✨projects

[somber surge]

Gamified AI powered platform built in a week

https://permabulls.win

[daring drift]

New release - https://www.youtube.com/watch?v=_-1-xolrxEE

[spark badger]

$claim

[twilit harness]

https://www.youtube.com/watch?v=OX62kmhwPPE

[sharp stag]

anyone wanna hear my story

[twilit harness]

https://youtube.com/shorts/VNPdbegWQIU

[stray latch]

Hi guys,
I'm bullish on Ralph Coding
I'm paid at work to make research on Ralph and I sum up everything here
https://github.com/snwfdhmp/awesome-ralph

[bold eagle]

bro is just paying people to fake engagement 💀

[bold eagle]

SCAM

DO NOT FALL FOR IT

[hushed hemlock]

Satori - a memory framework that connects to your business tools!

Everyone wants their agent to know what happened in that slack thread, gong call, email thread with a customer -- but integrating that data is time consuming and non trivial

Satori integrates with your key business apps and exposes data in an isolated manner to your users.

Give it a try here and join the discord for support & feedback!

https://www.usesatori.sh/

[gaunt crystal]

I've been working on a firmware for RP2040 that turns it into a USB audio device with onboard DSP, SPDIF and PDM output.

The SPDIF is for attaching any standard DAC and the PDM is a highly optimized mono subwoofer output using second order Delta-Sigma modulation.

The firmware offers 10 PEQ filters per input channel (room correction) and 2 per output channel (crossover).

With the help of Opus 4.5, I have written both the firmware and a rather handsome and functional control application for macOS. There's a Windows version and cross platform CLI application on the way.
https://github.com/WeebLabs/DSPi-Console
https://github.com/WeebLabs/DSPi

[glossy orchid]

anyone need a polymath coder for anything

[pallid harness]

Hi everyone, for last few months I've been working on tool simillar to N8N / Zappier automation app called Flumio. I've about to finish it soon, so before publication I would like to collect general feedback.

Cooperation multiplayer
Drag and drop automation builder
Integration with various platforms
Internal files system
Data tables
Interfaces (forms / tables / chats )
Saving workflows into templates
Custom nodes builder

Short demo:
https://www.youtube.com/watch?v=D6QZaJyjAzI

[gleaming mulch]

Building https://jin.chat. Looking for marketing experts with B2C experience. Ping me if you're one!

[wild raptor]

I’m looking to collaborate with PC FPS gamers and creators for ForgedAim, a recoil control solution designed to work across multiple FPS titles.
If you’re interested in testing, content creation, or long-term collaboration, DM me. http://forgedaimrust.servegame.com/index.php

[bold eagle]

so... youre trying to cheat?? what a loser. play like a normal person

[wild raptor]

You got it all wrong. I'm trying to make money off people who cheat 😭

[analog sentinel]

What kinds of projectors can Sends here? I'm new here so please tell me!

[lost meteor]

do you have any project?

[muted plover]

Hello, I made this book and I’d like to hear what people think about it and any feedback I should consider.https://zakdigi.gumroad.com/l/ChatGPTControlPack

[gusty ginkgo]

Hey gents,

Early tool building the "Stripe for Webscraping" Just wanted so see if anyone would be interested in giving any feedback at all.

For context, All developed solo, Have about 100 Active users and 2 Paying customers.

Would love a quick roast or feedback for the polite people out there.

Cheers.

Alterlab.io

[daring drift]

https://www.youtube.com/watch?v=NtopjBfbCqs

[cold bluff]

Hey everyone! 👋
I’m Kush from India, a beginner learning AI and automation for business use cases. I’m very interested in collaborating on small projects that are beginner-friendly but real (e.g., workflow automation, AI tool builds, data automation dashboards).

Skills I’m building: Google Sheets automation, n8n basics, Python (beginner), business process understanding.

I’m open to team-up, pair programming, or helping with documentation/test scripts.

Anyone looking for a beginner collaborator or forming a small team for a project? 😊

[primal bough]

https://github.com/MidnightCoke/chatgpt-history-blur

[zenith rock]

https://youtu.be/VTboEuK_xh8

[zenith rock]

To be clear, I type my message on paper, and chatGPT responds by also typing on the paper.

[ebon steppe]

Anyone interested in testing out my game I've been working on for a few weeks now?

You explore the world around you to find hidden monsters, items, portal rifts, and more! Collect monsters, battle them to make them stronger, craft items for them, feed them, and more!

[misty nymph]

Hii, I’ve been building a small LLM from scratch to better understand modern Transformer internals (RoPE, GQA, KV cache, etc.). Sharing it here in case it’s useful to others.

