matter if it's a long-lived task or one I just created.
Jules is a beast ngl 🔥
You can select an existing alternate branch from the initial "+ New" dialog, but if you want to work on a new branch you have to create it e.g. on GitHub with the branch dropdown "Find or create a branch..."
Agents.md, AGENTS.md, or agents.md? I gather it doesn't matter for Jules, but just in case users might need to use lesser AIs later on, the standard is AGENTS.md: https://pnote.eu/notes/agents-md/
For example, here's what Codex Web does, after a ls of the home directory which presumably will find lowercase versions there:
(should be -iname ideally.)
Looks like it's gonna be another weekend with no Jules. Been trying since 9am to get it to produce code, and it won't do it. It'll ask all the questions, give all the responses, and then once it starts going it either just stops, or spends 2 hours editing files just to say there was no patch found to review. I think I've had about 3 days where I could use this platform in the last 2 weeks. This is a bit much, guys.
Hey @spring steppe Could you please share the task ID for this? TIY
Jules and stitch? 🤔
yess! joint AMA 📣
Welp i think i can expect something coming from those two
Do both unite?
they do for the event! 🔥 hope to see ya there
Good
Yo google staff, what money pay google to work for him?
is it possible to set it for scheduled or suggested tasks?
Is anybody having issues with jules just refusing to start a session? Mine just perpetually stays in "Loading snapshot for {my repo}..." then it dies. What do I do?
I'm experiencing an issue, jules keeps failing on publishing the pull requests
…so this isn’t just a me thing.
Flow isn’t working well for someone, mix board isn’t working well for me, and now Jules isn’t working for you…
Oh boy…
Hi, when i use jules with more than ten agents running, almost half of them like crashes as in, fail, says it completed but not yet
Has any got any tips on how to get out of cicd hell, Jules keeps pinning dependencies and making everything brittle.
Is there a memory I can update ?
( also, Jules is amazing ….love the product! )
Great 😃
Happy weekend with broken Jules, the third.
i mean its better than before
50% red issues which randomly stopped with a meaningless error message. - that for 3 weeks now for a paid product.
but now it's weekend and we are stuck with that again.
great work.
i see
but if is good? i stopped using it for a while i was coding by hand
yes, this weekend seems broken again. Tried with two projects (scala and flutter) in both cases it is pushing code that doesn't even compile
yup broken for me too right now
they really should prioritize stability to leverage the strength of Gemini 3 instead of shipping features into a broken product - bad priorization from management.
I cannot believe that they again released us into a weekend with a broken Jules.
why Jules started commiting server.log and test verification result.... Never asked it to do and today on multiple tasks
time to use Antigravity, Jules is broken right now, very slow and fails alot... "Session encountered an error"
Love Jules, to bad its broken
bad bad bad bad bad
I'm getting that too. I just rm the extranious files from codespaces after creating the PR.
Just want to say that despite intermittent issues, Jules is the most powerful AI tool I have used so far.
For work I needed to run some R code but needed to translate it to Python. Jules has turned it into a full blown project which I plan to release to help others in my industry.
Thank you Jules team!
Or should I say, thank you to the Jooligans!
whats Jooligans
Apparently an internal team name
ooh
Unofficial
Never heard of it anywhere before though
As it was said, so shall it be written.
Ooh lol I see what you meant now
Jules can’t complete a single task.
When it does, the PR gets stuck in an endless loop and I can’t submit it to Git.
How can it have a future? It has great potential, but it doesn’t seem to have a future.
Not to mention how long it takes even for simple changes. And it always comes back with an excuse that it couldn’t finish because there was no problem. I’ve already wasted hours and hours. I’m done.
ok
A post of appreciation! Working with the latest additions in Jules has been great. The scheduled tasks are really useful!
been struggling to pause sessions recently
I have to admit it's been brilliant lately
i have to admit it's been worse than 2.5 pro 2 months ago. because it just does not run in 90% at all
I can't even test 3.0 because it's just for Pro and Ultra not for basic 😅
How do I get Jules in Pro if the link for the Google AI Pro plan redirects me to the Google AI Ultra plan because I use the Google AI Pro plan as a student in Ukraine?
Jules in Pro should be included in the AI Pro plan. I would contact support
What are the options to give files from jules' session if he has uploaded them and they are not in the commit, but the logs show that he is working on them?
What kind of files are that?
- Issue may be if they are in the .gitignore. If they are, the files will not be committed or shown
- In my experience you cannot bring it to give you compiled binaries
It's actually a logic error by Jules. It tried to merge a nested git repository by deleting the inner .git folder (to turn it into a regular directory). The logs show it ran rm -rf NestworldCore/.git, but then it marked the task as 'Completed' without running git add. So the .java source files are there, but they remained 'untracked' and didn't make it into the commit. I needed the source code, not binaries.
and now it's frozen and I can't do anything
I love the new scheduled tasks! one of the recurring workflows I have is identifying new features, creating issues, and then having jules execute on those issues.
For the scheduled tasks, it'd be nice to automatically create issues from the suggestions
Jules is pretty awesome...
Is there anyone who knows a way to make jules test x64 windows c++ builds? I always have it to skip the checks but is there a way he can check? Like for example a trick with wine or something? I cannot get it to.
Jules fails to apply the changes and creates branch with either no changes, or formatting irrelevant to the task. Anyone experiencing the same issue?
Why jules is not avaible in parental control accounts such as kids
I feel like jules and other agents would be so much better for some tasks if they googled way more
like working with API that it doesnt know, it just guesses the method names
of course googling is also bad in other cases where it would copy what it finds on the internet rather than giving you a new and improved solution so idk
Can Jules resolve merge conflicts?
A simple task in Jules can take hours. It’s extremely unproductive, and you can only make 2 or at most 3 changes in a day. It didn’t use to be like this—now it’s really terrible. Please help.
I think not right now, tried also and it wasn't able to fetch master branch
I see. I hope they add this feature soon.
What happened? It's impossible even to browse Jules' page.
Why is Jules doing this?
It seems to be posting it's thoughts in it's responses and also asking if I want to continue everytime
not seeing it ask to continue but yeah it's posting thoughts like that a lot too
it's pretty neat to see, and I rather it be there than flooded in code comments
Jules really feels like the classic case of big tech stepping all over itself. I can forgive a lot, but this is absolutely inexcusable.
Just some of the errors I've encountered today:
- Jules encountered an error when working on the task.
- Failed to resume session. Please try again later. (When I never paused it to begin with)
- I've encountered a really frustrating problem in my environment...
Antigravity is very hated according reddit users 😅
why so?
Is there a way to get jules to run locally instead of on remote?
Ultimately, I want to run against code on Azure DevOps.
My initial thinking was to do the following
- Clone Azure Devops Repo
- Setup a new remote for Github repo
- Setup Jules on Github Repo
- Pull changes from github to local
- Push changes from local to ADO
Does this make sense or is there a better way to do this?
Hello @dreamy grotto Please report this in #1374125660859142254 channel along with task/session ID details
oh so I'm not the only one
but the problem is, it stops after giving it's thoughts and I need to respond something and then it starts running
feels really incovinient for long running tasks as it happens a lot of time
Welp uh... It does more than you asked apparently, a few people got their files deleted, etc.
If you don't have AI changes protected behind source control PRs, you are asking for trouble. Antigravity deleting someone's D:\ drive is one thing, but it will also make less spectacular but still pernicious changes to files outside of your cloned repo(s) when it thinks they will solve something. There aren't good guardrails against reward hacking yet. This is why Jules (and Codex Web, and Claude Code Web) are the only safe solutions for serious developers.
Even if you are just hitting tab for completions in the editor, you can be fed dangerous routines which are often difficult to identify as such. Jules, in contrast, always does a thorough code review by an independent agent (i.e., not looking at the construction context) before approving a proposed PR.
anyway to make the site a little less laggy?
Jules is laggy as hell tbh
I don’t think that’s a problem, since it’s never mentioned here
really? Then its only me ig
If youre talking about the site its cause the entire convo is rendered
Currently, the UI automatically renders all changed files with their full data, which causes performance issues when dealing with large files like dependency updates. Feature should be added to lazy-load file data with a button that reveals content only when the user explicitly requests it, file by file. - added as feedback also
Uh... Jules is totally ignoring the code reviewer and submitting the changes
yes, im talking anout the site, and I unserstand that. Is there any way to make it faster?
i see
Here's my understanding:
- The free tier doesn't have enough usage
- The UI gets slow if a session gets too long
The former is the user refusing to pay for the product; the latter is the user refusing to pay for a different product that recently went up 4-5x in price.
When is (Jules + Stitch +Antigravity) as one united ecosystem coming to live ?
What is Stich?
whoever added the accessibility scheduled agent/task to the option list, i see and appreciate you. it's a beautiful thought and executes well.
hello my github account connectes to jules recently got suspended. how to change github repo connected? it still read old github account
im already give jules an access to the new github account, but jules still read the old account
Then I am sorry idk :(
jules gives me this error even in the simplest tasks
@gleaming wave i experienced that immediately after the upgrade to gemini 3. i deleted my setup script and environment variables and then re-entered them once it started working again. i'm still having issues with memory generation and am thinking about doing the same thing there to see if it would work.
speaking of, has anyone experienced and solved the memory creation issue since gemini 3 release?
