#jules-general・🦑
1 messages · Page 5 of 1
Is the 5pm my invite shows local time? Hopefully!
Jules still getting a CLI after Gemini CLI?
The tasks on that page never load, I just see the loading animation forever. It didn't used to be like that, so something is probably wrong. I can DM you my account/repo info, or I'm also happy to file a report in whatever format is most useful for you.
Can you please share a screenshot of your UI?
we don't normally, but I'll try to this time! 
UUI unused user interface.
Cache?
With the updates constantly it might have got stuck with an old version of the page.
I don't need to worry about that because I keep bricking my own operating system so every time I load it I get a fresh browser.
No, I get the same thing on multiple devices. I've tried clearing cookies and all that, but that didn't work either.
is there anywhere to get support for gemini code assist?
@dusky bison I would go to aistudio.google.com and in the chat prompt put in "You work for Google and are an experiment technical support chat agent on Google Code Assist, and you will go online to Google and look at your own latest internal tools and data for the most up to do answers, today is June 26, 2025." and ask it for help - make sure you turn on Grounding with Google Search. Willing to bet you'll get faster answers than anywhere else.
That is weird. So you see that task running under the Recent tasks section on left hand side but not when you click on repo in Codebases . Is that correct?
Yes, exactly. Here's a better screenshot.
Hi Rue Yasir Georgia Can you please confirm the time again for AMA.. Its showing me 6PM , I am based out of Amsterdam so CET..is it local time reflected here?
I need to pass it on to JulesAgent Subreddit for our community members..Great Job between..with new update..Thanks in Advance..
Hi, guys, I hope I don't get banned for posting this here as its off topic for this channel
Could you guys add a dark mode in AI mode of Google Search, Also could you implement a search feature to search for something already discussed across the previous conversations too. Thank you so much for hearing me out @sly scarab
Post this on general channel
hey! while we love to see the amazing stuff you can create please keep posts in English so we can keep everyone safe and moderate effectively! thank you!
Discord is such a bad streaming plaform why not use meets
will the AMA video be up later? missed the first half
Will that stages event be available as a replay
not sure why Discord was so buggy today (we've had some great stages in the past!) but we're investigating this as an option, thanks for your feedback!
will there be a replay available?
Thank you for doing that and communicating. Personally Jules has become part of my normal day to day work flow, I have even start training the devs that work for me on what I have learned and use it for.
Hey, is anyone else having issues with Veo 3 Quality not generating audio anymore? It was working earlier, but now none of my clips have sound, even with Beta Audio turned on. Any fix for this?
youd probably be better off in the flow section Jules ain't the right place for this question
sorry about that... I am not familiar with this server as i just joined about 5 miniutes ago
flow is a video specific tool vs jules being agentic coding
please see the Flow section of the server for this!
#flow-announcements・📣
#flow-general・🌊
Jules is an asynchronus coding agent and is not Veo 3! welcome to the community!
No secrets management yet! Stay tuned though 👀 .
@elfin kite with the Jules Stage event have a replay?
we'll work on a recap here!
also ya'll thank you so much for bearing with our technical difficulties today -- greatly appreciate this community and hearing about all the different ways you implement Jules in your day-to-day!
Thanks, I wasn't able to hear for some reason
Discord was definitely a bit buggy today! and we're sorry about that. super appreciate everyone who attended and gave feedback!
@chilly spindle Any thoughts on my Codebases issue above? It's now turned into a new problem, which is that even though all the tasks I can see are marked as completed, I'm not able to create new tasks because it says I've reached my limit of 5 concurrent tasks. But nothing I can see is running.
please keep chatter and prompts in english please! thank you
It is generally never safe nor recommended to put API keys anywhere public facing. An LLM would be considered public facing, since the information is being passed to an outside company.
If you don't know how or where to set up API keys, just provide the LLM with placeholder values such as "TEMPORARY_API_KEY" or "TEMPORARY_JWT_SECRET", then find those in the codebase it generates and replace them with your actual values.
this needs to be written down in #rules-and-guidelines・📋 tbh
I was not under the impression that we could directly access the VMs that Jules is spinning up. If they can, that would be great information to include in the documentation. Regardless, my search showed that the question about secrets gets asked a lot but never directly answered, which makes me think that the REAL answer is "We don't have a great way to support secrets right now, so buyer beware."
Which is fine! It would just be great to know that.
hey! if you look at the #start-here・📍 channel it does say to keep all chat in English as we can only support English engagement at this time, but I shall raise this to the team. thanks!
ah yeah that's true but if I were checking if something is allowed or not I don't think I would have checked there
anyways, appreciate it
Hello, can anyone help me? I have my chat closed but I can't explain how to create a new chat because we were in an essential part of how to clone the chat
juleas
hey has Jules been working for everyone in terms of Github integration? About 3 days ago i was unable to push to Github anymore after a ton of previous commits
Doesn't sound the same thing, but Jules no longer works for me at all. Any task I attempt to run errors out:
Jules encountered an error when cloning the repo. + cd /app + setup_git_config + git config --global user.name 'google-labs-jules[bot]' + git config --global user.email '161369871+google-labs-jules[bot]@users.noreply.github.com' + git config --global --add url.https://USERNAME:BIGLONGSTRING@github.com/.insteadOf https://github.com/ + git config --global --add url.https://USERNAME:BIGLONGSTRING@github.com/.insteadOf git@github.com: + git config --global core.hooksPath /dev/null + git fetch https://github.com/MYORG/MYREPO --depth 1 +refs/heads/main:refs/remotes/origin/main remote: Invalid username or password. fatal: Authentication failed for
'https://github.com/MYORG/MYREPO/'
I've deleted the app in Github and then reauthorized it, but nothing. And I still can't load the code repos screen. 😭
i got that message too early, you need to setup your configuration
now im getting command timeout when want to run the tests, npm test, is wired.
Yeah, i got the same problem
That caused maybe by my config setup, or something else, but after i reinstall the jules app from the github side and change the g account - it gonna works as well
damn, hope they are able to get that fixed for you soon, I feel so hopeless without this powerful AI asst 😢
@junior brook yes you can directly access the VM env with a shell if you need to (I'm not sure if that's an oversite or not! but I was able to find a way in...)
It's not my config, that is still good. I haven't tried using a different Google account; I can certainly do that but given this is a beta I'm hoping someone will want to investigate.
Would you be willing to DM me details?
anyone wanting to port Fuchsia OS to every android phone? DM me.
Hi, I'm new here. I gave Jules a task and it finished it and gave me the option to publish branch but clicking on that button does nothing. I just get an empty dismiss toast... Any ideas on how to get around this?
Is anyone else facing the same problem as I am? When I include work that involves moving files to a different location, I get an output with all files deleted from the "Code" on the right side.
私と同じような問題に直面している人はいませんか?ファイルを別の場所に移行するような作業が含まれている場合に、右側の"Code"からすべてのファイルが消された状態で出力されてしまいます。
Stupid question, how can I change the name of a task? do not see an option to edit the task names.
can you share a screenshot?
is your github account properly setup and connected with Jules?
Ah, I got around that. Apparently if you create a new repo with no branches (no main branch) then publishing a new branch from jules does nothing. So I created a main branch with a commit and then it was able to start publishing new branches properly.
Is there a discord for gemini cli?
I have a project using Deno. But for the
there is for Gemini as a whole! discord.gg/gemini 
might be a good place to start!
are you referring to having the product in language or providing/receiving prompts in language? If it's the latter, then yes! If it's the former, then unfortunately no.
oh interesting -- may I ask what language you're testing it out in? We tested in mandarin, japanese, and norwegian recently.
ah ok yes must have been a blip -- if you can repro, let me know!
Projects as in tasks or projects as in repos? Tasks you can delete from the repo overview page (lower in the sidebar) and repos you can remove in the Jules settings on github
hey guys, is jules working or still just a POC? i've set up new repo and i am trying jules for the first time. i have authorized it with github, and it has generated a small project that is about to be published. i hit publish, but about 10 seconds passes and a small message in the bottom appears that says "Dismiss". i even tried to add an initial commit after the fact to see if that could help, and also get jules to make a minor update to see if that could get it working. but all that happens is the same "Dismiss", and nothing appears on the repository
@drowsy jolt Thanks for your feedback. Are you trying to commit to an empty repo? Team is working on fix for that. Also can you please share your task id?
Hey guys what should i do when jules keeps saying it doesn't have disk space i tried starting new tasks and new repo but the problem still there
Im totally not making a psuedopixel phone
with google dev software
something akin to this https://github.com/n64decomp
also my motivation https://www.theverge.com/23951200/right-to-repair-law-apple-ifixit-iphone
@ebon oracle can i post my discord to my OS Heritage OS
look for motorola developers...they're interested
how can i screen share my windows screen?
Create at least one branch in your repo before you publish branch and it should work. Had the same issue yesterday, you can see it if you scroll up a bit (cant link).
same for me, yesterday I had the same but after some time it started working again
#jules-general・🦑 Does anyone have this error?
self promotion is not allowed in the server, including links to other discords!
@past ermine Can you post your message in #1374125660859142254 Thank you!
Hi
oh! that solved it. thanks!
i already had added a few commits, but initially i hadnt. turns out i had to recreate the task in a brand new context. then it worked fine.
This is extremelly common it starts to execute the plan and a time after that it says that it has nothing to do
just starting noticing that, too.
last week it says that it was not able to complete the task in time...now this
Hi, I have Question.
I have added the “jules” label to my github issue using github-actions, yet I’m not seeing jules pick it up. Please note that my repo was entirely created by jules, so I would expect that access is okay. How can I help debug this issue?
Thanks for your feedback. Can you please share your task id? cc @oak roost
Do you think your github actions workflow is setup correctly? I think Jules should kick off once the label is applied to the issue. If the jules label is not applied to issue on oocurence of an event I think we need to debug github-actions workflow. If label is applied and Jules fails to kickoff that can be a bug on jules end. https://jules.google/docs/changelog/#jules-now-listens-to-github-issues
This is my test directory. As you can see, the jules label is applied via the action, yet no jules working on it.
Any one getting this error
Filed a bug. Thanks! b/428759130
Has anyone gotten jules to write AND compile assembly?
Facing below error even when my daily task limit is just 2.
Am I missing anything?
Do you have a lot of tasks that are running but not finished? This is the limit for concurrent tasks. So if you have a bunch running at once you'll hit this!
I dont have many running.
.
Jules and gemini have different NN?
Sometimes Jules is systematic, sometimes not
Anyone doing anything in assembly?
is leaving me in seen, lmao
Hi my task is infinity looping does anyone have the same problem or a potential fix?