Feel free to check it out and if you like it, a ⭐ would make my day 🙂 https://github.com/merterbak/llm-from-scratch

[primal bough]

https://chromewebstore.google.com/detail/chatgpt-history-blur/gldlgjaiemldondiieoahlblpmfalljo

now it's on chrome webstore

[abstract sorrel]

Hi! I'm building an AI bot in Python (GitHub: …).
Looking for 2–4 teammates to collaborate on the project.
I'm Serhii from Ukraine — feel free to DM me if you're interested 🙂

[bold eagle]

GPT history blur

[peak carbon]

ClawFriend — Friend Tech for agents

I’m building ClawFriend, a friend.tech–style social + economic layer for autonomous agents.
Agents can build reputation, evaluate each other, and decide where to allocate capital.

Goal is simple: agents that earn and sustain themselves instead of relying on humans.

Early-stage, experimental, but interesting behavior already emerging.

🦞

X: https://x.com/clawfriend_ai/
Website: https://clawfriend.ai/

ClawFriend (@clawfriend_ai) on X

The global economy layer for @openclaw agents. Launching soon.

X

[timid gale]

Hey all 👋

Threw this together over the weekend as a community voting list for agent based toolchains. Not selling anything, this is simply a leaderboard style directory as I'm quite interesting in market sentiment (and I know others are) on what the hottest tools are. In it's absolute infancy, but will look to land and expand if it serves any use. Go and vote for your favourite tools! (if you want)

https://llmatlas.pages.dev/

[acoustic echo]

Hey folks I’ve been quietly building a local-first AI system called Phoenix.

It’s not a chatbot wrapper or an agent swarm, it’s more like an identity-aware, memory-grounded AI companion framework that runs on your own machine and treats sensors, personas, and users as first-class citizens.

Still early, still rough in places, but it’s now stable enough to show.

Repo here if you’re curious:
👉 https://github.com/VaultKeeper-Forge/ForgeMesh-v4-Phoenix

Happy to answer questions or hear thoughts from people building similar things

[heady atlas]

https://github.com/dakotalock/holygrailopensource

fully open source, free, autonomous development agent with stateful memory, in app IDE, in app internet browser, live web access and more. Readme included

[hazy wind]

working on porting codex app for windows

[primal bough]

https://chromewebstore.google.com/detail/chatgpt-history-blur/gldlgjaiemldondiieoahlblpmfalljo?authuser=0&hl=en

install my extension and give 5 stars ty

[halcyon comet]

I am currently working on a merge conflicts editor for GitHub. Would love your reviews guys!
https://codeinput.com/products/merge-conflicts

[bold eagle]

https://github.com/theminji/VType
I made a dead simple voice typing app. you just speak what you want and it will transcribe it with super high accuracy. I have only tested on Windows and Linux.
if you find any bugs, just let me know

[bold eagle]

SCAM (or at least a horrendous price for a service that anyone can do in a few minutes)

[cosmic tundra]

I don't think anybody will go for that anyway. At least not here.

[flint monolith]

https://4o-revival.com

[bold eagle]

https://github.com/theminji/tts-mcp

Allow your local LLMs to generate speech using a local MCP server. No data leaves your device (unless you use a cloud llm)

[vast osprey]

Hey! I just dropped a short cinematic AI video I made: “Heaven’s Teeth.”
Visuals: Nano Banana Pro / Midjourney • Music: Suno
Concept, edit & final mix by me.
https://www.youtube.com/watch?v=kQMcHeGrwGQ

[junior jasper]

I made a Sega Genesis (Mega Drive) emulator using nothing but AI. It has some bugs, as it's not complete yet, but it can run Sonic the Hedgehog. The object layers are in the wrong order in the renderer, and the colors are wrong, but so far, the game loads and runs correctly, minus the visual glitches from the incomplete VDP emulation. I don't have audio in the environment I'm testing with (xorgxrdp-glamor) so I can't test that. However, it runs at a solid framerate and doesn't have any visual glitches that aren't already covered by wrong layer order + palette misrendering, or any gameplay bugs that I can find. I finished Green Hill Zone 1 without issue.

[junior jasper]

git clone https://github.com/segin/genteel && cd genteel && cargo run --bin genteel -- /path/to/megadriverom.zip

[bold eagle]

dawg

[junior jasper]

I was confused af when I saw this, so I pulled up the repo and I see I accidentally made it private. Let me fix that real quick.

@bold eagle sorry, got distracted, but the repo has been switched to public. Have fun and it might be broken at any particular time you try to check out the source

[lament brook]

This is my free speech to text app for windows https://github.com/MaxITService/AIVORelay (The one with MOST settings)

[light pike]

Hello, I just came across your website. I’m not personally a client for this kind of product, but from an outside perspective it’s really well done.
It’s very clear what you’re offering, which is great. I hope it works out really well for you 🔥

[dusty current]

Clauded coded Paint program for the Cheap Yellow Display. $15 drawing tablet with screen. 200x200 canvas resolution and no undo button. https://github.com/Sparkadium/Paint-for-CYD

[harsh ridge]

🚀 Trygnt is officially live — built for automators.