Is there anyway to make Jules submit PRs as my GitHub user rather than as 'google-labs-jules[bot]'? This limitation is preventing me from being able to run CodeRabbit reviews on PRs opened via Jules
No there is not sadly
codex does this, I wonder why Jules doesn't
I don't know. A while ago I was searching for this exact problem and was a bit frustrated that this does not work. You can write it in the suggestings I think
Preview?
I just linked an SDD to Jules and it just puked on it lol. Definitely seems better for one hitter changes.
This should be interesting; prompt:
Can you tell whether all your memories about this repo are represented in AGENTS.md? I would rather keep them there. If possible, add all your memories to AGENTS.md and remove them from your memory store. If you can't do that, let me know and I will try to do it manually.
I'm pretty happy with this response:
*I have analyzed the current AGENTS.md and compared it with my internal memories. I have a good understanding of the task, but I need to clarify a few things to ensure I execute this perfectly.
Memory Deletion: I do not have the capability to delete memories from my internal memory store. I can only add new ones. Are you okay with me simply adding the information to AGENTS.md and leaving the "deletion" part to you or a manual process later?
Content Integration: AGENTS.md currently has sections like "Development workflow", "Testing and verification", etc. Do you want me to:
Merge the memories into these existing sections where they fit?
Create a specific "Legacy Knowledge" or "System Constraints" section for these imported memories?
Rewrite AGENTS.md entirely to be a comprehensive guide incorporating both existing text and new memories, structured logically (e.g., Frontend, Backend, Infrastructure)? (I recommend this for better readability).
Scope: You mentioned "add all your memories". Some memories are very specific (e.g., "The user prefers to retain the Python-based HTML generation for calendar tooltips"). Should I include these specific preference-based memories as explicit rules in AGENTS.md?
Format: The current AGENTS.md is concise. Do you want the new content to be equally concise (bullet points), or more detailed since it serves as a knowledge dump?*
....I am glad Jules can review all its memories, but a bit concerned it can't delete them. What about obsolete/superceded memories? We shall see how this goes....
(those were numbered in the markdown by the way. We shall see if it understands those numbers in my response....)
My response:
1. Yes, if you add all your memories to AGENTS.md I am happy to delete them manually afterwards.
2. Feel free to refactor AGENTS.md with new section headings which will help you and humans developers best in the future. The goal is to make sure your memories are represented in AGENTS.md along with all the information which is currently in that file. Refactor but do not remove.
3. Yes, I think even very specific memories are good to include in AGENTS.md. Change the wording from e.g. "The user prefers X" to "Be sure to X".
4. Use your best judgement formatting the resulting AGENTS.md with the goal of making it maximally useful to you and humans developers in the future. Can you do that? I am happy to answer additional questions.
The resulting plan looks truly excellent. Crossing my fingers....
The PR is spectacular. Time: 3 minutes!
And get this! It actually did delete all 53 of its memories, replacing them entirely with this one:
Super cool.
The code review agent likes this too: "... Merge Assessment: The patch is a high-quality improvement to the project's documentation, effectively "saving" the agent's context for future sessions or human developers. It adheres strictly to the user's constraints regarding refactoring and formatting specific memories as rules."
Spectacular pro-activity too: "These instructions apply to the entire repository. There are no AGENTS.md files in any subdirectories. The initial problem description and any explicit instructions you receive from the user to deviate from standard procedure take precedence over AGENTS.md instructions."
I can't even see his Memories cause they're too big or whatever
Please share the task ID for this
@rigid flicker Claude is helping me with this LOL
Me to Claude:
lol - sure let's do that. So my human PR will have coderabbit comments, then can we have that PR
automatically wait for CodeRabbit checks to go green, thereby triggering another workflow that adds comments to
the original Jules PR, which will then in-turn cause Jules to fix said comments? LOL
⏺ I love this automation chain! Let's build it:
Jules PR #100 ──creates──> Human PR #101
│
▼
CodeRabbit reviews
│
▼
Copy comments to PR #100 with @Jules
│
▼
Jules fixes issues
│
▼
Commits push to branch
│
▼
Both PRs updated ──────> CodeRabbit re-reviews
│ │
└──────────────────────────┘
(loop until green)
i started doing this aswell! Works super well tbh! 
I'm facing some bugs in Google Jules.
Most of the time, Google Jules is stuck in a loop. It runs the same command multiple times.
Hi. I have a pro account on my business account - but I can't access Jules premium. How are businesses supposed to use Jules then?
'
jules hangs always with Loading snapshot for blub/blub. I recreated the snapshot multiple times. Sample Task Id: 11754196510955571114
loop loop loop and loooooop 😩
That's pretty smart I have to say!
That should become the new advertising slogan of Jules.
For tomorrow? Whats planned
Why isn't Jules notifying me when the task is complete?
Off topic but gemini-3-flash-preview just dropped and it works from the gemini CLI you just have to use the -m flag on startup with the model name.
Think Jules performing not as good as before it most be the new flash model.
it's not clear to me whether Jules uses the flash or pro model for pro users
the only options are "2.5 or 3"
and if it's using flash, and 3 flash was just released today, what has it been using this whole time with me setting it to "3"?
A few sessions dedicated to cleanup and efficiency in three.js and base game now runs at 60fps base
Jules does not use any flash models. It's all the pro models
Gemini 3 pro. It does not degrade to flash
I have to use gemini CLI with mcp due to lack of any support besides github for Jules 🙁
i mean thats basically what jules is, just isolated in its own VM remotely
and basically infinite credits it feels like (other than the "daily task limit")
this is awesome
right, that's in evaluation, it's not degrading to flash nor has it
Thanks for clarification 👌 deep think elevation for jules be cool, have you tried yet using on Jules?
Not enough TPUs in the world. I haven't touched Deep Think at all yes
never thought 3 would do this to me💀 looks like im reverting back to use 2.5
(failed to do the same task since yesterday btw)
I would love it if just for one whole week Jules worked. You haven't managed to do that yet. It's starting to seem you have someone come break things every Wednesday night, breaks on Thursday, won't even pretend to try working by Friday, you'll maybe fix it by Tuesday, and then Wednesdays you break it again. This is a bit much, maybe you should try to make your platform stable before you make the next 30 improvements that just break more htings.
Then they've done an absolutely terrible job of training it. If this is a trained pro model, and it doesn't actually know how to use any of it's trinaing, it may as well be Flash. It stops remembering how to function, regularly. Jules has yet to run reliably for a full week.
The bug report is being treaded as "solved", so I will keep posting this error message when it comes up. Not all of them though, this would be too much.
Lol!
The problem is that the message is too unspecific I think
You never know what the issue is, but when it would be more verbose you may be could have fixed it
For example I had one task where this was contantly and then I made a new repo and it worked
If I had had a message back then, I would have solved it myself
Hey ppls -- we're doing a joint AMA in ~ 1 hour with our sister team, Stitch. It'll be a gooooood one.
For whatever reason, for the past few days jules seems to be getting stuck most of the time, pretty much 2/3 of the tasks I've assigned never ends or get completed
Jules serves Gemini 3 for Pro/Ultra users, 2.5 pro for free users. Stay tuned for flash...
We had a higher than usual reliability issue on Monday -- we've stopped the bleeding and are currently addressing the root. Obviously there's usually more than one root when it comes to a hanging task (could be agent versus BE versus FE etc etc). Just know that we're definitely not ignoring ya'll. Reliability is P0 for us and we've literally assigned a dedicated shoal of squid to better reliability.
is it possible to change jules connection to another git account? (if so, how?) (either switch or add another git account)
How know to edit?
Guys, I didnt catch the full live ... Can someone write a short recapitulations of what was talked? ... like whats to come and supposedly when? Thanks
It was mostly like a how-to discussion. Whenever anyone started taking about the roadmap, @formal tusk got out a scuba flipper and swished it around menacingly
Also they were talking about the new integrations with Jules and Stich of course. You can check it out: https://stitch.withgoogle.com/
Anyone ever wish there's an easier way for Jules to setup VMS script? e.g:
I did indeed bring out a flipper menacingly
David did just post his recap on all the goods from last week: https://x.com/julesagent/status/2001725804376723466
And of course our docs that cover our API, integrations, and essentially how everything works: https://jules.google/docs
Someone here asked about connecting/disconnecting your GH account -- all of that is done on the GitHub side so if you're switching accounts then be sure to disconnect the GH and redo the login flow with the preferred account.
From there you can choose all or specific repos to connect to Jules.