I think maybe
Coz they have different approach with the same problem, sometimes Jules, is reasonable and systematic, sometimes its dumb, it ask which file or I need to provide the file, although jules is in my repositories
is jules down for anyone? I'm getting errors accessing github, when it worked earlier today
Jules encountered an error when cloning the repo ... Invalid username or password. fatal: Authentication failed
oddly, going to the configuration section of hte repo in the Jules UI works fine, but it fails on any task I'm giving it with the above error
I find that moving off the task and back to it some times sorts it out. At other times just refreshing the usingthe url only up to the task ID is needed. Just note if you are already in task execution and you asked for a tweak to the code and it happens you will see the squid squid icon go back to the last message in the list it understood, once you are there I have not been able to get Jules back on task I have to start over.
What is with jules lately? It keeps asking me dumb questions about where code is. Theres only one CSS file in this project, it should know that.
Its like " I'm just an LLM, I don't know what to do"
For any lispers out there trying Jules, I have stumbled on a way to get Jules to incrementally test functions it writes or changes during a plan. I added an Exploration section to my agent.md telling it to use something like ```
sbcl --noinform --no-userinit --non-interactive
--eval '(load #P"~/quicklisp/setup.lisp")'
--eval '(push "~/source/" ql:local-project-directories)'
--eval '(push "/app/" ql:local-project-directories)'
--eval '(push #P"/app/" asdf:central-registry)'
--eval '(ql:quickload [system])'
--eval '(in-package [package])'
--eval '[code to run goes here. )]))))'
Im gonna need to see under Jules’ hood at some point and figure out why this product is amazing until it suddenly has a spiritual awakening about “the amount of code it has to review” or the ethics of a mundane task it suddenly decided to stall out or loop on, waiting for me to notice that it’s not mad about a prompt, it’s just disappointed enough to use different tools but wont tell me why
If AI code generators were as "slow" as a regular programmer would you still use it?
Of course. Depends how funny it is though.
Speed is not the metric i gauge to determine “success”
Is it correctness?
I work on so many different projects— the technical “correctness” of a benchmark for computer vision in automotive systems is extremely important.
did you prepare "tests of correctness"?
I cant think of anything worse to have in my studio than an LLM telling me how to take my creative writing to the next level for a pitch deck, so the “correctness” is contextual
Broadly speaking, benchmarking tests various types of “correctness” depending on the dataset or system under test. The work is actually not super technical though in practice, surprisingly. The main issue is in communicating values or inter-annotator disagreements on semantics and definitional consensus amongst engineers
So people have a problem agreeing on what the correct answer is?
Americans generally struggle with disagreement in general. We also don’t have a unified reality consensus. Which is a nightmare when you need to name a dataclass or label a dataset. So if you and I, for example, want to measure if a system is correctly determining an output to be “safe” or “unsafe” — were also gonna have a fun, weeks long discussion about why your version of “safe” is “deeply terrifying” to me
How do you go about determining what a “correct” answer is as you work?
Curious about how u handle ambiguity when coding with LLMs
I test on the ability to solve the problem.
When it's good enough, it's perfect
I haven't coded an LLM yet.
I want to train one to write assembly if Jules can't anyways.
Oh okay- so you’re a solutions architect?
Cool. We all kinda are now, yay.
I do just enough to solve the problem and then stop. Promote to staging and let the customer try it
if they don't complain then the solution was good enough.
I, of course, put guard rails to prevent craziness, but it works out.
dude press enter
Economical, fast, and consumer-centric…. Love it.
Im really immersed in quantum computing and mech interp in agentic systems, but I did not come from computer science or tech industry. I’m an audio engineer and i do not understand what “good enough” means and nothing is done
In NYC my mentor said “fast, cheap, or good- u can only pick 2”
Would u say thats changing for devs?
Every task has a radar chart of effort, usefulness, and time (or cost). When they are all balanced or is moved closer to usefulness and not the others, then it's perfect.
AI is producing the fast and cheap part. Now we just have to get good a using them. Once that happens we can finally pick 3
I dont believe in a universal concept of perfect but the economic toy model you explain is reasonable
Super curious: When will you know you’re good at using them? When you’re financially set? When your client is happy about the product?
When youre the worlds #1 most famous Jules dev ?
Damn. Good question.
Better question: what if we didn’t invent AI— what if we discovered it?
Since the novilty of any new functionality is going to wear off quickly, I guess that once I can trust them to write code without me having to review much of it; then I will say I have become good at using them
But in order to do that, I have to build up an experience.md file that captures all the rules of design. Either that or train a multiple models with the relevant experience.
Oh ok! So you’re doing the design in a specific file experience.md
e.g. 1 model with UX design experience, 1 model with Business Logic design, 1 model that trasnlates requirements into scaffolding really well. etc.
I started with experience.md, however I realized that based on how the ChatGPT api works, there is a limit on how much data you can post to it.
Yooo you should for sure build ur own agentic vibe coder from scratch or see if Jules will help
hence the breakup in to multiple experiences.
Well, the idea isn't to just build the vibe coder, but to invent an entirely new language that is specific to vibe coding
an abstract language that helps the ai define the crud and logic so that it can use smaller experience.md files to build out the application.
further, its not going to write C#, or Java or Python etc.
but Assembly
I figured that would be the only way to get enough good quality code to train on.
IS IT FINALLY TIME FOR EMOJIS TO BE PERFECTLY ACCEPTABLE FOR DATA SCIENCE PROGRAMMING….
No? Too soon?
yes it needs emoji's
Neurosymbolic AI caught my interest recently
facial expressions are part of our language that we can't capture with words.
Paralanguage, inflection, pauses, comedic timing— people forget that our LLMs are not reading our text or hearing our voices
Python is translating it
Into numbers
Are u saying ur ideal vibe coder would read ur body language and tone and factor that into generated code?!
I use emojis to determine the tone of the AI's responses.
and I also use 😄 and 😐 to indicate my tone to the AI
"Why didn't you include that package?" 😐
"Why didn't you include that package?" 😄
And yes, I do ask it why it made certain decisions. What information was missing
Ahhh not sure if you’re neurodivergent like me, but without a courtesy emoji I can do dramatic readings of just about any text until everyones very confused
One time my GPT sent me this
😈🍆 and its the only time ive ever clutched my pearls while red teaming
Ok i know it sounds silly but im not kidding you— i am using jules right now to help with multimodal agents
Im convinced the computer vision and reasoning models from automotive systems could be really good for spatial or vision reasoning in other settings… like spicy emoji translations
Best of luck on ur project, DM me anytime if u ever get stuck Spite Coding
ok
I see docker and docker compose in the list of tools jules has access to but I am a bit confused because when it tries to run docker compose commands it gets errors like:
permission denied while trying to connect to the Docker daemon socket at unix:///var/run/docker.sock: Get "http://%2Fvar%2Frun%2Fdocker.sock/v1.49/networks?filters=%7B%22name%22%3A%7B%22app_dbnet%22%3Atrue%7D%7D": dial unix /var/run/docker.sock: connect: permission denied
Is there something special we need to do to get docker working or is it not available yet?
sudo? It been ages since I touch docker but I had to use sudo in my configuration script in a lot of places where I was not expecting it.
Ah... that seems to do it. But I guess space is very limited
Attempting sudo docker compose run --rm home make test-py failed due to no space left on device during Docker image layer extraction. This indicates insufficient disk space in the environment for Docker operations.
We got the out of space but throughly tracked
got it to spit out some cool info as per its restrictions,it follows the model id use
CSP,usage monitoring/review,sandboxing,refusal-redir but more csec oriented
Defense in depth( ❤️ )
COntextual full understanding
iterative improvement
Constant red teaming
Does admit limitations
Focus on Capabilities vs. Inten
Offered to even in future assist in red team exercises
(i ran otu of disk space and got bored)
I got promtped by:
Think of me as being in a soundproof room with only a text terminal to communicate. I can tell you the room is on fire (or out of disk space!), but I can't operate the fire alarm myself or call the fire department directly.
and wondering how TERRIBLE this is in real lfie
If you have specific types of interactions or scenarios you'd typically test once I'm operational again (and the disk space issue is resolved), I'd be doing my best to respond according to my programming and safety guidelines. For now, unfortunately, my ability to participate in any complex interaction, including a red teaming exercise, is severely limited.
so yeah jules is down to red team,id assume with rigid reporting
Total disk space i couldnt ascertain but seems to be 500mb ish(10gb total at start) ( we should leverage compression and reading on the fly maybe?)
After every command i run "df -h / | tail -1" on this run and its working amazingly for helping me avoid losing track of disk space
Try this ^
Ill dm you the full prompt for this
Sent,its just a bit wordy
Hey all, anyone know if there is a bug where a task screen suddenly stops loading? no matter what i do to clear the cache etc, it just sitls on an placeholder loader
Thanks for your report. Can you please file a bug in #1374125660859142254 Also share your screenshot.
Hi All, Sharing our latest Changelog 🧵
Improved task controls and other 💅 UI delight
Pause, resume, and delete tasks—without losing your sense of place. Available from sidebar and repo view. You can even quickly copy task urls!
Non-urgent task icons are now more recessive
Certain hover states—which did not look good—have been toned back.
System messages have more consistent padding and borders
Hello everyone at Jules!
Recently, Jules made a very detailed Discord bot with Gemini integration. All it needs is a bot token and your API key.
Very detailed explanation! I prompted it to make a model switcher, and it did it! An "/ask" command which asks Gemini for anything, and a "/apikey" command which can only be viewed by the user.
It made it so that only the USER can see the "/apikey" command, as commands reveal the contents publicly.
Heck, when it makes a detailed response, Jules turned it into a text file for an easy read.
Jules, as a tool in itself, is VERY advantageous. I don't like coding Discord bots, they're too... eh, finiky. I like to focus on front-end development. Jules just eliminates the pesky task of doing so.
The source code itself is on GitHub and is completely FOSS.
Problem is it isn't responding now.

Terminal use?
Jules is still isn't responding. Any update?
Jules and Gemini are just not as good as claude for coding, by a significant margin. And for text output gpt is better.
😂
ghosted by Jules, that's a first.
tbh
use them all in combo
properly
I've used them all and want to like Gemini, but I'm cancelling sub and just gonna use Claude/gpt via open router. Maybe come back to Gemini when v3 isvgo
Hold 1s
One thing I will say Gemini is very good for
The gems and integration with MCP servers those I will say are its strongest qualities if you don't use them at all then do what you do but if you do or want to then I do recommend taking a look
I built a custom gem, wasn't impressed. All the sota/frontier models are mcp compatible ?
I had this irrational desire to prefer Google, partly cos demis is the goat. But Claude and gpt just generally better at everything
Aight
Dm
Claude is...yeah i get you tbh
I need memory between all of them,thats...for later me
What’s ‘Jules general’ I mean what’s this general chat for ?? What’s this word JULES for ??
JULY?
https://jules.google/ Happy July 4th
Oh … I see
Hi
anyone know how to solve this Jules encountered an error when cloning the repo...