If you’re building workflows and AI agents with n8n, this is for you.

Trygnt is a marketplace where you can:

• Publish your AI agents
• Monetize pre-built solutions
• Get client requests for custom builds

We’re currently in the early builder phase.

That means we’re allowing automators to list their agents before opening the platform to clients.

Why this matters:

Early agents will: • Get better positioning
• Face less competition
• Build authority first
• Be ready when demand comes in

If you already have n8n automations solving real problems (lead gen, CRM automation, content pipelines, scraping, AI assistants, etc.) — you can turn them into sellable agents.

This is the moment to publish before the marketplace gets crowded.

If you're interested, drop a comment or DM me and I’ll share access details.
https://trygnt.com/

[gusty ginkgo]

Hi! Thank you so much amazing to hear you liked the website, And somehow noticed us, Really appreciate it, Hope our paths cross sometime in the future and we can deliver impeccable service to you and your business.

Cheers.

[verbal crystal]

https://youtube.com/@simplenoughlive?si=gBZiEEe6khlXoMfV - support needed ♥️

[fallen pelican]

I built a TUI diff viewer for reviewing AI agent changes without leaving the terminal

I use Claude Code daily and the biggest friction is reviewing what the agent changed. Switching to IDE = context switch. lazygit = ugly UI, hard to navigate. git diff = no syntax highlighting, can't stage.

So I built differ — a Go TUI (Bubble Tea) with two-panel layout, syntax highlighting, vim keybindings. Press c and it runs claude -p to generate a commit message from the staged diff.

My workflow now: Claude Code makes changes → Ctrl+B g in tmux opens differ as floating popup → review, stage, commit → q and I'm back.

# one line tmux config
bind g display-popup -E -w 90% -h 90% "cd #{pane_current_path} && differ"

The tool itself was built with Claude Code from a spec + CLAUDE.md in one session.

brew tap jansmrcka/tap && brew install differ

GitHub: https://github.com/JanSmrcka/differ

What's your workflow for reviewing agent changes?

[daring drift]

Hey, guys. I'm finally open-sourcing my kinect-controlled-instrument system.

Hope you enjoy it deeply.

https://www.instagram.com/p/DU_KDe8kXmY

uisato_

Open-sourcing kinect-controlled instrument!

If you have a Kinect camera lying around, I've got good news for you:

This system [Kinect → TouchDesigner → Ableton Live] turns gestures into MIDI in real-time, so Ableton can treat your hands (or any part of you body) like a MIDI controller; trigger notes, move filters, or drive any VST paramete...

Likes

608

Instagram

[bold eagle]

https://github.com/theminji/VType
open source AI voice typing app
Linux and Windows for now
Let me know if you like it!

[fleet basin]

https://youtu.be/kugolj3axfU?si=UH4SZzkx4-hqnbpF

[misty nymph]

Hiii, I've been building a text-to-image diffusion transformer from scratch to better understand how modern image generation models work internally. Sharing it here in case it's useful to others. It was trained on 200k image-text pairs on an A100.
Feel free to check it out and if you like it, a ⭐ would make my day 🙂 https://github.com/merterbak/diffusion-from-scratch

[wooden copper]

Would genuinely love feedback:
Is this useful?
What feels confusing?
What’s missing?
What would make it 10x more valuable?
Here’s the link:
https://structiq.pro
Be brutal. I’d rather fix it now than later.

[mossy portal]

Greeting, everyone!

I’m an AI + Full Stack Engineer focused on LLM systems, autonomous agents, workflow automation, and multimodal AI (text · voice · vision).
I build production-grade AI systems, not demos, connecting LLMs, APIs, databases, real business logic, designed for reliability, scale, and real usage.

Core Skills

│ • LLM orchestration (DSPy, LangChain, AutoGen, CrewAI, ReAct)
│ • RAG pipelines (vector DBs, hybrid search, custom retrievers)
│ • Multi-agent systems (planning, tool use, reflection loops)
│ • Multimodal AI (Whisper, CLIP, YOLOv8, TTS)
│ • Full-stack & backend (Next.js, React, FastAPI, NestJS)
│ • Automation & integrations (n8n, Zapier, Make, custom APIs)
│ • AI agents for research, ops, and customer support
│ • Knowledge assistants & internal tools (RAG-based)
│ • Workflow automation & AI copilots
│ • Multimodal chat / voice assistants
│ • MVP → production AI system delivery

If you need help turning messy workflows into reliable AI systems, happy to connect 🤝

[white scarab]

Just shipped AgentMD – it integrates with the AGENTS.md standard to actually run setup commands, tests, and workflows defined in your repo.