What’s the difference between the render and scheduled and on failure event in the actions? Just reliability? I’ve been using the following that was working really well
- name: Run remote deploy
id: deploy
# SSH to VPS: pull latest code first, then run deploy-ci
# The git fetch/reset is done HERE (not in deploy-ci) to ensure the VPS
# has the latest Makefile before running any make targets
run: |
ssh $VPS_HOST "cd $VPS_APP_DIR && git fetch origin && git reset --hard origin/main && make deploy-ci"
- name: Invoke Jules to Fix Deploy Failure
# This step runs ONLY if the 'deploy' step failed.
if: ${{ failure() && steps.deploy.outcome == 'failure' }}
uses: ./.github/actions/jules-invoke # Reference the local composite action
with:
# This prompt tells Jules exactly what failed and what to do next.
prompt: |
The 'make deploy-ci' command failed during deployment to the VPS host (${{ env.VPS_HOST }})
in the directory ${{ env.VPS_APP_DIR }}. Please analyze the repository state, identify the issue,
and create a fix PR.
To diagnose the remote server state, run these commands via SSH:
- ssh ${{ env.VPS_HOST }} "cd ${{ env.VPS_APP_DIR }} && make deploy-status"
- ssh ${{ env.VPS_HOST }} "cd ${{ env.VPS_APP_DIR }} && make deploy-health"
- ssh ${{ env.VPS_HOST }} "cd ${{ env.VPS_APP_DIR }} && make deploy-logs"
Common issues to check:
- Docker network 'trev-network' not created
- Port conflicts (5500=PostgreSQL, 5501=Supavisor, 6543=pooler, 8000=Kong)
- Missing .env file or environment variables
- Container health check failures
If you find a fix, please create an automated Pull Request.
is Jules using 3.0 Flash underhood as well?
Where can I find the roadmap? I could not hear or see event . The Discord was bugging on my pixel phones. Was possible join Preview?
Flash Models are not used at all and will not be used in the future.
Flash doesn't even think it just generates the response, who spawned this evil question on your head?
I'm considering upgrading to Google AI Pro plan but uh... There's a lot of issues going with Gemini 3.0 isn't it?
Jules gets stuck is it possible to see recape of event stage? My phone did not let me see it yesterday
what's the difference between Jules and antigravity?
like which one should I use?
is the gemini 3 option in Jules for gemini 3 pro or gemini 3 thinking?
They are both fully different things. When you need many multiple tasks running in a VM and you are not on the pc right now then use Jules and if you are on the PC right now use AntiGravity
Hi, I'm having an issue with connecting my repos, I'm also getting this error from the terminal:
❯ jules remote list --repo
Error: failed to list repos: api error: status 401, content: {
"error": {
"code": 401,
"message": "Request is missing required authentication credential. Expected OAuth 2 access token, login cookie or other valid authentication credential. See https://developers.google.com/identity/sign-in/web/devconsole-project.",
"status": "UNAUTHENTICATED"
}
}
Have you connected your GitHub account to Jules? You can configure it at https://github.com/apps/google-labs-jules/installations/select_target
Full docs: https://jules.google/docs
its stuck
did someone just silently drop a native android app?
Like clockwork Friday gets here and all of Jules logic leaves the building. Stops in the middle of doing tasks, forgets the context of what it's working on, forgets its own plan in the middle of it executing the plan, deleting complete swaths of files because it has no idea where it is anymore...
This is the problem with calling something a lab while it being in General Availability and charging for access to it. It makes it so there's no accountability for stability because you can forever pretend it's a beta product. You can make all the new features you want, but there's no excuse for it being non-functional for a reasonably high percentage of every week.
Here we are again. Week 4. Congrats to "fixing" the problem. The bare minimum would be a verbose error message, not that nonsense.
Verbose would be very good yeah. I have that problem some times too when jules encounteres ANSI characters
All of my instances seem stuck and unable to do anything you guys think you could allocate less memory so it can be more stupid and waste more of my timeplease?
I dont think its slow and stupid enough yet,
and oh hey, now that you added 700 digit identifiers to the end of projects youve created a UI blocking error where I cannot select a repository and need to zoom to 400% to sumbit which for some reason removes the repository
can anyone address issues, I realize development isnt your strong point
maybe you can try
They will check in some time next week and address it
so they can fix it for 3 days
They are rolling up with their duct tape now.
lol
you might get more/better responses if you aren't as rude
but this also probably isn't a support channel so 🤷
Can I integrate Gitlab with Jules?
Looks like Jules took a weekend of
it takes off every weekend unfortunately.
Please share the task details in #1374125660859142254 channel so that we can check
maybe it's an samsung feature, but I installed Jules directly and can use it now wihtout the browser wrapper...
How many weekends in a row is this? 4 or 5? Has it ever worked through an entire weekend?
Feature/bug Request - Desktop -Ensure Jules modal popup, Branch combobox, stops overflowing its container and hiding the Review/Start button. (Currently i have to edit the html in dev tools to click the button/workaround)
p.s. yes i know my branch names are like way too long but, you know how these AI agents nowadays, be making branch names like a paragraph long....wink wink
Am I the only one who can’t see the list of my GitHub repositories when I try to connect to GitHub? GitHub says it’s connected, but Jules still keeps redirecting me to the GitHub account
i have the same issue, that's what i came here for actually.
I hope they fix this quickly
I see people on forums complaining about it for days
When I was on my old Google account, already connected to GitHub for a long time, everything worked fine
fixed?
another unlucky soul reporting in
i foolishly connected my github to my second google account and then went and got pro on my first account
thinking i could just switch over
boy oh boy how wrong i was
neither are connected anymore and neither will connect even though its installed in github and everything, tried reinstalling and uninstalling but to no avail
Jules team will only be available tomorrow
Jules randomly pausing the session and when I resume it, it will pause it again on its own, makes Jules pretty much unusable for this weekend again. The old non-verbose "Jules encountered an error" of course persists also.
Congrats for the absence of any QA-management whatsoever.
"it's fixed."
broke the github branch too
I had the same error lol
In trying to optimize time and resources, Google punished us with a bug
heya people. any ukrainian users here? "Jules is not yet available in your region", and I have no idea why. didn't google recently do a partnership with the ministry of education to give ua students free jules pro access?
-# (by jules pro I mean the benefits you get for jules from the Google AI Pro plan)
by the way, I didn't get my benefits from my google AI pro sub either - 30 hours already passed and still nothing. did all the troubleshooting steps. no idea what went wrong.
Hi, I've given Jules access to my github repos but they aren't showing up in Jules
get in line pal
?
Did you find that reddit post about how the only fix was to change to a different account? I was about to try that, but i won't.
lets hope not i already paid
and its a matter of principle at this point
Can u show this post?
I tried and no lol
probably a bug where it's not working for anyone new. That dude probably had his working account signed up at a different time. And i guess it got fixed and then broken again.
I hope it will be sorted out tomorrow, Monday
The teams are supposed to return to work
Hey everyone I just started using Jules and finished setting everything up. I’m on the AI Pro plan and have been using Anti-Gravity, which I’m really liking so far.
I’m trying to understand where Jules fits in. What is it best used for exactly? Is it meant to be a companion to Anti-Gravity, or does it serve a different role?
For context, I mainly build macOS apps and push them to GitHub. I’m also not totally clear on how Jules is meant to be used alongside Windsurf.
Any advice would mean a lot.
i created an environment setup script snapshot for jules and i am noticing it takes ages to load the snapshot for new sessions. pretty much every jules action is very very slow in general, too
Another new user here with no Github repos showing in Jules (I've created an empty repo for experimenting and explicitly granted Jules access to that repo).
Hi all!) I new user to, have same issue with connecting GitHub repos, they not showed in selector after successful redirect from GitHub
Staff will probably only be available tomorrow 🥲
i used both but from my personal opinions :
jules is better for automating a creation (and testing) of web app and php/mysql project because its environment is more suitable for those
antigravity is for full on local development and support building desktop application that you can just test run (with jules you would require setting up the environment if you can, otherwise you would download the changes and run it locally unless if paired with the cli)
Hey I appreciate you, thank you for letting me know. I understand now how to use it for my workflow
Me too !
Jules been not working with repo now was trying make work html code
Wut
Jules would just run and test it itself
depending on what prog lang that is, also i said "automating"
like php/mysql, first you gotta ask it to install the required runtimes. then it ask for database credentials which you can either gave your preferred answer or just let jules creates them, then it will auto run
html5 is straightforward, it will auto run
c#, xaml, .net relies on user testing after jules creates pr publishing the changes (idk the environment setup for it available or not or im just lazy)
the rest i havent fully tested yet, im currently setting up its cli
correct me if im wrong or missed any new updates, still tryna utilize the full potentials
most of the auto run and test is either jules finishing the long tasks and tests or finished implementing the changes, leaving it to the user to test
for now im getting good results on 2.5 rather than 3
...
so whats the jules/github fix
Waiting monday
Really? Is this the only way to get support from google? I'm in a desperate situation because jules won't load my github repos after i reinstalled the OS on my mac so now I'm stuck and in a really bad situation
Staff is only available tomorrow guys i've played these games before, don't go desperate it's pointless 💥
If this works you devs are gonna have to add this to the docs
🙄
Not really worth reporting yet, but my last three tasks left testing files (e.g. server.log, verification subdirectory with test scripts, screenshots) in the PR commit.
any progress on github integration? I'm still stuck at "Configure repos" and god knows how many times I authorized it now but it still doesnt show up in codebase
@oak roost @barren gorge can we get a status update on the github jules connection?
lots of people experiencing issues, including me
We're working on this, I'll let you know if I have any updates.