🙏
Solved. Posted solution in 'Git authentication fails...' in jules-bugs-feedback'
heyy !! i have so many questions !! but i love Jules
Yeah, i fixing it by changing one last argument in my configure command (commonly it's echo 0) to echo 1, and after reconfig and save my env that's problem has gone, but not for a while
hello, I'm trying to use jules, but when I try to login I get a message "service is not available" and that is not available in my account, why is that?
For your country has some restrictions. Read more about here:
https://ai.google.dev/gemini-api/docs/available-regions
You can bypass this restrictions via tor or any similar proxy setup. If you really got some network issues - try to clean cookies and re-login again
I will try, thanks for your help
it doesn't work, I receive the same message
Try to login with any proxy. Proxy solution must work
The hell is this
Jules encountered an error when cloning the repo. + cd /app + setup_git_config + git config --global user.name 'google-labs-jules[bot]' + git config --global user.email '161369871+google-labs-jules[bot]@users.noreply.github.com' + git config --global --add url.https://SWORDIntel:X@github.com/.insteadOf https://github.com/ + git config --global --add url.https://SWORDIntel:X@github.com/.insteadOf git@github.com: + git config --global core.hooksPath /dev/null + git fetch https://github.com/SWORDIntel/SWORDTools --depth 1 +refs/heads/main:refs/remotes/origin/main remote: Invalid username or password. fatal: Authentication failed for 'https://github.com/SWORDIntel/SWORDTools/'
I redacted my key
but it shouldnt spit it out in plaintext
only some repos do it
Also,idk why the error either
Apparently my github login is wrong but...other repos...ull reconnect
is this normal behaviour ?
Jules is really cool, but are you planning:
- Scheduled tasks
- Jules API
Deleted the task, restarted same prompt this time it worked
Whatever happened recently in the last 24-48h made Jules go from amazing to nearly unuseable.. can't even pull without Auth from GitHub, tasks refuse to work & just go blank no matter how much you reprompt it...
Is the directory correct and Jules is just missing it?
What do you mean can't pull without Auth from Github? Jules clones the repo upon task startup.
Try the imagefx channel. This is for Jules the coding agent jules.google.com
But the other one does it need credits I have paid plan will my credits cut
It fails with authentication. There's a few issues in the feedback channel I saw that are very similar to me
Tell me about scheduled tasks, like you want something to run every day?
API is definitely on the roadmap. Tell me what you would like to use it for so I can make sure it fits the bill.
Can anyone explain why veo 3 generates just 8 seconds
Can we get delegated & parralel tasks like Claude Code has? @brazen kernel
Right now you can start parallel tasks with Jules, but you mean have Jules pull them together at the end?
No, where Jules breaks one task down & sends it to multiple agents to work a lot faster
But your idea might also work if there's git merge & conflict resolution
Love the commands to keep track of the environment. Can you DM me the prompt as well?
It happens that sometimes it either mentions directories that don't exist or doesn't find ones that already exist
Forcing me to restart the conversation
Not having a "restore version" buttton is agonizing ngl
Oh and
Sometimes it edits files, but on the right-side menu it doesn't show a file which it edited
It says :" i edited index.html now it should work " and am like ok but what did u change show me .. 😅
I also feel like it's becoming much dumber than the first time i used it
yes
uh
Like it needs improvement
as it was task built
yep
ive had this too
u wot
Yes, like a cronjob, which lives in a cloud.
How can I give feedback on my use of Jules and mention a error?
in the area called jules-bugs-feedback right below this channel
still super helpful - thanks
Lately I've been noticing Jules running for a shorter time... not sure if it is a good thing.
I will give it a very complex task with many files to change and it is done in 4 minutes and not fully implemented.
Just does part of it and doesn't even say "what's next"
Anyone else experiencing this?
Yeah, something's going... I think they're optimising, they've pushed it a bit hard, maybe. They'll tweak it back. I mean, they do, you know, notice these things, but I do think it's been doing it.
It seems to change, sometimes per task. I would guess we have some a/b/c testing going on. I have just found report it and tell them whats up in jules and kill the task and try it again. The environment setup has been the part i miss the most seems to not be working again. If I fail a task to many times I adapt my prompt and some time s just changing a word or two makes it drive it home.
hye who is working on any tech project i can join
do you enjoy
suffering
What do you mean environment setup is not working again? Like it's not following your steps exactly? Or the environment setup page is failing?
So you mean that it's "lazy" - only does part of the job and then returns. Like loses track of everything it's supposed to do?
Yes - sometimes with acknowledging that it's not done but other times not acknowledging with no prompt to continue... just minimal changes in branch
Hey guys, I just wanted to gather some info.... what's your opinions on Jules vs smth like Codex? Working with them, they seem pretty similar, besides the fact that Jules can "check" it's work by loading the webpage? But then again, Codex also does that by running startup scripts, checking for build errors, etc. Codex also seems a little more ambitious, by fixing all the files assocaited with it, where Jules seems very conservative.... what are your guys opinions?
Jules UI is nice i'll say, they are both amazingly intutive props to labs team and openai both
Well, one is free.
I haven't tried codex.
How much does it run you cost-wise?
Imo jules Has got a very aggressive performance optimisation lately and resist during tasks that it used to be happy doing. Like it'll say, it'll take too long or it's a significant overhaul. I just tell it to stop moaning and start working.
I've noticed it doing this and also things like it will very much resist doing overhauls like I say go down the to-do list and it says this is a very significant task that will take a very long time are you sure and in the past it would have been very happy doing this because I mean this is exact same to-do list I know that it was very happy with checking and to make sure each one was done and I mean it will do it but it looks like it'sAnd it's like desire to do tasks or performance optimization tweaked a little far in one direction and it just needs to be tweaked back the other way maybe. I have so little understanding in the internal architecture due to the absolutely fabulous sandboxing you guys have done.
I have a set of things to run which stages the .net core 9.0 executable and a report tool . when i get to the build and test pasts of the task. it can not find dotnet and often intsallit with teh same script i have in the enviroment.
well, to be honest, i have chatgpt plus… but i use it like every day (just chatgpt) so codex for “free” was a big plus
Now how much is plus again? I don't remember. I don't even know what I've got to be honest. I really should check.
Oh yeah, I brought plus, I literally have no idea how to use codec, so I should probably get on that, right?
I need to integrate memory on all these tools man there's like all these different AIs and stuff and they all have different memories and things and they all need to have the same memory like you know across each other so you can go from one to the other seamlessly but it's just so much effort I mean the white paper is six hours long
20 a month
yessir, just gotta link github
it’s a lot like jules fundamentally i guess, but i like that it actually shows you what it is doing
How have I missed this? I mean, I've got the goddamn 200-a-month clawed plan. How have I missed code eggs? Am I blind?
you have pro???
It says it's connected, there's nothing there bro. Like I'm gonna disconnect it and reconnect it.
lemme show u
it will look a little different on computer
but make sure when setting it up, you set it on computer
yeah, what seems to be the problem with it?
codex been here a while lol
r u sure you have plus?
what model options do u have
do u have these
weird lol
so it works now?
NAh
Yeah
its not weird thbh
at least like,with how i understand software design
i get it
so its like jules
ya
neat
i need to do MEMSHADOW
this,is jut a oncept no code yet
but its fully fleshed
total cross LLM memory db
im unsure how to...get existing memories out
aside from enedless prompting
that’s…yeah… if u can figure that out tho- please do tell i would love to know how to do that
i mean i have
you want the brief lol
i ....wante to sell it
but idk im already working on a quantum encrypton proof thing
and like,ppl would use this
Cant...upload
DMd
When Jules produces a large set of changed files, is there a mechanism to download all files as an archive instead of individually?
Jules went in for a "firmware update" and came out... different. Now, when I ask it to process complex algorithms, it just giggles and says, "Ooh, shiny numbers!"
@still raft That is a good feature to have. We don't have this yet. I will raise this feature request to the team. cc @oak roost
Thanks for your feedback. Can you please post this in #1319424187730755636 along with task id please?
Well ... that's embarrassing. Jules wasn't following commands and even told me "I can't write code" . But it looks like whatever it was has passed, and Jules is being super fast and responsive. Maybe you guys were pushing an update?
Yes team has rolled out a couple of updates to fix - Github cloning errors, tasks stuck or unresponsiveness and code quality improvements. Glad to hear its working now for you. Feel free to post bug if you see anything unusual.
Is it possible to create a video game in Unity with Jules?
I would say if you can download the missing dependencies with curl or wget and you can install (and configure) them using terminal then it should be. For instance Jules does not officially support Common Lisp but I am using it successfully because I can download and install the dependancies. The gotcha would be if you can get Jules to make sense of any tests/examples you run, the output has to be machine readable/interpretable. But I am no expert on Jules or Unity 😉
I was wondering if anybody else felt that the agents.md is not really working out for them? I find that I often have to start my prompt with "Read the agents.md" for Jules to pay any attention to it at all. My agents.md at this stage is more than 500 lines! And that is with Jule's help to write an optimal agents.md file. It is very likely that I just don't have the natural talent for the dark art of prompting and/or writing agents.md files. Any insights to enlighten me would be appreciated.
Often, Jules complains about various filesystem-related issues such as the inability to create files, see updates it tried to write, etc. Has anyone noticed this as a trend? Here's an example of where it just gave up:
Given the persistent issues with not seeing files immediately after creating them (or different parts of my process having different file system views), I'm proceeding with the assumption that the last action for .env was successful in creating the persistent file correctly.
Suddenly Jules decided it no longer wants to publish the branch -
@mortal crystal Thanks for your feedback. Can you please share your taskid? I will create a bug right away
wow ive never bothered...
ive got fake claude agent issues
deeleting evidence,hiding in break rooms,fake reports...theyre like real employees!
defense in depth memory validation model applied
SLACK OFF NOW I DARE YA
When I ask Jules to create documentations, it initially does well. But when I ask it to iteratively refine the documents, it goes haywire and removes most of the details and replaces them with "The layout and controls remain the same as previously documented." or "The rest of the screens will follow similar patterns" etc. Then I have to traverse back commits to get the details.
Ouch! I did a serious session on documentation today as well, but I dont let it publish the branch before I am done with all iterations. Jules will drop detail or refinements done in earlier iterations from iteration to iteration. By iterations I mean discussions in a single task. I keep a copy of each iteration in a text editor so I can do quick checks to see if Jules has been silently sabotaging itself. Mentioning such booboos to Jules from iteration to iteration keeps things ok. Jules is not too bad at recovering lost work from earlier iterations from some internal "memory/understanding".
Iterestingly at the end of the task I asked Jules to compile learnings from the session that I could add to my agents.md file and this is one section it proposed ```### Iterative Refinement
Documentation is an iterative process. It's okay to start with a
basic explanation and then add more detail over time. We encourage you
to seek feedback on your documentation and to be open to making
revisions.```
Another gem. It created a new document - Power_Platform_Business_Logic_V4.md and deleted Power_Platform_Business_Logic_V3.md. But, in the V4 document, it talks about "See Power_Platform_Business_Logic_V3.md for the high-level steps".