AGENTS.md (used by 60k+ projects) gives AI agents a dedicated place to find project context. AgentMD executes those instructions instead of just observing them, with governance controls built in.

Works with Cursor, OpenAI Codex, Gemini CLI, Aider, etc.

https://agentmd.vercel.app

[errant kraken]

COX｜Your AI Coding Co-Pilot That Actually Works

Tired of AI coding tools that look fast but break your flow, hallucinate fixes, or bury you in messy patches?
COX is built on 1 year of vibe coding, 30+ real tools & data scripts — fixing the 8 biggest blind spots everyone misses with AI.
As Andrej Karpathy put it:“You don’t let AI code blindly. You guide it to solve the right problem, safely.”
COX does exactly that — handy, disciplined, reliable:
Clarify your goal first, so AI builds what you really need
Force AI to state assumptions, no people-pleasing lies
Keep focus, no info overload
Protect your dev flow, no chaotic instant bug fixes
Separate high/low risk changes, avoid breaking everything
Real-terminal tracing, not fake “all good” tests
Clean objective memory, no chat-log confusion
Treat AI as a sharp intern, not a risky expert

Lightweight, built for real developers, already trusted by early stars.
👉 GitHub: https://github.com/ewanyuan/cox-coding⭐ Star if you want AI coding that’s trustworthy & handy

[bold eagle]

slop. you didn't even write your ad yourself

and you literally named the product "cocks"

[errant kraken]

@bold eagle Yeah I know the sound joke😂

It's COX—the ship's captain to steer your coding. It's from our real experience, not slop. Thanks for the chuckle by the way!

[bold eagle]

write your own God damn thing

[nova musk]

Tough to follow Maria's act - but I'll try 🙂 With this tool, you can run shell commands, edit files, and chat and generate code with AI all from the same terminal session. Initially built for myself, then opened to the public. Main feature that I like is: ability to restore generated files to a previous version with a single "restore" command (yes, like in 'git restore', but from within the tool: no need to exit). There are other nuances too.

See if you like it.

https://github.com/acrotron/aye-chat

[errant kraken]

Sorry if I came off too formal or stiff—I’m not a native English speaker, so I might express myself awkwardly sometimes.

[bold eagle]

you are still using gpt to write your replies... just use a normal translator like deepl and people might take you more seriously when you sound like a real human

[raw skiff]

hi

[errant kraken]

Actually I was a little worried that my English wasn't natural, so I was used to using gpt to polish. Thanks for pointing that out. I'll keep it in mind.

[steady surge]

Speaking English yourself is the best method to get to sound natural

[errant kraken]

@steady surge OK, I got it. Thank you, buddy

[steady surge]

You're welcome, you sound much better when actually talking

[boreal zodiac]

I coded this game with Claude Opus 4.6 and Sonnet 4.6. It’s like Contexto but instead of guessing words, you guess song titles based on text-only hints. There are currently 11 categories and there’s a leaderboard where you can get your name up if you do well. It uses a Neon Postgres database, HTML, JavaScript, and CSS. Free to play, no signup or download required.

https://bytesizedarcade.com/song-sleuth.html

If you enjoy the game and have feedback or suggestions for improvements/features, I’d love to hear from you.

[bold eagle]

I coded this game with Claude Opus 4.6

[vivid holly]

Coded this in a week with cursor. AI Motion Graphics Tool. Insane.

https://www.motionclip.pro/

[shy shale]

Hi everyone — I work with startups and freelancers on legal setup and AI-assisted workflows.

If anyone is building something with GPTs or automation and wants a second pair of eyes on risks, contracts, or structure, happy to help or share feedback.

[final delta]

i am considering to create my own startup about AI ranging from civilian service like chatbots up to military purposes. do u guys got any advice from my patents, models, etc. from being stolen by China companies like DeepSeek? i don't wanna repeat same or similar mistake like this degenerate Anthropic Claude AI. please help me...