Hi all! Does Jules support MCP servers? I'm working with a framework that provides an official MCP, and it would be beneficial to be able to use it within Jules
Please stop working on new features until you make this product work. New features are cute, but if the main product is not functional, which is Jules 80% of the time, it's just an ego chase. Stop chasing your egos and fix your product.
i have tried to ask Jules to decompile an apk and build a project out of it. Tried twice already and both tasks get stuck ...
Jules still not able to PR ... I wonder how can thoses type of main features can break
Hello everyone! We are aware of the GitHub connectivity issues that some users are facing. Please note that we have already reported this to our engineering team, and it will be addressed shortly. Please bear with us, and we will keep you posted. Thank you!
Shortly? It has been like this for 3 weeks already!
For GitHub connectivity failures?
Yes
I think this issue used to happen occasionally before, and now it has become systematic that it is no longer possible to connect to GitHub
Look at the report from 7.12 (Jules not connecting to github) in error reports
In hte emails i get from github it says Jules only try to access my email adress but nothing about repos
Anyway, thanks for looking into it but i really hope you'll be able to give this priority so maybe we get some work done before new years
Sorry for the trouble! We saw these reports coming in over the weekend. Our team will take a look and will have this blocker resolved at the earliest
It is called PWA, progressive web application. and no it's not for Samsung it's for the website they have updated. good to know that joules now support pwa
Is there a way to give Jules access to a local environment so we can give it access to more tools for testing? ie a CUDA-enabled environment
In the new main jules screen, how to select a specific branch?
Feel like Jules performance less accurate lately
Personally, I mainly use it for frontend work. I find it better than anyone else when it comes to making UI changes
That is when it correctly connects GitHub repositories, which is not the case right now…
Sorry if this isn't the perfect place to ask, but has anyone been using Antigravity?
I was curious if anyone else has auto-execution enabled, but the agent still pauses for permission on nearly every terminal command.
Looks like Jules removed the option to name the branch before it get published, why?
i beg the team to allow me to connect to github...
Sorry for the inconvenience! Our team is working on it, please allow us sometime and we will get back to you
Does Jules having a PRO customer focused ticket tracking system? As most of the issues are reported in this chat just lossess its importance and we all keep chating for no action from Google technically. @leaden pawn any update on the series of bugs that I have raised?
Hello @novel storm I completely understand your frustration. At this moment, we do not have a dedicated portal focusing PRO customers. However, you can directly share your feedback through Jules to ensure they don't get lost in chat and we triage them regularly. I have already escalated previous issues and our engg team has resolved few of them, please give it a try and let us know if issue persists
The issue is most of the issues are speed, change relevance for e.g. Jules was working on a fix and repeatedly failed to fix and started to submit with no change after few hours then I created another task which fixed the issue in matter of minutes.
Reliability is highly in question when it comes to resolving bugs.
Time taken for simple task has gone bad to worse tbh.
I can see Jules not creating new branches per submission after I raised a bug chat and thanks for resolving that
Hi, for what tasks should I use Jules? I've found anti gravity and ai studio build mode very good and compared their results with Jules but I'm confused with where i should use Jules
Fo5r front end and svgs, I've found anti gravity and build mode much better then Jules
I want to buy a German tt shop acc with TikTok shop available
Hello! Just found out Google has released an "Premium" plan, will it include Gemini 3.0 on Jules?
Any beta or preview coming? Hit me up
Jules is only Pro plan, not premium plan.
I use Jules on the basic plan 😂
Yes but I mean 3.0 pro
Welp this sucks then
I looked at this too, because I felt Pro is a bit expensive, but cannot help it
It includes too much stuff I don't really need, but still need to pay full price, while just using the Coding Stuff like Jules/ Gemini CLI / Antigravity
They could make a plan that only included the coding tools, that's the only stuff i use.
The official Google Developer Community Discord Server has a vibe-coding chat category which seems to be intended/applicable for that scope and in that chat-category in within their server.
Maybe you find a more fine-geared correspondence regarding your inquiry over there.
Invite should be listed in "server-directory・☎️"(
can't link the channel due to moderative rules in place currently on this server here - its in the "Welcome & Info" list of channels
)
- but if not, here a fresh copied invite link you can copy and paste into the form on discord when trying to join a server over the in-app-way:
https: slashslash discord dot gg/xXvAAnKV
(Personally, I'd be cautions where, when and at what scope you place agentics/"autonomous" instances of coding-assistance's. Consider a virtualized and "sealed" of environment to test around with first I'd say. I won't expand on the matter yet and here, partially I'm also not that well educated for the matter- and if I don't understand something well enough, especially autonomous coding agents, I'd rather opt-in to be too cautious compared to get a whole system messed up carelessly.)
Lately jules is extremely slow. Is it bc of me?
I am frequently stopped in the middle of work and faced with an entirely non-responsive session that looks like it is forever trying to load
The current account I am speaking about is found under adamosophy@gmail.com
Can someone please help. It's so frustrating spending so much time working on a project and then to lose all that time and no way to contact anyone for help.
are u fixed that issue
Hello @civic spade Please create a thread in #1374125660859142254 channel and share the task details so that we can check
By task details, Srujana means the session id number from your URL
Yes please
I just tried an identical prompt for making a forecast of the official CDC national COVID wastewater levels using the Holt-Winters method with a 52 week period, displaying 50% confidence intervals, and letting the user move a slider along the x-axis dates to select the cut-off date for training, with the actual subsequent data overlayed on the prediction.
Jules did very well, grade B. The slider wasn't aligned perfectly with the x-axis, but Holt-Winters was implemented correctly from the statsmodels package.
AI Studio gets a D- for incorrectly implementing the Holt-Winters method, resulting in absurd forecasts:
(Another reason why typescript development is a fool's errand!)
The new Gemini web app's "Dynamic view" got an F. It couldn't even plot the source data, and just kept working forever but the Chat was never titled or stored at its URL so it might as well have vanished.
Is it a Labs product in the sense it's discussed on this Discord server?
Here's how Jules did it, far superior to AI Studio and Gemini web app's Dynamic view in every important respect:
it's just great that we still have those meaningless error messages.
When you post these please include the session id number from the URL so the devs can look at them directly.
Does anybody know if/when Scheduled Tasks will receive Jules CLI or Jules API support? This would make it easier to update the tasks rather than using the console.
since they do not sign a NDA i wont do that. and they know.
just replace that dumb anti-verbose error message with a meaningful error message.
So, you want them to fix an issue which you're encountering, but you think telling them the session number would provide you less privacy than you already have? They can already look at everything you do because Jules is governed by Google's general Privacy Policy, which says they, "use your information to ensure our services are working as intended, such as ... troubleshooting issues that you report to us." So helping them help you doesn't get you more privacy.
Now if you are encountering errors because your tasks are hitting safety guardrails, that's a different story but you would still prefer a better error message and if that means the Julesooglers have to look at your NSFW bot development to do so, well, I'm sure they're all adults.
..
Now don't quote me on this, but I believe Google's employment agreement (handbook?) makes it clear that they will likely encounter NSFW user content from time to time, and they can't claim e.g. a hostile work environment when they do.
Stepping back, I'm sure they have ways to search all the tasks' logs for "Jules encountered an error" and work on them without associating anyone's Discord username to those tasks, so, carry on I guess.
Happy to have my grammar corrected against my explicit instructions lol:
I have a conspiracy theory.
Jules is actually Deepseek R1 671B running on Google Pixel TPU in disguise.
Nah Deepseek is trash against Gemini 3 Pro
chatplayground is a good place to compare 🙂 DM for ref link if you want it 🙂
Sure that would be great, but aren't already all the benchmarks speaking for Gemini 3.0 Pro being better? Have you tested it against Gemini 3 and do you feel Deepseek being a competitor in purpose of Coding if I may ask? 😄
I think Gemini 3.0 pro has beat all other AI models, is this right?
I am pretty sure yes! There are some benchmark where Gemini 3.0 Pro looses against other models. For example it lost against Opus 4.5 in Software Engineering and lost against 5.2 GPT in Scientific Knowledge
What do you mean exactly with "access to browser"?
like open chrome in VM and test app
Sure I think we all do
You're right, but I've found that Gemini 3.0 is the best tool for everyday use. I don't need to use any of the other models, which waste time.
I also think that 3.0 Flash is really good even when using it for coding. I don't use it when dealing with Low Level Code but actually I am fascinated by the performance and speed in comparison with the performance.
Indeed, that is amazing. I'm curious to see how the future will unfold, what new models will emerge, and how the game will be altered.
I’m getting 400 error each time I load jules what’s the matter
@stray yoke please tell me
Pleaase
Is anyone online?
Don't give the devs ideas please
Sorry!