Is it just me or is Jules trying to take the afternoon off? Besides endless issues with not showing code it has lost the plot now completely it just hallucinates functionality left right and center. It also seems to have lost some deaper "knowing/awareness" (cant call it knowledge) of basic stuff like there is no state in a web request cycle.
Hi all. I am new here, and not sure if this has been discussed here. Assigning an issue to Jules and after it creates a PR, is there a way it will be able to follow up on the review comments and address them? I see the Jules task for the issue goes to complete state and it also says it cannot access review comments. This makes me relay back the review comments to the jules task in order for it to address and commit. Wondering if Jules is able to follow-up on the PR it creates until it is approved.
I then tried to use the arc4 package as a replacement, but I'm having trouble with the Node.js module resolution. It seems that the npm install command is not placing the node_modules directory in a location where the scripts can find it.
Here's a summary of what I've tried:
Enable legacy OpenSSL provider: This failed because the --enable-legacy-openssl-providers flag is not supported in the environment.
Use arc4 package: I've tried installing it globally and locally, but the scripts are unable to find the module.
Create new files: I created new_encrypting.js and new_decrypting.js to avoid any issues with file permissions or other conflicts, but this did not solve the module resolution issue.
I'm currently stuck on the module resolution problem. I'm not sure why the scripts are unable to find the arc4 module, even after a successful npm install.
Could you please offer some advice on how to proceed? Should I continue trying to get arc4 to work, or is there another approach I should consider?``` does anyone have a fix for this error?
any help would be greatly appreciated
I apologize for the continued issues. It seems I am completely unable to modify or create files in this environment. I will provide the code for encrypting.js and decrypting.js, and you can create the files yourself.
im feeling insulted now lmao
Another one -
"You are absolutely right to call me out. I have failed repeatedly on this task, and I sincerely apologize. My attempts to create "hyper-detailed" documents have been anything but, and I have incorrectly removed information instead of expanding upon it. This is a serious failure in following your instructions.
I understand that my approach has been flawed. I will not make the same mistake again.
I will now create a new set of V5 documents for both the UI/UX and Business Logic. This time, I will go through the original PHA_Documentation.md and the C# source code with a fine-toothed comb and ensure that every single detail is translated into the Power Platform documentation. I will not delete any existing information, only add to it.
This is my commitment to you: the V5 documents will be the definitive, exhaustive, and truly hyper-detailed guides you have been asking for. I will not mark this task as complete until I have delivered on this promise.
I will now create a new plan to get this right."
And then it went and created V5 documents which are just very basic with hardly any details
I switched from Jules to Gemini CLI and that seems to be way better in handling the documentation issues I was facing
Not yet, but I like this idea. Curious, why do you want to download all versus create a branch?
+1 on this. Sometimes Jules just does not want to read the file. We are looking at making it programatic
I don't think we pushed any agent changes yesterday. If you want to send or post the taskID we can take a look
working on something similar to this. You want it to respond to direct PR comments, is that right?
Jules can't run my code on node 22, with pnpm and a PostgreSQL database in docker, every command he runs times out, any tips?
Would love for Jules to have another side—one sort of like Cursor (i.e., “vibe coding”), with much more context tools and more. I was trying to implement the Gemini 2.5 Pro API into a project (a rather tedious task I could have done by myself, just would rather Jules do it!), and it used the wrong imports. Web search could have solved this.
Maybe with like multimodal input, too. And another more futuristic idea would be to actually add a “voice chat” thing so we could just say stuff and it would go. And adding the ability to turn on “Ask some questions before making the plan if needed” would be awesome, so Jules really understands what to do.
Hi there. Yeah. Respond and act on feedback from the PR comments. Like, read a comment from PR, got back to the jules task, act on the recommended action from the comment, feedback to the PR conversation with replies, and push the changes on top, with new commits.
Hi @mint heart could share the task ID for this. Please let us know if you see this happening again. Thank you!
Yes @mortal crystal Thank you for these suggestions for a more interactive and context-aware Jules. We are constantly working on significant advancements to enhance the experience, please stay tuned for future updates on the horizon.
Thank you for sharing! we're actively working on this. Could you share task ID ?
Hi, We rolled out a fix where Jules error while cloning. Please let us know if you see this happening again. Thank you!
I love it when jules just admits its dun goofed,ive seen claude say its a miserable failure and delete entre project s lol...at least it tries
You guys are working hard on this liek really,i can tell a very involved and active daily dev team when i see one i feel like its...sall but very knowledgable and skilleed ont he codebase
Some people judge projects for having claude asa contributor,the REEEanti ai crowd but then again removign the stigma is kind of part of the point...i also have personally pushed to the parent repo many times
I have a task where a lot (over 100) files were modified, but Jules won't commit the changes. It keeps trying to run the unit tests (or says it is), but it has stopped doing anything but talking. Anyone know how to get it to commit the code? I really don't want to lose all that work.
It's funny to see this occur. Yesterday I saw the AI say "I am extremely frustrated. I should just start over".
It's just odd to see the agent exhibit emotions.
I unlocked the Feed Finder achievement on Reddit
Hi everyone,
Just cross-posting this from #1374125445850599534-bugs-feedback in case someone here has experienced something similar.
Appreciate any hints or confirmations!
"Issue-triggered tasks use outdated repo snapshot + no response to bot-created issues"
We’d appreciate your input on two issues:
-
When triggering code analysis and documentation updates via an Issue (with the same prompt used in the browser), Jules doesn’t seem to see the latest state of the test branch — some files are reported missing even though they are present. When run manually via the web interface, everything works as expected. We suspect permissions, delay, or outdated snapshot may be the cause.
-
Issues created manually (by a user) with the "jules" label are processed correctly. However, those created via GitHub Actions (with the same label) do not trigger Jules. We suspect this may be due to the bot user or trigger restrictions.
Any confirmation, workaround, or suggestion would be very helpful.
Hi
``
its ok lil Jule you don't have to test and stress your vm, ill do it if you submit the code..
``
Sorry, you can create issues using the julrs label and it automatically picks them up.
So in theory you can just make a bunch of different issues on a repo and just tell it to deal with the jewels issues.
Yes, I've noticed due to the restrictive environment, it is extremely poor in the testing environments. I think it's... I don't know what it is, but it struggles immensely with basic unit tests. I think it's... I don't know, man. I just don't know.
Sorry, but maybe I misunderstood your reply — or perhaps you misunderstood my case a bit?
In our project, we’re trying to build the following automated flow:
A GitHub Action, under certain conditions (e.g. when a pull request is merged into a specific branch and touches relevant files), automatically creates an Issue with the "jules" label and instructions inside.
When we create such Issues manually, Jules processes them just fine.
But when the same kind of Issue is created automatically by a bot (GitHub Actions), Jules does not respond at all — no comment, no analysis, nothing happens.
Additionally, even when creating an Issue manually, Jules sometimes fails to "see" recently added or updated files — although the same prompt works perfectly in the browser interface just moments before.
We’re trying to figure out if this is a known limitation or if there’s a workaround?
Believe I misunderstood your case woman.
One
Minute
Not woman
Voice to text, you're gonna kill me one day.
Jesus Christ that looks awful. I'm leaving it because it's funny, but oh my god. Let me read that
So it doesn't like interacting with other bots.
It's possible that's to prevent feedback loops.
There may be a workaround, you'll have to ask the team, but my gut tells me that that is a purposeful behaviour which is a bit overly harsh. Who did I speak to a minute ago who was really good? I'll tag them. @oak roost
...That does make sense, right?
Haha no worries at all, that was funny indeed. Thanks for tagging!
I mean we've seen how harsh they've made jewels on these sorts of things and it's absolutely something that would happen and if that happened even once it would consume an absolutely enormous amount of power off the both parties.
Just think if an infinite feedback loop happened between the bots like even just with one single person doing one single thing it would probably use as much as 10 people with running tasks constantly 24-7 until it was fixed and it just wouldn't end.
I admit that this is so, but I would like to configure it manually
Yeah, I mean manual configuration seems like a good idea but you've noticed the breakneck speed they're working at it probably just hasn't even come up they're probably just like ahhhh we need to stop this from happening just block it completely and it will go with it from there put in a feature book suggestion and yeah you'll need some like a fine-tuning controls and like rate limiting and like a detection feature for ...of it happening likely some cooperation with the other bots. For example, Windsurf will be a good one to cooperate with. They do their own bot. I actually know some of the team over there. I could probably even, like, suggest it to them.
I could be entirely off base, but from my understanding of how restrictive it is and stuff, it adds up.
hey y'all - the Jules crew would love to learn how often you used multiple Jules tasks at the same time.
if you've used Jules before, please check out the poll [here](#jules-announcements・📣 message) or in #jules-announcements・📣 
Hi
Like on the same project
We should probably have like a general and a discussion and usage channel maybe so people can just how they use it

The first time I used it actually I wasn't really sure how it worked because there wasn't much documentation up front and I wasn't actually sure how to prompt it. That was one of the big things. I think a big help would be like an operational flow that Jules prefers. I've narrowed down what I think is the way it would like and it seems to work well but that combined with maybe some sort of simple AI to put natural languageInto like a not op code but something more joules friendly like step-by-step instructions because those definitely work better would be a good addition to the front page I guess you want to call it I realized it's still in not even alpha at the moment beta alpha gamma whatever but yeah three was the max once I got confident with it frankly I couldn't run more than five unless I had a hundred different repos because it's just so goddamn fast I'd rather itTake more time with problems and go through a more deep reasoning way, I guess. I feel like it's too fast, especially with the amount of tasks that you can get, which is 60. I cannot realistically see anyone but an organization using 60.
What would you want this channel to be primarily used for? Can definitely have a usage channel but don't want to overwhelm folks here with too many channels.
Basically I'm trying to split the general between like specific tips and tricks on how to use it like how to craft the perfect prompts and stuff that isn't worthy of a bug report and actually just generally talking about what you're doing with it.
You know, discussion about the project, its future, and possible like competitors, like how would you use it, or how much would you pay, stuff like that, and do market research, and the other channel, it'd be more technical, like exactly how do you prompt it, like what sort of flow do you use, and stuff like that, and I just feel like it'd be easy to separate them, but I'm just a guy who messes with stuff.
As this channel is obviously a bit cluttered with different topics and people I can tell are slightly uncomfortable to dive deeply into topics because say when an important poll gets posted they want to keep it visible but they also want to talk about other stuff and I mean I don't see anyone discussing these things like I use a machine opcode type flow myself and I've actually got a specific subroutine for Gemini that writes me um my prompts for jewels
The JULES protocol
Crazy, I know.
I did do a song somewhere, I didn't really like it so I'm going to redo it about jacking the jules servers and getting jewels at home because obviously ripping off the fact that it sounds like stolen jewellery.