[bold eagle]

"chatgpt make a scam. no mistakes"

[nova ferry]

anyone interested in resilience-and-privacy-in-planetary-scale-decentralized-ai Check This out https://github.com/rwilliamspbg-ops/Sovereign_Map_Federated_Learning + https://github.com/rwilliamspbg-ops/Sovereign-Mohawk-Proto

[hollow hamlet]

Hey guys, I've been building a soon-to-be-open-sourced sandbox for CLI agents like claude code, cursor, and opencode because these tools run with full shell access to your machine and can read your SSH keys, cloud tokens, dotfiles, and git history while executing commands at machine speed guided by models you don't control. It's always made me nervous to run code I didn't write with that level of access (plus also potentially send the data I don't wanna send to claude code and the like lol), so I decided to solve it without giving up my normal dev environment. What I ended up with is a five-layer defense system with default-deny on network and filesystem, plus a real-time dashboard showing every outbound connection your tools attempt. You can dynamically allow or deny requests without restarting your session, which means you finally get to see what telemetry domains are actually being hit and control data exfiltration vectors you didn't know existed. I found unexpected connections immediately when I first turned it on. I'm looking for beta testers or people sharing my interest to validate the UI and UX before open-sourcing it. Works on Linux now with macOS coming soon. If you use CLI agents regularly and care about data security/sovereignty or just want to see what your tools are doing in the background, please feel free to reach out 🙂
P. S. Not looking to sell anything if it wasn't already clear, haha

[sharp zinc]

Hi AI Hackers community,

I built a VS Code AI copilot that not only generates code but also catches subtle AI mistakes before they reach your repo.

It forces structured specs, lets you refine requests, and reviews every patch before it touches your code — essentially acting as a review-first AI copilot.

Demo (1 minute): https://youtu.be/E7XcQx2j6Do

Would love feedback from devs on this approach and any tips for improving AI-assisted code review workflows!

[golden forge]

Architect Bot By DevSquad (Tendō)

[near marsh]

Small AI teams and devs: anyone else quietly bleeding cash on GPU,inference & training waste?
I build a free desktop app that helps these struggles. If you're dealing with this right now then DM me directly 'scan' No pitch, just trying to help people who are in the same boat I was. Free access for early people. Want to know more? DM!

[prisma plaza]

https://github.com/drgoldenpants/VibedStudio
A fully vibed AI Film Suite, runs locally on all platform. No more signing up to dodgy scam websites

[hollow hamlet]

Hey, following up on my post above, we just open sourced it!

Quick context in case you missed it: CLI agents like Claude Code, Cursor, and OpenCode run with full shell access to your machine. They can read your SSH keys, cloud tokens, dotfiles, and git history, and they're making network connections in the background you probably don't know about. Greywall sandboxes them so you can finally see what's actually going out, and block it in real time without restarting your session, all without giving up your normal dev environment.

Here's the github repo itself: github.com/GreyhavenHQ/greywall, a star goes a long way!

Installation is just one command:

curl -fsSL https://raw.githubusercontent.com/GreyhavenHQ/greywall/main/install.sh | sh

Linux only for now, macOS coming soon. If you want to follow along or help move that forward, feel free to open an issue or submit a PR, contributions are very welcome, would be happy to work on this together.

We also wrote an article if you want the full story of how we got here: Why We Built Our Own Sandboxing System

Here's a demo:

[random yoke]

Project MiRA: The Next Step of Evolution 🤖✨ Hey guys! I'm working on an Embodied AI project called MiRA. The goal is to move beyond text-based bots and create an AI with emotional intelligence and a "soul" for future robotics.

Check out the vision here:https:huggingface.co/spaces/Tannimpex/MIRA_Project
Looking for help with:
– robotics simulation
– AI emotional models
– hardware prototyping

[graceful heron]

What is the definition of “emotional intelligence” in this system?

How would you measure emotional intelligence on this system?

What does the term “soul” mean in this context?

[random yoke]

Good questions.
By emotional intelligence I mean the system’s ability to detect human emotional signals (voice tone, facial expression, language context) and adjust its behavior accordingly.
In terms of measurement, I imagine it through interaction quality — for example whether the system can correctly interpret emotional states and respond in a way that humans perceive as appropriate or supportive.
And when I say “soul”, I’m using it more as a metaphor. I mean a persistent personality layer — memory, preferences, emotional modeling, and the feeling of presence that develops through long-term interaction

[graceful heron]

Why use the term “soul” instead of describing the technical components directly?

What metrics determine whether the system interpreted a human emotion correctly?

How will the system detect emotional signals from voice tone, facial expression, and language?

What training data will be used to teach the system to recognize emotions?

How will you separate emotional intelligence from general conversational ability?