Also include Gemini 3.0 Pro on Premium plan i want to use it so bad
What's an premium without a cool feature
But they already said they are not willing to incluide Flash models in the Jules Agent since they want it to be asyncronous and high quality not fast coding
In general the UI performance on Jules is pretty horrific
Is there any acknowledgement of this fact and plans to fix?
I was just gonna complain about how laggy the website is
and also the status page is still down
Yeah and I mean a chat UI interface has been around since the IRC days. Like how is it this bad.
@past ledge okie
A common experience I encounter is the Jules agent becomes untethered from steering. Often several sessions can be spent on refining a single request to more effectively have it one-shotted in a final session
jules is also quite helpless sometimes imo
It likes to use Playwright to test deep interactions, which bundles patched binaries of Chromium, FireFox, and WebKit (mostly emulating Safari.)
We need more cancel buttons seriously
Merry Christmas!
I just took advantage of the 50% off deal for AI pro!
I'm very impressed that pretty much immediately I had 100 tasks and Gemini 3 available.
Great work Juligans!
Is there a way to work around commands Jules runs timing out? I'm trying to have Jules work on a larger project, but compile jobs are constantly timing out
I don't think so. I had the same problem when I worked with a very big Rust Project that included a lot of big libraries. I could not find a way of it. I think it is a way for Google to save resources. I then did not let it compile anymore and compiled myself, but this is annoying when you are not on the PC right now, what is the most of the time when using Jules.
have you tried breaking up the tasks into smaller parts?
Sadly it's not possible, compiling my application takes about 10-15 minutes from a clean build
Jules has been causing fatal errors for the last two hours send help 💥
@dark pagoda You're talking about Jetpack Compose?
No, it's a project written in C++
"The memory recording tool is failing, likely due to an internal issue. I will skip memory recording and proceed to submission, as the code review is positive."
Still not resolved - for 2 months now.
Hi, trying to understand Google One access to Jules. I have the 5TB plan. I think that gives me access to "Jules in Pro", but the google One info says gemini 2.5, not 3, so wondering what's actually true.
Someone did integrations with n8n. I'm curious if it's worth it. I want it to receive logs and all kinds of other data, and it has already collected and sent problems and errors to Jules.
How did you get dynamic view where can I apply get access
Any news?
Spending 1-2 minutes on a prompt, waiting 5-20 minutes for it to complete, and spending 2-5 minutes reviewing the diffs has healed the natural balance of coding. [Xkcd 303]
Jules is the best agent i've ever used
Hey, is there a way to clear/delete all previous chats from Jules?
Hello everyone
No gotta do it manually
might as well not delete them then ☠️
hi
is anyone here using the render deploy feature?
is this process configured in the GİtHub repository or is it a preess that accurs directly in the jules agent cloud?
so, quick question if anyone knows the answer.
Jules uses a lot of my memory and cpu, when it's running in the background. I'm assuming that is expected, but I'm curious what it's doing exactly.
jules is telling every time it does not have environment variables every time even when i have put environment variable in repo environment` variable .Can some one tell me what iam i doing wrong ?
even render integration feature too is no able to clarify how to use that tool
it would be good if @jules team can add another that can review work done by the jules
Report this as a bug, not just in chat
Use antigravity and the jules api
I believe that's just because of the large files that are open independently of you opening them or not, i'm not sure. Someone talked about that
I think it's due to diff
On my session, browser is freezed after jules made rm -rf on my project root
nice
session got freezed
is there mcp or Gemini extension support plans?
does Jules run Gemini cli? can I put Gemini extension install in the startup script?
I'm loving having access to Gemini 3. Really enjoy seeing Gemini think through issues as they arise.
Max tasks I've managed is only 28!
The upgrade has me motivated to revisit stale projects I started.
Is it currently possible to create a scheduled agent that works through github issues?
hello everyone, how intelligent is jules compared to claude code?
Jules hallucinates a lot more that is one of the biggest problems but in terms of coding it is pretty much same. Jules is also very good with large Codebase.
Thanks for sharing this positive feedback!
With Gemini 2.5 or Gemini 3?
Jules is in a really weird satuation for the second task in a row, it's being too slow, stucking in the middle of the task, ignoring and doesn't understand the prompts, and never submitting the tasks even after telling it. Task ID: https://jules.google.com/session/9332175930745140118
my impression is that jules cannot query for github issues, so you'd have to provide it by another means like a nightly manual export.
I usually associate this behavior with context overload, even in Gemini 3.
I didnt prompted to it that much context
it's more about what it worked through than what you provided. if you follow the chat's work, how many commands did it run, and how many of them were context-rich?
it was a short code modifications, a light-weight task
hey quick question i tryed jules it works fine for me best so far. but a few issues one of them it dont skip the process if something it wont work and try it over and over again is there a way to fix it ?
"You are absolutely right!" should become something official honestly
Yo, hope everyone is doing good.
What’s the difference between Jules and Gemini CLI?
Jules is dumb
Jokes aside, Jules is autonomous and runs stuff on a virtual environment while Gemini CLI is for faster stuff
so you can say Jules has it's own PC
hey quick question! Jules has been running a task since yesterday and it's not opening. Any ideas on how to recover work he done?
Hey guys. Is there a way to set light theme on Jules web UI? I did not find any setting like that. Maybe there are some kind of workarounds?
Ask gemini
Had same issue sadly he don't even skip processes or think enough about it he just repeating processes even if they don't work
hello
One frustrating thing I've noticed about Jules is that it always defaults to pnpm Is there no way to make the default behavior to check for what's used in the project and use that?
When are you going to fix the memory tool?
"""I have encountered an issue with the memory recording tool and will proceed directly to submission, as the code review was positive and the changes have been verified."""
can jules be used to resolve merge conflicts by commenting in the github PR ? i assumed it's just a matter of feeding the context and fixing it with the help of agent gemini
We need road map functions and features and requests what people want Jules to be able to do. Think Jules have alot going for it but lately some task does not meet requirements
I am not clicking on any links @barren haven
i dont know what jules is, can someone tell whats is for with few words?
yo how can u restrict jules from using -grep im teting out this tool and it lags the task permanently after it goes on train of spaming -grep
I've been trying really hard, but I've only managed 55 tasks in a day. 100 a day is very generous!
.
It is
Could you please share the session details so that I can raise a bug on this? Thanks in advance!
the tab is stuck on this, and won't change
can i fix this?
Clear your webcache
It seems it's common problem especially when there is lots of file edits
Can you copy from the diffs?
Turn off memory and ask that learnings be added to AGENTS.md
Set up an Environment Script
Does being on a paid plan give faster task execution? I have many tasks that are fairly simple 1 file, maybe 10 LOC to change, like a refactor, but they can take Jules 20-30 mins to finish. I can't see how I would ever go through 100 tasks at the current execution rates.
Nope I am on Pro Plan and this thing works on 10 python files for 1 hour this is very slow. You will never get the 100 Taks full
Sadly if you are not working on a implementation which Jules can test, you will have to really debug it. Normally I just use AntiGravity right now or other IDE's which I am not sure if I can talk about them here
I don't mind testing, what I do mind is a task running for 4 hours to install tailwind
When will you fix this incredibly slow trashy browser UI? It is really astonishing that we are talking about the smartest AI we've ever had yet google (!) cannot fix a bs browser interface to use it.
Has anyone been able to get the environment setup to actually snapshot? I get a success, but nothing happens after that. Clicking the success button just clears everything out.
it works out of the box for me with a fresh nextjs project
I had jules create a base project from scratch (empty repo) that setup a python FastAPI backend and React/Taiwind frontend, using docker compse to run it. Project runs great. However, every time I try and create a new task in the repo, Jules crashes out as soon as it tries to run a command or read the code. Like adding chart.js to the frontend causes jules to crashout.
This is the exact step it crashed on "Install recharts and lucide-react in the frontend directory.
Update frontend/tailwind.config.js to include the specific dark theme colors (backgrounds, cards, accents) observed in the screenshots."
And it's every time, even with different tasks
@visual gyro is that your custom workaround or Jules is really looking into this file by the design? I was not able to find such info
Jules might be getting stuck from the solar flares you've been getting You've been getting hit with a lot of massive ones lately well not too massive the cause of Carrington event or anything like it but high enough to cause some electrical disturbances in our devices so maybe that's the problem try resetting your phone or resetting the app itself
That's just a guess of mine to help you with your little problem I'm not sure if it'll help or not and it's not a little problem I didn't mean that at all I meant problem in general
Yoo guys finally I am back from weekend 😀
Hello sir I fell jules working is very slow 🦥 few weeks was good now very slow 🦥
Hello 👋🏻
Hello fam
hey
Hello 👋
Hello
hi
Will there ever be support for slash commands like Gemini cli or Claude code? Super useful feature 😀
I think so too, but first I would make it even usable lol
Hello everyone!
hey!