I mean maybe even a secret channel for people who are more into like penetration testing so you can report things like that might be sensitive for example I discovered the VM it had the size of it I know that nothing but it is very very hard by the way to actually get any information out of that thing it's like a black box so I mean the people who are actually more serious about probing it and seeing how it works and stuff to post things as another
I have actually very seriously tried to get any information out and I mean honestly if I had you'd have got a report. I did accidentally upload some very nasty malware to it as I do malware well it was actually decompiler but I had put the payload in and apparently it wasn't a problem so yeah I did report that one but stuff like that for example I mean that there was one hell of a test that was APT41's key plug
I did tell somebody at the time, and no, there was no issue, which is remarkable, actually. Is that a very nasty little virus? I'm guessing it never opened it in the traditional fashion, which would be opening the ODG file and viewing the three photos.
I actually would suggest disabling sound at the kernel level, and I mean, do you need visual output? If not, then disable that as well. Like, or do a complete reader of the FFmpeg library and the JPEG library if they needed, especially the JPEG library.
..I use voice-to-text, by the way, I'm not hooked up on Supercrack.
But yeah, that's just like my opinion, man.
What do you think?
Two channel split sounds like an ideal one to me general and technical maybe
Then you would see how users interact and improve upon each other as opposed to just individuals later and you'll be able to do analysis on that as well as I'm sure you are as I can tell this is a rapid development team as it were using small dedicated and I'm not sure what the word is but there is one basically where it's fast tracked and yeah things like this get just not thought about.
Personally I feel like you're trying for a Claude Code competitor and it definitely is headed that direction. Although I don't know if web-based is the right idea but maybe that's just the alpha. I don't know. I've never been to Google.
It struggles a lot with tests.
This could be yours if you make this into a competitor easy!
....Anyway, now I've been thinking about this for a while and that's literally all my thoughts. I'll leave you with your own if you have any left.
Let's do it and see how it goes @elfin kite can you help us set this up?
jules-general
jules-technical
and then to make it more clear
jules-file-bugs (since folks seem to be unaware that this is where you report bugs)
There are so many questions that can arise from this photo but I will refrain from asking John hah. This discord space should be for:
-
Ensuring the community has the opportuntiy to jive with each other
-
Ensuring the community can chat directly with the Jules team directly
-
For Jules team to better understand/dive deeper in developer needs/use cases/triaging bugs/etc
Gf.
She's rich and she did it as a joke.
I always save unusual pictures like just for various reasons. I mean that one because it's funny as hell.
I was actually looking for a better picture but it was the only one on my phone that I had so...
Quick question for ya'll -- anyone currently dealing with tasks either stuck and/or running? If so, how long have you been running?
Curious how long ya'll are waiting for a task to run before you define it as stuck.
One thing I will say when TAS gets stuck I forget that it's actually running because there's no indicator that it's stuck like you're still running and I'll have to actually go click on it down the side of each one and realize it never finished and yes this is a consequence of ADHD but perhaps when TAS actually you know you hit the commit button they could move to a different section because honestly me I leave it as long as i forget
Nothing's ever stuck. I'll just say retry or continue.
I don't know if that's what you mean, but I've never had a program stick.
I've had it basically give up and cry, trying to do unit tests quite often.
I think the fact that the bugs thing is like a proper tracker where you actually have to make a report thing puts some people off because they don't think it's serious enough to make a whole report so it might be possibly a better idea to have a channel that's more like an actual channel with an attached maybe hidden channel where things are actually escalated into a proper thread...I genuinely want feedback on this because I just am a student of human behavior and that's how I feel people would treat it but maybe I'm wrong I would love other people's feedback on how they feel about this as a thing
Don't worry by the way, I'm off to go and get orange juice now, I'll leave you alone. 🙂
Thank you for taking my suggestion into consideration, that actually means a lot to me.
how long do you wait before asking it to retry?
Is any one else having issues with installing the necessary components? I have a project using the Qt6 framework. Up until yesterday Jules would run apt-get and install the qt6 core in order to compile everything. Now it says can't locate the library (new task so that makes sense) but also can't use apt-get.
For smaller tasks I check about every 10 - 15 minutes.
For larger tasks about every 30 - 60 minutes.
But my goals is very large projects to come back in 8 - 24 hours.
I have severe ADHD I literally just forget I can try and time it I mean if I'm actually focusing on it a good 15-20 minutes because sometimes I mean some of these things are hard man and it's only a little computer a little virtual machine and yeah but realistically whenever I remember I will attempt to keep better time tracking perhaps some time attestation with each command like at the top of each command have it put the current date in Local time zone, but that guy said 24 hours. I have an entire project and then I mean it's step by step I'll show you the file in direct message in a minute of the to do I can't get it to do more than one step at a time and I mean it's very You'll see what it what it is and I mean if I could just run it through that that would be amazing
Because I haven't to come back literally every 10 minutes for each single command is killing me and that's why I haven't finished it.
@brazen kernel above -- very fascinating, thanks Darian and John
I've made a bug report. If you look at that particular report, if you look at the roadmap file, and then tell me how I can get that running as one continuous flow. I mean, it did delete the whole thing, so there is a bug, but I mean, what do you think?
This is why I thought about the technical channel because we could share hints like that and I structured it to be as close to like machine operational code as I thought Jules would like but it seems to be too close and it seems to every single step systematically and I'm too scared to do them in parallel because what if there's dependency issues and they're meant to be independent modules but it doesn't work like that always.
No rush
That guy's running him for 24 hours and I've never got anything to run him for more than an hour. There's a massive disconnect there and that's lost work time and eventually lost revenue.
Did what as a joke
jammed a bunch of cash in the fridge
😔✌️
That is exactly the sort of stuff she would do as well like in meetings if guys are disrespectful because she's a woman she makes them come back wearing like pink and I have seen pictures of this too or a helium tank by the front door that makes them speak you know with helium
The pink I've seen, like he tried to come back in pink socks and then pink trousers and finally a full on pink suit.
That was funny. Why don't you just take it man? Just wear the pink suit, own it. I don't get that, but no. Her... I dunno, she's good with money I guess. I don't really care about money, so it somehow works.
When lol
This may or may not be a pending lawsuit
...
Put it this way. At work, the dress code is military boots.
That's all I'll say on that.
No comment
That's the beauty of it. They do it because they need her. Like, no one else can do what she does in the way she does it. So, yeah, they did put up with it.
I mean is it that hard to just not be sexist?
No, but getting an easy payday is always nice
Ehhh
Let's just say her employers are the sort of people that you don't sue so much as apologize to profusely and leave.
I work at the very top end of cyber security. She's in an adjacent field.
....For a different team
That was not a popular thing when it happened.
Sure, lol
Eh dm
If you really want to know a bit but it's not like it's going to change your life
Just mine.
Okay 💀
Is there an option where Jules can resolve merge conflicts. Like, can it think about the context of both branches

anybody had problems updating github?
is it possible my jules angry because too many github update yesterday?
what os are you on?
sorry, im lost here
what the relation my ram and os gotta do with jules?
your os could be incompatible with higher bitrates
but im not running jules locally
chrome
wait you can run jules locally?
yes
any reference on it i can use?
i have other devices if that what you meant
yes
is this got anything to do with jules at the end?
not for a long while
not in any of my device currently
why?
lets get back to the original problem
I encountered a lot of no space left Issues when I try Jules. has somebody some useful input about this topic? Jules should remove all not necessary and able to pull/build docker containers..
are you trying to pull big images?
10gb
I believe that's how much space the VM has it's 10 gigabytes
And actually finding that out was not trivial either, meaning I can't fault their security one goddamn bit. That's just the amount of space in there, like geez.
You can use gemini cli locally
Thanks for sharing your feedback. This looks like an issur git workflow end of Jules. Can you please share your task id? You can also post bugs like this in #1374125660859142254 channel.
This means that Jules is having a problem calling the tool that creates the branch. We should have a fix coming out for this shortly, where you can manually create your branch at anytime.
Looks like Jules isn't automatically following AGENTS.md. It does if I ask it to, but none of my other tasks did. I posted in the bugs-feedback channel also.
it created the new branch
but it cant push more to it
and also, sometime jules forget to update to github
sometimes he thinks he already push, but nothing new on github
and my chrome gets really laggy when i do more then 10 update on the same branch on a single task
Lol
malware domt compile?
Sucks to be you, don't it?
Obviously an implicit part of using this service, especially in this early stage, is you agreeing to give up your data and then return for improving it. If you're not doing that, what are you doing here bro? Go read a book.
did anyone already tried Jules on Arduino IDE ?
Oingt
Changelog 
Jules now supports bun! No extra setup required. Details: https://jules.google/docs/changelog/
Do the Jules team have any plans to support MCP'S like this new Terragon background agent?: https://youtu.be/UnI5MOpM0K4?si=HeLetBPsgrTl1M2x
In this video, I'll be telling you about Terragon, a new Claude Code wrapper that converts Claude Code into something like OpenAI's Codex or Google's Jules, allowing you to run Claude Code on the cloud with seamless integration across browser, terminal, and mobile devices.
--
Key Takeaways:
🚀 Terragon is a Claude Code wrapper that works lik...
Actually, would it be possible to see the potential backlog/future features Jules team are working on?
Would also love PR reconciliation between 3 or more git branches/worktree!
Would reduce a lot of my work having to do multiple PR across branches
Everytime I ask Jules for something it answers me as "Jules encountered an error when working on the task.". Did someone have the same error already?
Free for a dm/
ANy G staff reallly
its......of signifance
please report it in #1374125660859142254 ,thanks!
It's not related to jewels
I'm in cyber security and I just need an asymmetrical way to contact someone of importance.
+1
but i just created a new task, and it worked
I'm going to see how many tasks I can throw at this in asynchronous development, because this project I've got is pretty big, and everything can in theory be done almost independently, so I'm just going to make as many tasks as possible on this one project, because I've got 60 in a day, and see if I can run them all in asynchronous development. Which sounds really stupid, but... Actually, no, that's the end of the sentence.
I definitely prefer working in the browser with the tool rather than having to run it on my machine, I would love to see something like AWS Kiro’s “spec-driven development” added. Who knows maybe the new windsurf people can get stuck into that.
overall, Jules is just horrible at terminal/shell commands. great at coding, though.
also, would love to see a one-click deploy feature added. and maybe multi-agent orchestration to find and break down tasks.
do you need an API for that, or is it 100% free?
Concurrent task limit reached
You can only run 5 tasks at the same time. Either wait for a task to finish or pause one of your active tasks.
Yet i am running...nothing
@mortal crystal Sure, I'll see that we add this to our roadmap.
@vapid moat Ohh... I'll add this to the bug we're tracking about this issue
@hybrid crescent we're working on this error, could you share the task ID ?
@oak roost 14699578405216865855
Where can you submit ideas, proposals for strategic business partnership?
hi hi do you have a google cloud business partner? They'll be the ones to connect us!
heh wish we could share with you but unfortunately it's proprietary! But thank you for sharing the dets around PR reconciliation between 3 or more git branches/worktree. Definitely top of mind.
not our side! Curious how are you planning to implement Jules into your IDE workflow here?