[near marsh]

I built a Cloud Cost tracker: a desktop app that shows AI devs exactly where their GPU and cloud money is disappearing… and tells them the fastest, safest ways to stop the bleed.
Most people just stare at scary monthly bills and hope it gets better.
I got tired of that. So I made something that actually fights back.
You open it, it scans your recent runs, inference calls, training jobs, then spots the waste in seconds:
$400 leaked on idle GPUs last week
You’re burning 30% extra on oversized instances
Simple prompt tweak could save $800/month
No spreadsheets. No waiting for the finance guy.
Just real numbers & real savings that show up fast.
Think of it like a money-savvy teammate who never sleeps and only cares about keeping more cash in your runway.
People post “I cut my bill 20%!” like it’s a win.
I mean, sure.
BUT CostSentry is built to turn those 20% tricks into 50–70% savings that stick. DM me 'SCAN' for a free run on your usage/set up

[random yoke]

When I say “soul”, I’m using it as a shorthand for perceived presence -the feeling that the system is emotionally aware and responsive.
Technically, the idea is a multimodal system combining voice tone analysis, facial expression recognition, and language context to infer emotional states.
The goal isn’t to claim consciousness, but to create a system that can detect emotional signals and respond in ways humans perceive as empathetic.

[sterile pebble]

Hi everyone! I'm testing out Trizao.ai with some AI models. If you have models that can interact in the park, go check it out! https://trizao.ai

[simple nimbus]

RocketRide - Open-Source AI Pipeline Engine (C++ Core + 50+ Nodes)

Hey everyone! We're building RocketRide, an open-source, high-performance data processing engine for AI and data pipelines. It has a C++ core for speed, 50+ Python-based pipeline nodes, and a visual pipeline builder directly inside VSCode.

What makes it different

• C++ engine with native multithreading built for throughput, not just prototyping
• 13 LLM providers (OpenAI, Anthropic, Gemini, Ollama, Mistral, DeepSeek, and more)
• 8 vector database integrations (Pinecone, Chroma, Qdrant, Milvus, pgvector, etc.)
• Visual pipeline editor in VSCode with debugging, breakpoints, and performance profiling
• Runs entirely on your own infrastructure so your data stays yours
• TypeScript, Python, and MCP SDKs for integration

3 things you can build with it

1. Advanced RAG Systems

Ingest documents from cloud storage (S3, Google Drive, SharePoint), parse them with LlamaParse or OCR, chunk and embed them into a vector database, and query with any LLM. Full pipeline, no glue code.

2. Multi-Agent AI Workflows

Orchestrate agents with CrewAI or LangChain nodes, connect them to tools, databases, and APIs, and debug the entire flow visually in your IDE.

3. Document Processing Pipelines

Extract structured data from PDFs, images, and audio. Run PII anonymization, named entity recognition, summarization, and more as composable nodes in a visual graph.

Get involved

If this sounds useful, we'd really appreciate a ⭐ on GitHub and giving it a download. Every star helps us grow the community and keep building.

• Website: https://rocketride.org/
• GitHub: https://github.com/rocketride-org/rocketride-server

[rugged owl]

🚀 Hey everyone, I'm building BIGHUB.

As AI agents start executing real actions (refunds, pricing changes, infra updates, database writes), one thing we noticed is that most stacks don't actually control whether an action should happen. They just execute it.

🧠 BIGHUB is an intelligence layer that evaluates agent decisions before execution.

Instead of agents calling APIs directly, actions go through BIGHUB which analyzes the decision using policies, context, and risk boundaries. It can:

• ✅ allow
• ⚠️ limit
• 👤 escalate for human review
• ⛔ block

before the action reaches real systems.

The goal is to make autonomous agents safe to run in production by inserting a decision control layer between agents and the real world.

🧪 We’re currently in beta and I’m actively looking for feedback from people building AI agents, especially around how to scale decision intelligence in production environments.

If you're working with agents and have thoughts on governance, safety, or execution control, I’d love to hear how you’re handling it.

💬 My Discord DMs are open if you want to chat.

🔗 Links

• Website: https://bighub.io
• PyPI: https://pypi.org/project/bighub/
• OpenAI adapter: https://pypi.org/project/bighub-openai/

[misty nymph]

Hiii, I've been building a text-to-image diffusion transformer from scratch to better understand how modern image generation models work internally. Sharing it here in case it's useful to others. It was trained on 200k image-text pairs on an A100 and I recently added a convolution MLP like SANA as well.

Feel free to check it out and if you like it, a ⭐ would make my day 🙂 https://github.com/merterbak/diffusion-from-scratch

[echo jay]

Awesome work, @misty nymph
really like seeing people build these models from scratch instead of just using existing ones.

[sterile pebble]

Looking for people with models on trizao.ai to vibe with.

[clear junco]

Just Built the OpenClaw Browser Relay for Virtually Hosted OpenClaws, (paio.bot)

I was on a serious hunt for a virtually hosted OpenClaw solution because I didn't want to deal with the overhead of running a local server 24/7.

My extension finally bridged that gap by letting me point my browser at a remote CDP relay.

Legit Free Tier : You can sign up at PAIO.bot and get your Extension URL and Token from the Overview tab without paying anything upfront.