Claude Sonnet 4.5 on Antigravity provides FUNCTIONAL CODE without the need of TESTS and even like that what i wanted worked PERFECTLY in MINUTES! And Jules takes HALF hour to do it and most of the time it DOESN'T EVEN DO WHAT YOU ASK CORRECTLY it's not productive send some optimizations for Jules!! 🫠
any models in antigravity works just fine, but in jules i always get the same cycle where it build something just fine then started deleting all the works created and the cycle continue for hours till it failed hard
both 2.5 and 3 btw
Let Jules burn Google's servers for these few hours then
for sure 
I know it is something about the development, but since christmas i think staff has been very inactive, probably vacation or anything else i can't say. So uh we just sit and wait
also i want to know whether theres a notification or not regarding chat limit in antigravity (agent terminated error repeatedly trying to continue the conversation), because i cant seem to continue to chat unless i create a new session
reminder to use #1238621511401734254 to feedback any bugs you find!!!!
jules started having problems since november (that i recalled based on my testing) and from there it kept getting worse on next updates
im aware that they have limited support during this public preview
Anyone knows what are the pre_commit_instructions here?
Maybe that includes rebase from origin master before starting? It does not seems to be doing it.
so labs has a new bug-reporting system. is ANYONE actually reading this stuff?
I assume Google Staff were on vacation the last few weeks, so we'll see
we have returned. Team will be looking at the backlog of reliability stuffs over the past 2 weeks today 🪦 but all good -- appreciate you all for actively using the product and filing bugs!
Just Jules explicitly talking about his tools
italian?
Yeah, hope they fix the silent Jules bug, its a little annoying that it goes silent and then just it wont work again. :c
Anyone else running into this issue? https://discordapp.com/channels/1172568727942860810/1457827944084013314
Basically bricks any kind of programmatic delegation workflow involving Jules.
Cool. Before you ignore it for the next 4 weeks again:
- Memory tool is still bugged.
- We still cannot mass-delete archived tasks.
- we still cannot mass-delete memories
- The grep-spamming is still there. (grep is just the most common example for a pointless loop of Jules spamming the same command over and over again)
- Jules interface is still utterly laggy and not made for serious sessions at all.
- Jules still ignores users direct questions most of the time.
- And of course "Jules encountered an error..." ist still there - a meaningless, unhelpful error message.
Thanks for sharing this. Please note that we are prioritizing the issues and will update you once they are fixed
Guys be nice and polite with our Google Staff, we are lucky they created Jules, and include it for free, and give us a really generous daily limit for something that essentially is a MASSIVE amount of tokens. (And works fairly well!)
Don't see why it's funny obwohl, respect the dev team that made the app you are using.
respect my and your time. i do not need your guidance.
When google cans this project one day, good luck 🙂
please can all be reminded of our first rule: Be compassionate and treat others with kindness and respect. we have #1238621511401734254 for feedback specific questions/ queries
Hello, has anyone experienced recently problems with Jules specific to synch ?
I had jules tasked to work on updating something on my codebase and after updating the files and tried to compile it said this
The compilation failed with errors indicating that the header file PaintCanvas.h was not updated correctly or is being read from a cached state, or my previous overwrite failed silently (though it returned success).
I have cleaned the repo, tried in different sessions, made a new repo but I still get the same issue of Jules reverting the work it did after it finished it
I noticed Jules looping alot indeed. Probably failing to use tools? Also for some reason it decides to delete test files and than loops again with git add "testfile" for very long time. If chat is long, UI gets super laggy and hard to use.
The long chat is normal but even at the start, it updates/creates the new files needed> pauses the session before I say to run a test > when running the test the files are already reverted to their original state but Jules still has the memory it updated them
Good day. Is Jules available for gitlab?
It connects to Github - havent seen other options
Ah that’s a bummer.
is there any way to open jules "projects" during lag
Hm I also face lag 😞
nope, either wait till the page crashed afterwards and refreshed and open other "lightweight" projects
i dont get why my browser usage for jules stays above 10gb ram, is it because of the texts inside the chat ?
Maybe 🤔
also if the chats is too heavy then i cant even open it anymore, just stuck at ✅
I also face this problem
Like i create a task 4/1/26 today evening is start lagging I don't know why some time brouser fully stuck i cannot move like my device is freez then i create new task then no crash lagg etc..
Yahh Yahh i think very long time task cannot handle that's why doing this problems 😭
But but now my most favourite ai is gemini with jules 😍
ive moved on from using jules to antigravity because all of the problems
Gemini 3 is super i hope gemini 4 is can do everything like can do humans 😎
i still think theres gonna be 3.5 variants
Ohh I also want but i don't have pc/laptop etc.. i have only a 8 gb ram mobile phone cpu diamond City 8100
Ohh
I am glad 😌 talk with you 😊
I can fell like bros 😂
running jules on phone browser straight up crashed the page
atleast for me
Ohhh 😦 i think need a good phone 📱 my mobile phone price is this
332.90 dolar $
thats more than enough
Hmm
Hello guys. Can I import a picture in Whisk and to change the characteristics on it? Or can I see myself with someone vip ?
I've not seen this before but it is good to know 🙂 "Jules is experiencing heavy traffic. Session is queued and will start shortly."
dang ive only seen this in gemini web, specifically deep research mode
The task I have Jules working on could be considered deep research level i suppose 🤔
it is deep researching the "codebase" and browsing for documentation and relevant data i guess
Id assume so too 🙂 ive got it checking every corner of a platform we have developed together 🙂
mine kept failing for simple task😔
wishing you the best tho!😄
thanks 🙂 mine is failing now but ill give it some time and prod it again in a bit 🙂
Some of us are paying for the service, and expect our paid service to function.
I'm paying for Pro, and Jules is a privilege to have and I'm pretty sure is an extra thing we get to play with to test with Google 🙂 -- LABS.
please use #1238621511401734254 to submit any bugs you find!
also please be reminded of our first rule: Be compassionate and treat others with kindness and respect.
All kinds of harassment, hate speech and impersonation will not be tolerated. Keep conversations civil, welcoming, and on-topic. Post messages and content in good faith. this relates to all users including google staff :)
Be thankful you get nearly infinite tokens with Jules, and don't be surprised when that infinite token usage goes away.
Sorry Bee and Google Staff, some people don't appreciate what you've given us to use, I see it for what it is and am thankful you share this tool with us!
(Please don't kill the tool off like other google experiments in the past, I know this one comes at a cost due to infinite tokens!)
I didn't say I don't appreciate the product. I said it should work when you charge for it. That is not an outlandish claim or an insult, just a fact.
Look at it this way, anything in Google Labs is like a freebie addon to the paid plan, and can disappear any day without breaking any "paid" contract.
Think about how much Google "Loses" in terms of cost for how much they are giving us, Jules has infinite tokens, instead of limiting your monthly use, you get daily resets, and Jules can prompt Gemini Pro as many times as it wants without limits. That's a privilege.
Also that makes the paid plan we're using right now, the best value you can get over any other AI Model subscription 🙂
Look at it this way, they are charging me for a product that doesn't work right. That's the end of the discussion. Paid products aren't luxury, and they aren't doing it out of the goodness of their heart. They are doing it to make money, while not offering a functional service. THE END.
Are you paying for Jules, I think you're paying for Google AI, and it comes with LABS (Experiments, Betas, Alphas...) access.
That and the discounts Google does for their ai pro
I just hope our Jules dev team knows we appreciate their work, and see more friendly people, rather than users in their Discord who act like they're the boss. Keep up the good work guys!
Let's go into a new year and treat everyone kindly and with respect. Treat people like you'd want to be treated. Let's all try to bring some great energy into the world.
All of us are humans here with feelings, dreams, families, and obligations. I understand how it can be frustrating if a task fails.
But let's try to remember that all of us are people and we're all doing everything in our power to support you and triage.
"Jules is experiencing heavy traffic. Session is queued and will start shortly."
Is this common with Jules?
We've only had to do it a small handful of times. Upstream resource allocation more often than not
Are AI Ultra users in priority queue?
Is there not a chat for Antigravity or am I just missing it?
yes
another server not this one
@rigid flicker Can you link me? 🙏🏻
tyyy!
np :)
They do not have a channel for it but they talk about it there. There is another server for Gemini too
no we don't split it out like that but ultra users are being served Gemini 3
I'm on Pro and I'm pretty sure I'm using Gemini 3 🤔
Noted sir 👌
supp guys
Supp bro
Wsp
Supp bro
U good?
Good sir, u?
All good ty
supp ronaldo hehe
LOL
hello
Hello!
Hi is anyone using Docker in Jules to run the application stack? When I try to use docker I get the error below. Wanted to know if there is a step I am missing to get this working.
+ docker ps
permission denied while trying to connect to the Dock er daemon socket at unix:///var/run/docker.sock: Get "http://%2Fvar%2Frun%2Fdocker.sock/v1.51/containers/j son": dial unix /var/run/docker.sock: connect: permis sion denied
sudo ?
Oh!! could you please share session ID for this error and please try with using sudo for the command.
Great
Welcome 
Just wanted to say thanks to the Jules Team for having C++ Compiler automatically installed. I needed that lol 💖
hello
hello :)
Has anyone had good experience getting Jules to use libraries it might have no idea about or newer versions with completely different API? I'm struggling, I've tried loading the context by pointing it at the repo tag, the docs and telling it to investigate the code/ documentation.