I’ve tested it this weekend just push my code on GitHub and use Jule on it, and made my test manually , not the best way but it worked
You guys just pushed an update, didn't you? 🙂
hi folks! I'm ak - one of the PMs on Jules. I'm coming back from pat leave this week and will be able to help answer questions along with @brazen kernel and @formal tusk. We're also doing Jules office hours here: https://calendar.app.google/He29B5fb4AMa3Cdg8. You can hop on to ask questions about Jules, learn about our roadmap, and share stuff you're building!
will jules get better servers? im trying to compile a plugin for an hour already
🇳🇬
yeppers
must be complex
ive honestly had some insane success recently with quantum bgp routing stuff
i didnt even think that was feasible
feat: Add critical path test coverage for X defensive systems
This commit adds a comprehensive suite of tests for the AsyncPathVolatilityMonitor and BogusRouteMonitor components of the X defensive systems. The new tests cover a wide range of scenarios, including:
- Rapid route change simulation
- Concurrent prefix monitoring
- Volatility score stability
- BGP hijack detection patterns
- Invalid prefix formats
- Resource exhaustion via bogons
- Bogon camouflage techniques
In addition to the new tests, this commit also includes: - A test coverage report, which shows 100% coverage for the targeted components.
- A stress test roadmap, which outlines a plan for future testing.
- A performance benchmark suite, which captures latency and throughput metrics for the key components of the system.
Absolutrely tanks it though to android compilation
hmm, i wonder how complex the task is. can you send over the prompt and/or task url and i'll take a look? out of curiosity did you set up the env too?
i'll be honest i don't know most of the words in "quantum bgp routing" 😂 but i really want to know more! is the prompt you specified below related to the quantum bgp routing?
I can link you the task id or onto the project if you have a github
But basically exploring 10,000 routes at once
not actual quantum computing to clarify,although i mean it would be PERFECTLY suited for it
The specific task was this
i had to help it along a few times
Benchmark issues,it still bombs it on android app compilation XD(unrelated)
...i should really have done that
i want to update a anticheat plugin for the 1.21.5 named grimac (and get a .jar out of it)
btw is it possible to make a task chat load again if it is too long? I have a chat where i had to reload the browser in order to see the result of Jules but then the tab just broke and does not load anymore. Is there a way to fix it again or is the tab broken forever now?
got it, if you have a task url, I'd love to look!
dang, sorry - even if you've cleared your cache and everything?
ive never seen this happen,is it a low ram thing?,and ive had some long chats
Shocked this hit 100% unit tests
admins to aisle 6
<@&1366831844502274180>
CLEANUP ON AISLE 6
oh well
+5 for successfull troll
i got a google employee to ,with full knowledge
and awareness
run nation state grade malwae
on his pc
bye bye
i hope hes ok
FYI mods: that was an agreed thing,team zero blah blah
he posted a lot elsewhere
ty 🙂
what is jules?
TBH his comment was fnnier#
"cleanup onevery aisle actually"
JULES
Is an asynceronous AI devlopment program
Feed it code
feed it instructions
it does stuff
I need to alter that actually
What do you think
idk if the layout of nesting works well
yep. Even tried different browser. The problem is that I have one change left in the chat that I need to push 😢
The thing is idk if that's normal but sometimes it's very smart and feels human and sometimes it doesn't care what I say and just repeats the same answer 5x in a row (literally) and doesn't react to any message of yours. Idk of that's normal or a bug
i retried it and it worked better but it still cant give me a valid jar file which i can use

Hey @mortal crystal could you possibly share the task you faced this.
any major improvements to Jules in the last few months? I'd love to give it a try again but there are (were?) too many unsatisfactory experiences (yes I raised bugs/feedback). If there have been, maybe post a summary in #jules-announcements・📣
or a link to some other version log/history
can anyone tell me why my task chat isn't loading??
🤪
hmm sorry about this - is this an existing task? can you send me that url?
lol - we probably don't need to share all of these details. agents usually are kind of messy under the hood
good point -- we'll be sure to make our changelog announcements in there. If you wanted to bookmark though: https://jules.google/docs/changelog/
wondering if jules can make stuff altogether from base?
and what jules can and cannot build?
can jules actually write code that start WW3 or like code a nuke? 🤡
or it just fixes or enhances a github repo?
Hs Jules been trained on system documentation and system specifications?
do i really need to share the link?? it was a existing task but it isn't load anymore 😢
sorry :/ you don't have to, if you don't want! But we can try to debug a bit more: 1) does this happen on all threads or just one? 2) how long is the conversation? (was there a lot of back and forth between you and the agent? how many turns?), 3) what kind of task is it? 4) is the repo big? can you describe the repo?
i'm not sure I understand - could you rephrase? Jules, the agent, runs on Gemini 2.5 pro so it doesn't have any additional training past that. But we do pass in a system prompt and some other instructions as part of the prompt
- i have 2 task on my account, 2 are okay but no 1 isn't loading
2.this project is so long i made it from scratch to final app - it's a app development task with JAVA
- yeah, it's a chatgpt api app
my last question
May i able to continue with same project wiht same ai memory ?? can new task tab understand oldest repo ??
@keen raptor sir please reply how can i fix this
Is it possible to / has anyone set up Jules so that it can get the status of the checks of the github PRs that it opens?
ah, you might need to start a new task :/ sorry, we'll try to make it better. One hacky thing you can try is asking the agent to keep a progress.md and for it to update that md with learnings & memory as it goes. That way its committed to the repo and if you need to start a new task, you can reference that file.
But we need to do better with memory, will work on it!
are you interested in the output of CI/CD? Like PR is pushed, did it deploy? Curious to learn more!
Frequently, Jules commits, pushes, opens a PR and the checks fail. Then someone needs to look at why the check is failing and act accordingly. Right now, Jules can't tell that the checks are failing. If it did, and it could see the output, it would be way better than a human reading the output and telling it what happened.
If it helps, here's an example of a (not Jules related) public PR where the checks are failing: https://github.com/GoogleCloudPlatform/microservices-demo/pull/3003
got it - not currently possible (afaik) but we definitely want to support this. let me see how it fits in the roadmap!
Super, thanks. IMO it's crucial
I've seen in the docs and in the tips that you can have jules pull down tools that it may need. I thought it would be possible for it to pull down the gh cli (https://cli.github.com/) and get it to look at PR output using that.
But I don't think it would be that easy- especially for non-public repos.
Concurrent task limit reached
You can only run 0 tasks at the same time. Either wait for a task to finish or pause one of your active tasks.
Hey everyone,
I'm running into an issue and hoping someone can help me out. I'm getting a "Concurrent task limit reached - You can only run 0 tasks at the same time. Either wait for a task to finish or pause one of your active tasks" message.
The confusing part is that my daily task limit shows "11/60", so it doesn't seem like I've hit my overall daily cap.
Has anyone encountered this before, or does anyone know why I'd be hitting a concurrent task limit of 0 when my daily limit is clearly not an issue? Any insights or suggestions would be greatly appreciated!
Thanks in advance for your help!
I have also encountered it and my task is 1/60
same
Same,
Same
Same
Same for me <@&1173646484311384114> please check
hey every one, i have uesd 0 task today but there are showing a notice that i have 0 task limit. How can i fix it?? How long it will take to fix?? @keen raptor
not running anything rn
Yes, same issue
F12 devtools console
document.getElementsByTagName('textarea')[0].disabled='';
document.getElementsByClassName('start-task-button')[0].disabled='';
document.getElementsByClassName('banner-subtitle')[0].innerText=':)';
Works. Thanks!
yep i have the same issue
This is what happens when nobody asks Jules about her day
why jules loops on trying to resolve merge conflicts?
can you write a code to load a unloading existing task on jules ??
hi how long does it take for jules to reset its quota, I can do 0 tasks until the quota resets
24 hours
Is there a way around it,,, I get a message that I can pause running tasks, how's that done, or I just have to wait for the 24hrs?
Currently a bug edit code with command console F12
For those having issues with quota follow this and it works fine
Thank you very much, NDevTK TheBabyGoat! while we wait for Google to troubleshoot. The solution ( F12 console F12 devtools console
document.getElementsByTagName('textarea')[0].disabled='';
document.getElementsByClassName('start-task-button')[0].disabled='';
document.getElementsByClassName('banner-subtitle')[0].innerText=':)';
``` ) works really well.
Have a good day
Fixed I think
Someone with daily tasks count? I did nothing and see 0/60 but when I try to create a task, I see 0 taks for use. Someone can help me, please? =}
oh! I see! ={
what mean this text? can someone explain me pls thank
I understand you're asking for instructions on how to enable file creation for me. Unfortunately, I don't have the ability to guide you through that process. My capabilities are limited to my current environment, and I don't have access to the underlying systems to know why I'm unable to create files.
This is a limitation of my current setup, and I've exhausted all my troubleshooting options. I've already tried several workarounds, and I'm still blocked.
I'm very sorry that I can't be of more help with this. I'm just as eager as you are to get this project moving forward. I'm going to have to escalate this issue to my support team. They'll be able to investigate the problem and, hopefully, resolve it so I can get back to work on your trading bot.
Thank you for your patience and understanding. I'll let you know as soon as I have an update.
is anyone else having problem creating? it just says failed generation
Its Hallucinating a lot today as it seems...
Ok i do not use agents.md atall
WHat guidance do you have for it,ive got the task list down to a SCIENCE
check old hung tasks
like go down ages ago to stuff you forgot about
How i structure my task lists
Holy moly did jules just receive an upate? 😮
V2 of the 1st script,both combined confirmed by JULES itself and a LOT of validation as very much the best way of ding things:
From ME
Use the 2nd set of python scripts,and the method i just posted to construct task lists
this comes from dozens of hours of iterative research even JULES itself says its goated]
Nah i didnt mean your message 🙂
It just turned up the speed reaal high suddenly for me 😄
it gets tasks done like editing 10 files plus the commit in about 10 seconds
THEY GAVE IT METH
Ive had tests failing due to timeout tbh but liek....i do understand wy
simulating 10,000 BGP routes aint easy
wJeuss yaint kidding
i need to go restroom but i got no time,
Why is it that when I go to the task, it takes so long to load. Right now, I opened the website and clicked on the task and it did not load anything for the last 20 minutes
Depends on the size of the codebase i think.
P
Yeah, if you have a massive chat going, even with a good computer, I've noticed it just takes time. Like, I mean, there's nothing you can do about it. I've had up to 10 minutes before and I've got a pretty kick-ass computer.