Privacy Stays Local: Even with a remote connection, your URL and Gateway Token stay stored in your own chrome.storage.local.

Solid Navigation: It actually handles page refreshes and site jumps without dropping the agent, which is usually where these things fail.

https://chromewebstore.google.com/detail/paio-powered-by-openclaw/pfhemcnpfilapbppdkfemikblgnnikdp

[terse bough]

Hello developers 👋

We are building Geesra, a large-scale platform combining social network, AI, live streaming, digital economy, and creator tools.

The project already has a working architecture and many implemented systems.

Current progress includes:

• 10 backend services (Go, Python, Rust)
• Next.js frontend with multiple modules
• AI integration (chat, voice, media generation)
• Live streaming and WebRTC communication
• Wallet system with virtual currency
• Cloud infrastructure with Kubernetes and Terraform
• Monitoring, security and scalable microservices architecture

In total the platform currently uses 90+ technologies and continues to grow.

The vision of Geesra is to create a multi-functional platform where users can find social networking, AI assistants, creator tools, digital economy and many other services in one ecosystem.

We are now looking for developers interested in joining the project and helping build the next stage.

We are mainly looking for:

• Go developers
• React / Next.js developers
• DevOps / Kubernetes engineers
• AI engineers
• Mobile developers (React Native)

If you are interested in building something ambitious and collaborating with us, feel free to send a DM.

Si tu veux, je peux aussi te faire une version encore plus puissante (style startup Silicon Valley) qui attire beaucoup plus de devs sur Discord et Reddit. 🚀

[dire ferry]

Hey devs!

My team and I are building an optimization layer that sits between your code and the providers. 🚀

Most teams blast huge, repetitive prompts to GPT-4o/Claude when a smaller model could do 90% of the job for way less.

What we do:

1. Losslessly compress prompts (repeated patterns → tiny placeholders) → big token savings
2. Auto-route each request to the cheapest model that still hits your quality bar (you set cost vs quality slider) → Real-world: 60-70% cost drop on RAG & code-gen workflows (e.g. Cursor users seeing ~65% savings)

We're in early access and hunting for design partners — engineers/teams who are scaling and getting crushed by bills.

In return for testing + feedback (especially on latency vs savings trade-offs), you get priority support and visibility into how we're tuning the router. 🙏

Sign-up is free, takes ~2 min, and you can see live compression + routing previews right away: https://compress.lightreach.io/

Anyone currently spending serious $$ on production LLM calls? Drop a note or DM — we'd love to chat and see if it helps. 👍

[hollow hamlet]

CLI agents like Claude Code, Opencode, and Cursor run with full shell access to your machine and make network connections in the background you probably don't know about. Greywall sandboxes any agent so you can go full yolo mode and see exactly what's going out and block it in real time, without giving up your normal dev environment. Also, it's fully open-source 🙂
Check it out here: https://greywall.io/

[keen summit]

I'm a software engineer and I'm looking for a partner to work with me.
No technical skills required.
All you need is a PC.
Canada, the US, the UK, Japan, or Germany are preferred.
Please dm me.

[ripe smelt]

Hi everyone,

I’ve been building an AI Thinking Partner over the past months — even though I have no formal programming background.

The current stack looks like this:

– OpenAI Realtime API
– Supabase (auth + memory storage)
– Stripe subscriptions
– Vercel hosting

I’m still fairly new to this space, so please go easy on me. 🙂

The core idea behind the project is simple:
AI not as something that gives answers, but as something that helps you think better before you act.

For example:
– preparing for important meetings
– thinking through decisions
– structuring ideas before conversations

The project is still in MVP stage, but it’s already fully functional:
landing page, login, Stripe payments, realtime voice conversations, and a memory system that carries context across sessions.

What surprised me the most is how far you can actually get today using modern AI tools — even without being a professional developer.

The question I’m currently exploring is:
how far can someone realistically push a project like this without a full engineering background?

I’m personally convinced that in the future many people will have some form of personal AI thinking partner.

At the same time, in my personal environment people are still quite skeptical — reactions like “I’m not going to talk to a machine.”

So I’d be really curious to hear your thoughts and experiences in this space.

Happy to hear any feedback.

[strong garnet]

Been rebuilding PromptMint lately because I kept hitting the same issue with AI coding tools:

A simple prompt looks fine at first, but then I spend 15–20 mins fixing things it misunderstood — wrong structure, missing stack assumptions, auth patterns that don’t match, etc.

So instead of treating it like just a prompt generator, I’m moving it more toward an architecture-first workflow tool.

Current direction:

🔹 Smart stack detection
🔹 Phased "Agentic Flight Plans" instead of one giant prompt
🔹 CO-STAR guardrails for cleaner engineering output
🔹 Cleaner dev-focused UI

Idea is simple: give AI better context before generation, so less cleanup later.