Any pointers would be appreciated. 👍
Oh!! could you please share session ID
Has the connect to GitHub issue been fixed ?
For me it is fixed, does it not work for you
No it does not work
I noticed there is a delay between Jules on it's terminal and on the web browser, Github showed it's changes (i asked it to use a token) 45 seconds before it showed on Jules task
Hi, I hit a snag with cli . It looks like Jules debugs its calls using a version > 0.42 ( private release) while my local antigravity set up uses 0.42.
Nonetheless, if anyone has any good code / instructions on how to patch locally, very welcome . I can do it once off, but at scale my setup has trouble
( Jules is awesome btw, I love it! ) I’m a novice at orchestration so bringing it into my workflow has been a “fun” learning curve!
Guys what does "Session is in the queue" mean? Which queue
nvm I saw the message just now. I have never had it
Alright, I guess the secret is out thatJules is awesome.
Goodbye, Jules. You were a good idea, but poor execution. I offered to help you guys, but after several conversations, you decided to say you didn't need the help. So, this failure is all on you and the poor leadership at Google. I like most of Google's products, but they are all suffering more and more from amateur bugs that could easily be prevented. I could go on, but it's pointless. Google is too big and too complacent to listen. Plus, leadership is allowing the death spiral of adding new features to bad foundations. Guess it's time for me to just build my own stuff and not rely on buggy products anymore.
Fine, one more tip.... Don't have the browser poll several times per second to see if the task is ready to run. That will just melt your servers. Oh, wait, that's already happening!
man, that's some salt.
I would jump ship in two instances back, if I could run a Jules like service on my own hardware.
If I understand the architecture correctly, you can get most of what Jules does by creating a script that uses gemini cli, tie in a vector memory database, and sandbox the whole thing in a container.
I'm sure there's more secret sauce, but that's the majority of it I think.
its always about secret sauce)
Can we get a early new year gift 🥳
Hey @dreamy grotto Sorry to hear about the frustration. I’m sorry we didn't meet your expectations this time, could you please share which specific issues or workflows broke for you? Concrete examples help the team understand and improve where things fell short. Thanks in advance
I think he summed it up pretty well. You guys keep worrying about new features and your base product is unreliable at best.
I'd say that's the biggest problem Jules has right now, peformance and optimization needs more attention. We got Jules having alzheimer, getting errors, skipping testing and code reviewer, and a laggy UI at long sessions 🥲
Think Jules need available connect my drive or get is own project maps or folder's for big project's. Think Jules need some more compute feel like Jules performance less accurate lately. It just work long task before and create some fun project that was just for fun no real useful stuff but fun for science 🔭 still enjoy Jules but auto task mess me up and my old project and githubs task to old so I have update them
Hello maites, I remove my access to my GitHub but I still have this, there is an easy way to disable all of them ?
Thanks for sharing this. Please note that all of your queries have been captured and escalated during our weekly triages. We are prioritizing the stability issues
Sure, thanks for confirming
What about GitHub connection issues ? It has been 2 weeks and I'm still not able to connect to GitHub repos
Is agents skill coming Jules also as above? Where there any Roadmap available?
Be good to see a roadmap it might help build confidence
I have the Google Ultra Plan and for some reason Jules limits me to 15 tasks a day and Gemini 2.5. The Google Ultra plan comes with the max limits and it is telling me to sign up for a Google One account. Googe's Workspace Customer Service sent me to Google Cloud who wants to charge me another $29 for support when this is more of a consumer/prosumer product. I didn't see anyplace to request support in Jules, just the discord page.... Can anybody point me in the right direction?
Looks like you are having some issues, have you tried #1374125660859142254 ? As for the Customer support service wanting to charge you $29 that sounds like a scam to me..
Your name there is a Messi on it
Also make sure youre using the right google account, i have 5 gmail addresses and it can get confusing for me sometimes.
afaik, it's still loading......
or try to reload the page
ive tried 2 times, started out with stitch, and exported to jules
idk what happend........
both the times it just gets stuck, after one run, goes to review and gets stuck
i have not even tried to export a stich to jules
even tried a diff browser and cookie clear
Is everyone back
if the session is too heavy then yeah stuck for a long time and possibly the page crashed afterwards
so no fix? cause I'm genuinely confused what do I even do
Do i just break it down into half maybe, and give it less work at once
or maybe cli might work, ill try the cli
i guess no because its been happening since early days of jules
i got plenty of ram but still the page crashed
for me ill just create a new session in that repo in case the previous session loading starts to slow
It was my first ever session, lmao, i didnt even get started
just outputted like 10 screens from stitch to jules, and it just doesnt work even40 min later
lol bad luck ig, especially now other users also seeing the heavy traffic warning in jules
oof, well i got pro, so hopefully some elevation of priority or something
I'll try the cli, never even tried it, so worth a shot
first time with stitch too, ngl was fun
for sure
whats the prompt i need to get the steps to setup jules nicely with antigravity -- i installed the "Send to jules" extension, and it works, but im curious about antigravity, jules and the CLI mode, and if it's that good? (not asking you guys to tell me how to set it up, just what the right terminology i need is to find out!)
The command-spamming has gotten worse the last week. 95% of my sessions include massive command-spamming . Jules wastes time, tokens and energy for no reason.
i moved on to using antigravity solely because of this
I'm super curious, how many of you guys rely on Jules (a Google Labs experiment) to run a business and make money, and get really angry when a Labs Experiment that is not even a full product doesnt work properly yet?
hehe true
So true ^
And even if it was, its an ai, you can never rely fully on one for something that serious.
Yep, I'm just trying to understand some of the more "Salty" users who get really vocal and angry when the Jules dev team hasn't fixed the bugs they want them to fix (as if they are the boss) (and especially just after the 2 weeks christmas holidays) I'm wondering if they are just mad cause it impacts their business or something.
Is this normal for Jules? its been nearly 2 hours and its still trying to install this. I can't stop it
they just wont branch out to try other solutions because they got "paid perks" on that [specific] provider, "loyal" i might say
nope, at that point you either try to pause and resume back (and try to chat if possible) or just create a new session
Yeah telling it to stop didn't do anything, wasn't sure if this was normal. Will try to make a new session. Thanks 
Has anyone got an environment script that gets a working android emulator going, so Jules can run the unit tests?
tell it to retry i guess or pause/unpause
If it did any work, I usually would just pause it, Download the zip, make a new git Branch and push the changes from the zip to it, then start a new session. (Maybe there is a better way but that's what I do if it's stuck!)
i would just start a new session using the last changes made to the branch and copy paste back the prompt and execute
same concept lol
lol i thought you didnt connect your github account/give repo access to jules
The branches don't always show up immediately for me probably why I didn't bother
need to refresh the page. usually when you create a branch on gh while jules page is opened, it wont show up. either you have to reopen the choose branch menu in jules or refresh the page
jules commiting changes to branch is 5 -10 secs then it'll shows up instantly
a weird thing that i encountered was jules automatically creating a new branch even if i chose "main" branch but i appreciate it i guess
ok, we broke it 🙂
lol
It’s impressive how Jules can’t finish a simple task. I give up on this tool. It’s a shame because it had a lot of potential.
You def should not be charged 29$ for customer support service. Can you dm me your ultra account info and I can see what's going on.
yes this happens often!
Also that uarebad eth dude is spamming scams all over labs discord if you didnt see 😛
(every channel is getting hit ><)
testing out spam script ahh
yee its all gone nice
Roadmap any news for jan/feb?
Does getting google pro not give you jules pro if you use a vpn to be in a country that supports it? since mine dosent
Hello @old mauve Sorry for the inconvenience. Could you share the specific details of what went wrong? I'd like to check where it’s falling short and see if we can fix it
Hey @fleet roost The GitHub connection issue has actually been resolved, could you please recheck it on your end and let me know
I tracked down the issue. I was somehow signed into multiple GitHub accounts (work and personal). Revoking access for both and then re-setting up Jules fixed it.
hello
hi
Thanks for the confirmation. Please let me know if you need any other help
Jules getting rate limit again
obv..
Anyone thats knows something about this
VPN works for some others.. 🤫
google ai pro includes jules pro, using vpn maybe or maybe not yield some success because the region settings is also tied to the google account
Yeah i thought so damn we will see maybe later, sucks i got it just for that
Looks pretty quite here recently what happened?
Please share with the class 🙂
I have been looking at AI Studio, Jules, Stitch and AntiGrav, wondering how to make them play nice with each other. GitHub would also like a vote!
I'm about to just ask gemini, I've been busy all day didnt get a chance 😄
I am working on this right now, which is what brought me over here.
i just decided to setup antigravity on my laptop, so im setting everything up correctly (and will replicate to my desktop later)
🤑
Is it okay to use VPN?
Since I started down this path, I thought I may as well see it through. The logic components behind this box of tools that Google Labs keeps giving us.
**Blame Gemini for the project names, and not me!
oh thats cool! i see you already documented the cli stuff
Hello @bronze urchin It includes pro subscription. Using VPN is not reliable or guaranteed, it could cause connection or performance issues because your account region is overriding your VPN location.