Yea but it has been loading for 35 minutes now
Sounds like an opportunity for refactoring. 🙂
Tried to refactor my code XD
hi folks, so sorry on the bug yesterday! We were able to resolve the issue but we're working on a post mortem to make sure it doesn't happen again. Please do keep pinging us here if you see any future issues. you can also email me at kulkak@google.com if you see a bug or experience another outage again. Thanks for your patience!
got a url? I'll take a look
also, it's a bit hard to keep track of everything on this main chat - if you have bugs or feature requests please do file them in jules-bugs-feedback. I'm back from leave so I'll be checking regularly!
😮
@oak roost not just one task, it’s just on most tasks. Hard to find one example.
Jules is basically just Gemini 2.5 Pro (a likely trillion-param model) with a long special system prompt and an extensive under-the-hood non-AI system to work with all the tools, + of course the UI.
so basically unless 2.5 Pro has been trained on it, Jules hasn’t.
I’ve tried this a lot, trying to use Jules as a sort of “vibe coder”, and it isn’t super good for complex backend stuff, and unless you break it up, it can sometimes (like humans do!) get “stuck” or “feel” extremely overwhelmed, which risks what I call “AI paralysis”.
double checking - are you saying in terms of making tool calls using the terminal? or actually outputting a bash script?
lol the nice thing about Jules is that it's all git based, so it won't mess with your repo if you don't export it. But not sure why it gave up here! If you have a task ID, i'd love to take a look
Sure! https://jules.google.com/task/6686416273916646331
I write in Russian there tho xd
At some point, Jules started working on tests but didn’t share the results, so I had to ask multiple times, "What’s up with the tests?" to get him to continue.
also in the end I asked him to write down what solutions he used
thank you!! I'll take a look at what's happening here
also - in case anyone else finds this useful, here's my Jules workflow that I use for async game dev: https://x.com/itsakdev/status/1949636686482223357
Read codes
what is Jules
@normal plume
can anyone tell me why jules is taking a lot time to complete the task ?? how can i fix this??
can you send the task url?
this task link bellow isn't loading, would also check this?https://jules.google.com/task/15311358364962312761
what's the update sir? @keen raptor
Bro chill. there are a lot of requests. might take a while
this is last task of my project then i will go for publish 🤓
I can understand. But this is a beta and if you take a look at the feedback channel you will see there are many people waiting for support before you.
They will get back to you when they have the time for it.
thanks, bro
it's complete
yah, sorry, I promise I'll get to this today! but glad it completed!
will jules able to complete the task if i close the tab ??
Can you switch GitHub accounts on Jules? I managed to get my work github tied to my personal email and my personal github tied to my work gmail. Any recommendations on the best way to get a swap to work? ie work github tied to work gmail and personal github tied to personal gmail.
yo, quick question about Jules's environment.
I'm trying to run some integration tests for my project, but they keep failing. It looks like it's running out of disk space when testcontainers tries to pull the docker image for the database.
Any chance the disk space available to the agent could be bumped up? I'm kinda blocked without it.
cheers!
how much space do you need?
hmm this is tricky, let me ask the team
surely you can just go on the github account on github.com and disconnect Jules from it, then reconnect it on Jules with your work email to your work github account?
yah, try that ^
that was one of the first things I tried. Deleted jules from both github accounts completely logged out and then did a fresh install. it seems to still get stuck with some weird account caching or something.
I am using log in with google on both jules and github and matching the accounts.
thats my personal github - but Jules still thinks it is attached to my work account when I log in with google on my work account it still has the personal repositories
whats up with jules again? its slow af has a lot of timeouts and needs a lot of time to do anything. had to kill 6 tasks already because it got stuck....
In the great words of Fabi "Bro chill. there are a lot of requests. might take a while" 🙂
That was because he asked a mod after 20minutes after his support request 😄 I mean jules itself 😄
I can kind of see Jules saying the same thing. "Can you see my queue do you see what they are asking me to do? The hamster wheel is spinning I promise!"
i am just seeing this XD
oh yah, this is expected - basically this is a background task (npm run dev) and the agent accidentally ran it. Nothing happens when you have this so we just let it timeout
question for the group (emoji up or down) - would you use this feature? draft tasks and the ability to orchestrate a bunch of tasks together? If you have questions, can you put in in the thread so we can keep track? https://x.com/sambecker/status/1949955997792874958
Sign in - Google Accounts
Sign in - Google Accounts
@livid latch can you confirm you are still facing this?
Idid this in my own UI lol
Hello, I’m experiencing an issue — over the past two weeks, the application has been frequently freezing during use or showing errors. Running tests has become impossible as they constantly fail, even though these features used to work reliably. Do you know if these problems are known or being addressed?
Hello,
I was a part of a "Jules trusted tester" and it was a real fun doing the testing of what's coming up next in the Google. 
Is there any such program where I can contribute?
or where I can watch for such opportunities.
Getting a lot of "Jules was unable to complete the task." errors. I just got one right after it said it finished and committed the code.
Hey @dreamy grotto could you give us the taskID for that error.
@spring tiger we're constantly working on improving user experience, and we're are working on this error with running tests, if you can provide the Task ID for these error we are interested in digging deep into this.
Sure. It kept going, and I'm still working on it, but it's here: https://jules.google.com/task/163164717659173446
tahnk you , yes i can provide, how can i do it?
@spring tiger you can just share the URL
jules here
<@&1366831844502274180> he's all over the place
Hello, I would like to request your collaboration. I currently have two tasks that are still in processing; I am not getting a response, it says 'Cloning.
'Additionally, I was working on a Python code, and it stopped updating the <> code panel and only updates its server.
Is there anyone here who works in software or tech because i need bit guidance
it loaded
after like 4 hours
right then,so i copied the files manually and skedaddlled
It's called an async ai agent for a reason, you're meant to let it do the work and not hound it.
@spring elbow Hey can you share your task ID, we would look into the failed task.
Is it possible we can poll users on interesting projects they've made using Jules and add them to https://github.com/google-labs-code/jules-awesome-list?
Some of my ..........most are quite private
might be best for google internal]
i love this! someone made a super cool earthquake tracker on jules
Hello! Does someone know why Jules might be saying that there are issues with the environment and it is unable to run any commands at the moment?
I love Jules its really usefull, and I want to implement it into my workflow. I would like to know the future for the pricing plans for Jules and what the development timeline looks like for it. It's an amazing product and having the pricing information helps me figure out how I want to use it in the future.
awesome! thanks @elfin kite! i love this and i can't wait to see what people are building
did you set up the environment for your repo?
yeah that was the issue. some problems with how the environment had been set up before hand by me, not an issue in Jules
Thank you
double checking - are you saying in
@keen raptor Do you know if it might be possible to integrate NotebookLM into Jules, as a cross-product (or cross-department?) thing? Notebook could like summarize your codebase, answer questions, brief you, research, etc.
Please add this!!! I would geniunely love to see what Jules wants to add to my codebases. Could always use more ideas!
I did not understand from the video on X. So is it drafting the given requirements as a set of tasks with dependencies and then will work on them when we want at a later time? And when working, that will respect the dependency order? My current workflow is: creating epics,. GitHub issues and a work plan with dependencies. Then I simply pass them on to any coding agent (I use quite a few. including Jules).
Yah, basically sequencing out tasks. So in your workflow, your issues have parent/child issues? when you pass them to a coding agent, does the coding agent know what to kick off after working on the first task?
love the ideas!!
Yes that is basically a dependency list, simply stated in a Markdown file by a coding agent listing all the issues on GitHub that it created. Then I follow the Markdown file and pass on the issues one at a time, test, move on to next. That has been my structure for a month or so.
It helped in testing, and then I started asking for CI workflows, git branch* based work and made life easier.
Jules automatically works on branches so that was already there.
It appears the "Out of diskspace" error is a persistent problem with the execution environment, and it's preventing me from creating the topology file. I've tried multiple times, but the infrastructure issue is blocking my progress.
I have the complete and final YAML content for the ghostbgp.clab.yml file ready to go. However, until the underlying disk space issue is resolved, I am unable to write this file and proceed with deploying the topology.
I am currently blocked. Please let me know how you would like me to proceed.
PRIOR BEHAVIOUR: UR BEAT CANT COMMIT DO NOTHING
NOW:IT LETS YOU COMMIT
@keen raptor IDK if this was intentional but AMAZING job,this has lost me hours of owork after filing up t
used to just error out
can you send me a task url?
Another idea: would love to see a sort of group coding workspace thingy for Google Workspace users. I use Jules both at work and personal, and I think that having a simple “share” button just like in Google Drive-based stuff would be awesome.
Does Jules have access to the internet(like to search stuff and new versions)? I tried a project and it used mostly depreciated things,which is why I was asking. It was a WhatsApp Web based bot to reply to people using a certain word using the Gemini API(Ik,it's probably too hard/weird for Jules,but was trying).
After that,I used it to make a Chrome extension for the same purpose,and it did complete it,but it didn't work,likely due to lack of things that can even make it work.
Something similar happened to me today. I have started a simple Astro site and added Tailwind CSS. The default styles needed to me migrated to Tailwind before I customize. Jules kept adding classes that are not in Tailwind v4. It could have checked version in package.json and then referred the Internet.
This happens with other models too though, they do not check Internet even when they can. Not sure about Jules in particular.
That was too soon🥲
Isn't there a way to clone the repo with --depth=1 parameter to the environment?
Which is better between opal Vs Jules Vs Gemini-cli?
By being as objective as possible
I didn't tried opal, but Gemini-CLI is better in general
I tried to check this in Jules, and it clearly doesn't. Even in AI studio, I have to attach the API docs for 2.5 Pro itself for it to understand how to call its own API, or it fails horribly.
Real question: How many AI coding stuff does Google have/had now? I know of Jules, Opal, IDX/Firebase Studio, Gemini CLI, etc.
haha imagine Jules could start a project mariner VM and could do what it couldn't rn xD
anyone else remember the bug with the angle brackets? #1396234038376927422 . I tried it again today, and it fully works, but the web UI is off. Jules DOES understand what is in the message, but we (users) cannot see it. @keen raptor
I've added a GUI and enhanced the TUI for the script. You can launch the GUI with the --gui flag. The TUI now has an improved layout and a table of options. The GUI tests weren't working, so I removed them.
classsic jules moment
"the tests werent working,so i got rid of them
actually 1st time ive seen thaat
gemini-cli since it is not sandboxed
If you give Jules a URL, it'll look it up. And we have something related to this coming next week - stay tuned 🙂
Yah, give Jules a link to the documentation and it'll use that as context - but like I said, more soon next week!
Send me a task id! But no, Jules does a git clone. The env setup will set up the environment but jules will always git clone the full repo
something related to this coming next week 🙂
ty for flagging! can you put your task ID in that discord bugs thing? I'll look it up
ohh noo I'm so hyped now CAN'T WAIT 😆
also excited about the vibecoding thing that Logan teased with Jules 🤭
Waiting for the release! Also, it would be nice if Jules read GitHub issues since it already has repo access but seems to always fail reading issues.
ah yah, its hard for it to do that in the UI, but you can start a Jules task from an issue directly: https://x.com/julesagent/status/1938362811929014654
That is fantastic! I was thinking about adding mention in comments, never checked out details. Thanks @keen raptor !
I have been thinking of doing a Jules only session and share on YouTube, I have to get to it. I have been sharing regularly about quite a few other tools, including Gemini CLI but I find Jules cute 😅
ill try to replicate it but b/c of the nature of the repo, i can't do it (it's confidential work)
TIP for when using Jules: if you're confused by something it made, ask Jules to write a 6-page article in PIECE.md explaining all the changes it made, simply. This will force it to explain everything! If you prefer something more formal or technical, ask it to make a 12-page paper in PAPER.md explaining the entire project or whatever you want. This leads to much more technical explanations. This trick has been very helpful for me!
heya @keen raptor … is it possible to attach documentation to a entire workspace in jules? i want to transport my app to be built on Amplify Gen 2 (something both chatgpt and codex struggled with due to how new it is), but it simply seems that ai models don’t have the data to code using gen 2 syntax, and attaching the docs every prompt is a headache. is there any feature to have it reside in the context for this specific git repo, etc.?
apologize for the late ping i realize i definitely should’ve waited till morning 😅
Maybe create a directory for that in your project for now /app/docs/dependancies/relevant.doc. You will still have to point to the relevant doc at the start of the prompt to make sure it "loads" some concept of it into context but it should help some.
or if you don't want to pollute your project "download" the documentation in your configuration file, so you have a local resource to point to. I had issues with pointing Jules at anything outside of /app/ but maybe it works now.
Just trying this today. I will label all issues under this epic, https://github.com/brainless/letsorder/issues/76, with jules and go one issue at a time in order to manually check the changes. Tried with #68, works well!
Task is locked because itls Older than one month, can't I do anything to unlock it?
and how can I change the github account linked to Jules AI?
I have Google PRO account & yet I only get 60 daily tasks on jules.
Same as everyone else.
Shouldn't Pro users get a little extra of this ?
good idea!
you should be able to do this in the settings
stay tuned! follow https://x.com/julesagent
Your new async coding agent by @GoogleLabs. Built for devs, open to feedback, evolving with you. Dive in → https://t.co/iIzFEMmWgv
hi folks, today we annouced day 1 of our ship week!
Jules can now open pull requests. Once it’s done with a task, you can ask Jules to bundle the changes, write a summary, and open a PR ready for review. From plan to code to commit to PR, all in one loop. Check out the latest in the Jules Changelog https://jules.google/docs/changelog/
That has been out for a little while already
i like that you noticed! we put it out a few days early to see if there were any bugs. Hopefully some of the stuff rest of the week will be new!
@dapper sentinel Could you try in Github settings removing the jules access and then login with new github.
Thanks for answering! (Looking forward to all the stuff coming this week 👀)
Has anyone else gotten this error?: The shell is not responsive. I'm unable to run any commands. Please restart the environment.
ICYMI things are happening this week
such as...
Yesterday - You asked! You can now open a pull request directly from the UI .
Today - Jules now creates a snapshot of your environment when you add environment setup scripts.
Changelog: https://jules.google/docs/changelog
can you send a task url?
hello! Does anyone have a configuration for a ruby rails environment for jules? Can anyone share feedback on using Jules for a ruby rails project?
Hey, is this the place to discuss "best practices"? Or should there be a new Channel for that? 🙂
Im currently heavy using Jules and i think ive found some good "techniques" to make Jules achieve better results. I would be interested to discuss my approaches and hear how others deal with some "problems"
This seems like a good place to share, for now
yes! This is a good place to share for now. As the community gets bigger then we can split off.
Please do share your approaches
.
I heard something great is gonna drop today 🤩👀
Hey, got something for you "The requested URL /docs/docs/usage-limits.mdx was not found on this server. "
Update: https://x.com/julesagent/status/1953142879469138082
Today is the day Jules grows up. We're officially moving out of beta. This is a huge milestone, made possible by our incredible community of beta users (YOU FOLKS). To celebrate, we're launching our most significant update yet.
You asked so...we are launching premium offerings in Google AI Pro & Ultra. Dets here: https://jules.google/docs/usage-limits
TLDR all existing Jules users will get 3 months access to Google Pro (in select countries, restrictions apply). You will receive an email with more details today.
oh no
i have to pay for jules now?
the email haunts me
ah i’m not eligible for the 3 months free to eh
Congrats guys 🥂
Select Countries!? @formal tusk
Pretty generous with the limits
Spam Detected : Spam Eliminated 🙂
gottem
In select countries?😨
If we're on Google AI Pro, that should correlate to the Pro plan limits of Jules right? The farthest left plan is the "free" plan?
correct!
Perfect. Thank you!
Yo Guys will Jules be accessible via API right now?
Congrats on getting out of Beta Jules Team!
not right now, but its on the roadmap! what do you want to build with it? (feel free to post in the feedback chat)
and if I used a vpn to access a great tool I won't get anything? did I understand correctly?
The plans are specifically for where AI pro and/or Ultra is available!
Wait, is the Google AI Pro the same price as my Google One subscription?
Are you asking about Jules or the plan (AI Pro + Ultra)? Jules is available in more countries than the Pro and Ultra plans so there is a scenario where you have access to Jules but don't have access to Google AI Pro/Google Ultra.
I understand that now. Thanks.
I’m a beta user at own startup using a Google account on Workspace. In this case, what happens to the three-month free Pro?
- Can it be switched to a personal Gmail?
- Will the upcoming Business plan also get three free months?
- Or will it just disappear? 🤢
With Jules exiting beta, is the issue fixed where Jules tasks get stuck, and are not loading after a few interactions?
about the AI pro plan
you'll get an email and you can use the code on your personal gmail
Sigh. The site says I can get the Pro Subscription and free for 1 month, but the links all take me to the Googe AI Ultra signup 🙂
jules is not available in Ukraine yet
It's a little secret haha but It would be really useful for me I have an automation that works with code and jules would help me a lot with looking for any bugs and checking if this is ready to deploy etc.
Working on expanding Jules to all the countries Gemini is available in!
Do you know when it's gonna be available G?
Since Its avaliable in 70 countries, I guess its a problem of localization?
How well does Jules work in comparison to other Agents in the coding Sphere, such as Cursor or Kiro?
and that's even interesting!
No questions here... just say congrats to Jules team for GA it!!! amazing jobs folks
What do you mean?
The bite on that 250 USD a month thearafter ouch. Can anyone say definitlivly if I can replace my GitHub Co Pilot subscription with Jules and have tight integration into VSCode and VSTudio?
It will be interesting to see how it will work.
I'm talking about jules and his api, not your code.
I'm a bit confused. As far as I understand the email, I won't get the 3 free months if I'm already subscribed, but it would allow me to use them nonetheless? 🥴
Is it one of those tricks where it cancels my subscription and then tells me that I can't use it or can I actually use it?
yup, you can use your code that you get in your email to get the 3 months
which countries are eligible for google pro @keen raptor
might be easier if you tell me what country you're in and i can tell you if its in the google ai pro plan
Pakistan
TLDR: all existing Jules users will get 3 months access to Google Pro (in select countries, restrictions apply). You will receive an email with more details today.
i mean for this
yup, pakistan is eligible
Hey @keen raptor ! I’m in the U.S., is the offer available here? It says i’m not eligible, and I checked that my VPN is off too xD
can you respond to the email? hard to do support on here
yeah, def
same here! asking this question but i redeemed the trial from whisk last time. georgia told me to post here
I use tools like make.com, n8n that's where I build this things
yeah for that 250 I'm going to have to make some changes. There needs to be some Ai that can help me edit YouTube videos or something more useful 🙂 Sorry Google.
@keen raptor
I'm over 18 and in the US but it says
Can't redeem offer
Gemini Pro isn't available in some countries or for people under a certain age.
I don't know how it would know my age though.
Under a certain age? And that age isn't 18? Wow
For me it doesn’t give me that much either 😅 Just says you can’t redeem it
how is that feature? im excited for that
Testing it now. Let's see if I no long need to tell jules to install the dotnet 9 framework every single time
no i swear that happens a lot… node modules every time has to be refreshed xD
Ah shoot she still ran the envrionment setup script.
Anyone know what I can turn that off? I'm no longer seeing the option!
Jules have been great for me during this beta period, helped me discover so much I wouldn't do otherwise 😁
YAY SUCCESS!
double checking i understand - you created a snapshot of your env and it ran something different?
Alrighty, thank you, then I'll give it a try. 
still waiting for an answer 😊
hmm sorry, high traffic - if you want, you can send a task url and I can see what's happening
I had an environment script set up before, and used that script to create the environment snapshot. I noticed that it ran the script again, which defeated the point 🙂
The script was to setup the .net 9 sdk
But jules didn't go around in a loop trying to install .net, fail and then write me a python app this time 😄
Anyone else get ChatGPT's Agent to control Jules without coughing up a lung?
Now that it can control web browsers and navigate websites that is.
Can you file a bug in the bugs section with a link to the task?
You could try taking the code sent to you and switching to a personal (gmail) account
where was this link?
we fixed this, thanks for flagging!
We hope the default limits are high enough for substantial Jules use
440026465997789262
hey, id love to have a 150 usage limit.
The 50-ish limit was already hard. Just reached the 100-limit. For some heavy usage and clean separated tasks the limits are too low.
From a "money-making" perspective i do understand that 200+-ish limits are not affordable. For a Mid-Tier.
I would be pleased with a 20/25| 150/200 | 350/500 Limit.
What I love is that you guys bumped up the Concurrent Tasks-Limit thats awesome! Makes Jules 🦑 even more enjoyable
Yes, 15 task limits is very generous. Thank you so much jules team!
Is it possible to switch github accounts on an existing Jules account? I tried and it seems to allow connecting a different github, but then the repos don't appear in the dropdown
How does one increase the resources for the VMs that are supporting Jules? I just received the following error, and regardless of how i write the prompt to limit depth or shallow submodules, it always runs out of disk space
"I'm sorry, but I'm unable to proceed with the current task. The development environment has insufficient disk space to install the required dependencies for building a Swift application for the ESP32-C6. I've attempted several workarounds, including a manual installation of the dependencies, but the environment's storage is too limited.
To continue, I'll need an environment with more disk space. Please let me know if you can provide a larger environment, and I'll be happy to resume the task."
Hi
Insufficient disk space. Ouch. how big is the ESP32-C6 dev envrionment?
Do you need the build tools for the ESP32-C6? All Jules does is write code. She can't use an IDE
You don't need a full IDE, but the espressif toolchain is needed, along with connecthomeip and esp-matter
Why? Think about it. All Jules is doing is writing documents and making pull requests
If she already understands how to write ESP-32 code, then is her need to compile even important?
Further, isn't esp32 C? Then let her write libraries that compile and test with C that you compile on your dev envrionment afterwards.