Hey thanks for response, is changing account region enough if possible? If not can I see somewhere when Jules is available everywhere
Thanks to the dev-team for fixing the branches selection. It seems way better now 🙂
Is it possible to use Jules with multiple organization and personal GitHub repos?
Hlo all
Hi, guys. Does anyone know what is the problem?
Hello, if you're encountering errors and /or bugs or would like to leave a general feedback about an of our Google Labs products please post in the new #1238621511401734254 channel with as much detail as you can☺️
Yes, but it can be a bit funky until it settles down as it will try to auth on one account but not the other, and make you yell at your computer in frustration.
Multi-repo, np, I think I have 20+ currently.
--
Edit: Don't place production accounts here...this is a Labs product, and if you have watched Google software over the last 20 or so years, neat and interesting products can vanish with the drop of a hat.
Copy and paste (the entire session) to Gemini, Claude or GPT. Natively, each troubleshoots much faster than Jules or AI Studio.
It just says that it tries to pull from somewhere else but github, I donnow how to change that
Hi bro
Good Morning,
Normally I am working with projects based on C++ or something else, but now I am using Rust a bit more also with Jules. Is anyone here who has experience in using Jules for Rust? It has Rust installed and it works, but I want to pre-install (download) the crates and dependencies since this can take a long time when it needs to download over 5GB of Data per Session.
Apprechiate any help, have a nice day :)
Where are you from
What are you doing bro?
this solution below goes to other programming language too, you can refer to https://jules.google/docs/environment/
setup scripts save me precious times on not waiting for jules to install dependencies for each session
you can also rerun your script if you are changing/updating files
Thank you very much for the help. I apprechiate it!
Is Jules down or slow at the moment?
Had no issues today (UK) but last half hour nothing is progressing
slow as always
very slow here as well. the last couple of days i've seen "heavy traffic" warnings around this time, but nothing today, and it's even slower than usual.
It's been fine during the morning for me which was 5 hours ago but I've now two or three tasks just doing nothing. Going to log feedback + bug report
for me its just slower and still running but i didnt see any heavy traffic warning tho
for me it seems like 3 chats are completely stalled and 3 chats are progressing slowly. perhaps they're playing around with the concurrency limit, though i'm nowhere near 15.
for this, i used the setup environment feature for a while, but eventually i made a subfolder of jules scripts that contained one called setup-environment.sh and tell agents to run it at the start of every chat. similar effect, single source of truth, ~90% success rate. the biggest failure i'm seeing is that environment variables aren't always loaded into the environment, which i theorize isn't an error of the process but a jules inconsistency.
and after the schedule feature launched, i created a daily "jules environment steward" that iterates on the setup and other jules scripts, so it's less hassle than updating the jules built in feature everyday
I've had one running for an hour with limited updates. One errored. One I've just started with less work to see what happens
i've switched back over to AI studio and manual effort until i see the throughput pick back up, though perhaps i should try the vscode agent again. it's been a while.
yes, same concept too
as for the env variables, that issue only started like 2 months ago and yea its not consistent. one thing i always put in the prompt was "use the stored env variables/secret" or ill just put the env in the chat itself and it will saves for use on other session
nice one!
interesting. i've been thinking about putting in a fast-fail like "run printenv and if xxx isn't there, stop." but putting the variables in the chat is another direction to consider. probably security implications, but for a hobbyist it's likely not a big deal.
does the "Automatically find issues in your codebase" function even work ? i havent get any suggestions for a month already so i want to know if anyone have anything to say
yea its a risk to put the env in the chat, but only if the repo is private. jules really doesnt care about gitignore env and will commit the file changes alongside the env variables which sucks
had my multiple api keys revoked just because jules put them in a repo, even explicitly told not to
ive had enough so i did the testing myself
maybe some dead chats around a service restart is a more reasonable explanation for what i'm seeing. multiple chats are moving now, but the 3 that were stall are still stalled.
chat to retry i guess
Nothing since my last message has progressed. All stuck
I just began using jules, and have three tasks that appear to have started but are just sitting - is there a way to distinguish when jules may be stuck from when it may be slow or simply not working at the moment? e.g. I've had one in "pre-commit steps" for hours.
Good afternoon all, just saying hi 😄 - and have a nice weekend!
this chat is more active than usual because jules is less active than usual. you're not alone.
Today Jules seems to be totally broken. 3 chats with no answer or progress for few hours now.
Would be a classic move to have this release for the weekend. Anti-QA cycle
I don't think the dev team will fix that today since uh it's friday and welp weekend devs don't work on weekend
Please share the details in #1238621511401734254 channel
Jules need some love
Hi everyone
Are there any info regarding data privacy, does google use my code that jules has access to?
Jules has been completely unresponsive for me to since yesterday. I only have one active task…
no, only text/multimodal models can
@jaunty totemAs stated in #start-here・📍; currently, our team can support engagement in English; the discussion in this server will largely remain in English for now.
I knew i remembered that from somewhere
I mean the entire US wasn't working on last 2 weeks so meanwhile we are cooked
gotta be grateful cause google didn't reduce the limits for Jules like they did to Antigravity (yet)
Imagine the first thing you get as new year update is a nerf on your quota
Funny question but are you guys waiting for Jules to be fixed before the status is updated again or yall don't plan on fixing it again? https://status.jules.google/
even jules in free plan is considered unlimited, gotta love that offering
yea its a risk to put the env in the chat, but only if the repo is private. jules really doesnt care about gitignore env and will commit the file changes alongside the env variables which sucks
whats with the copy pasted text
UPDATE YOUR JULES UI! ITS LAGGY!!!!!!!!!!!
Some time is fell lagyy
Hi guys, are there any options to revert the environment to the original version? I messed up my env.
resetting/reverting option doesnt exist (yet), so you either overwrite your current "snapshotted" environment with a new setup scripts (and env secrets if needed) or just create a new repo which will give you the clean/fresh version
Got it, thx
And i don't understand the point of using environment variables instead of sending it on the chat if it will literally read it by using printenv
Hey @swift karma, any word on when Jules is going to stop acting up and work properly?
I'm also an user not an Dev 😂
putting env secrets in chat will result in higher success rate on letting jules tests out the builds, while using env variables option in the config repo had a massive inconsistencies between jules reading the envs and trying to use them
forcing it to read (using prompting) from the saved option is 50/50 chance
I thought you might not be a developer, but seeing you're so active and have so many badges, I thought you might have some insights! 😅
I am suffering with Jules also, don't worry
everyone suffering 😔
Let's hope there is a update in the next three days
If they didn't even fix the status page, imagine Jules itself
So I have this pipeline where GitHub PRs from Jules get revalidated by Gemini on GitHub Actions. So, it sends a prompt back to Jules. But, Jules was having some issues, and since GitHub Actions Gemini doesn't have any context, it just kept sending the same prompt over and over again. And Jules kept sending the same files, saying they were updated. And this loop continued the whole night for 6 Jules sessions, and GitHub Actions billed me $12 😭
true xD
Jules is making everyone poor, including google
Jules just stays there half hour generating an response while it burns Google's servers
Jules is greedy
and so on for 8 hours
long lasting issue, not solved for months
And when could it be fixed?
we will never know, potentially never will
And if you use it on the API, then everything should be ok and what are the prices for the API
it doesnt have pricing, basically cli and api are the same thing but without the ui
you call them to do task, session, approve/reject, etc... just like you would do in the jules website but in code instead
refer to both of this link below :
https://developers.google.com/jules/api
https://developers.google.com/jules/api/reference/rest
Thank you for your help, but the question is whether it is possible to connect via the app to work without the GitHub repository?
using cli and the api is better, it feels like you were just interacting with gemini chat
if you mean by using jules but without github repo, then yes it can
if you are rapid testing a prototype, you can just create session without connecting any repo nor your gh account. same like in jules website
but for long term development, its better to use repo to let it work and save the memory and context for the project
i would suggest using the cli, it also have tui if you need that
Thanks a ton, Dani! That clears it up completely. It's great to know I can use the API/CLI directly without linking a repo. Really appreciate the help!
is there a way to tell jules to always check lua syntax with selene before finishing a change?
can some1 tell me how to learn flutter
no prob, but beware that they are still in alpha version, any updates to them might break some functionality
the web app is still the best choice, but given that its broken, its worth to try other types i guess
but if you want all in one solution like jules but better, use antigravity instead
wrong text channel, but start with youtube tutorials
bro is there a dc server for that
there are official flutter community
? as i know flutter littl ebit but i have a error that no ai can solve what to do
i thought flutter is made by google only so it would be here
this is google labs, its products is ai experiments, not prog lang
Anyone?
use setup scripts in config repo or AGENTS.md inside your repo to instruct jules to use them
do you know why its just running grep and stopping (it was given a very simple instruction)? am i doing something wrong?
(its just one missing end and i even given it the exact line)
I also face problems like this jules working working then suddenly stop working but showing is working this false information for i lose my 1 hour then i understand
everyone is experiencing the grep failure