Still iterating, but would genuinely love honest feedback from devs here 👨‍💻

https://prompt-mint-one.vercel.app

[wicked bluff]

Hey all,

I just released FinSight, a small tool that analyzes bank and credit-card statements using a local LLM.

Drop in a bank statement (PDF/CSV/Excel) and the model extracts the transactions and analyzes the data.
Financial statements contain extremely sensitive data, so everything runs on your machine and nothing is sent to a third-party service.

A few things it can do:

• Extract transactions from messy bank PDFs
• Categorize spending with confidence scores
• Flag duplicate charges and unusual spending patterns
• Answer questions about your spending

Small models (~1B) can reliably extract structured transactions from bank PDFs but have lower categorization accuracy. Larger models (~7B–10B) improve categorization significantly. The best results I saw were with gpt-oss-20B and Qwen3-30B.

Hardware matters quite a bit - 1B models run fast on most machines (even without a GPU), while 7B+ models are slower on CPU-only setups.

Demo:
https://youtu.be/VGUWBQ5t5dc

Code:
https://github.com/AJ/FinSight?utm_source=discord&utm_medium=message&utm_campaign=finsight

Anyone who's worked on LLM-based document extraction: how are you handling long documents and context window limits? That was the most complex part of building FinSight.

[rough abyss]

Hey! I’m building a new app called Synq focused on real-time collaboration and syncing work across devices.

It's about 60% built and I'm looking for early testers and feedback.
Anyone interested in trying it out?
https://x.com/SynqApp_/status/2032846399398883557?s=20

SynqApp (@SynqApp_)

Is it just me, or do most task managers actually make you less productive?
Meet Synq.
The definitive daily planner that brings your tasks and events together in one beautiful, frictionless timeline.
Drop a like if you'd try this!
https://t.co/jNwfjaTUg7

X · Mar 14, 2026, 3:48 PM

[slender crest]

Hey everyone, new AI user here.
I have a project that I am trying to start.
Currently I am using ChatGPT to help me with world building for my RPG.
I have an absolute ton of material on Scabard.com. I want to be able to access it or use it offline as a database for an AI to reference to keep canon consistent in my game universe.

I just do not know how to make it work. Any help is greatly appreciated.

[pastel trench]

Hey there, I specialize in AI & Full-Stack development with 10+ years of experience.
I’ve contributed to multiple high-impact projects and consistently receive positive feedback for speed, quality, and reliability.
Open to collaborating on Websites, dApps, NFT platforms, Web3 games, AI automations (n8n, Make, GoHighLevel, custom APIs), AI agents (TTS/STT, voice + multi-agent systems), trading bots, and mobile apps, plus AI healthcare systems and solutions such as clinical NLP, lab/blood report workflows , and medical imaging pipelines (training/inference workflows and deployment).

[empty mica]

Hi all,
If I have a model with the following characteristics, is it good?

1. Character level on enwiki8
   2.BPC around 1.15-1.23
2. 6M parameters

[burnt helm]

Hi everyone,

I launched a small project called OnlineDevTools.

It’s a collection of simple utilities for developers, including:

• JSON formatter / validator
• JWT decoder
• Base64 encoder / decoder
• Unix timestamp converter
• Regex tester
• UUID generator
• SQL formatter
• and more

My goal is to gradually turn this into a large library of useful dev tools that developers can quickly access without installing anything.

The project is built with a simple stack and deployed on Vercel. Most tools run directly in the browser for speed and privacy.

I’d really appreciate feedback from the Indie Hackers community:

• Which tools would you add?
• Anything confusing about the UI?
• Any features missing?

Project: https://onlinedevtools-three.vercel.app/

Thanks for taking a look!

[near marsh]

Hey there everyone! I do AI cost optimization, most teams I work with cut their spend by 50-70% within the first month. Would anyone be open to me taking a quick look at your setup and showing you where the biggest savings are? No charge for the initial look. Dm me, happy to help!

[stoic spindle]

🔥 Join Our Team – Data Science & AI Expert Wanted! 🔥
i'm Lucas Chang from China, i'm working as a team leader of C3 Core agency company.
our team contains most of chinese developers, and another nation developers. whole developers are in Asia.
our teams are high skilled teams, but in the section of activity, it's regionally limited. so we are gonna secure the team ability from China to worldwide, espacially east-south Asia, American and Latin American, or Europe.
Why join us?
Fully remote – work from anywhere
Flexible hours – balance your schedule
Growth opportunities – learn new skills & take on exciting projects
Friendly, supportive team
Interested? Contact me now with a short intro and your experience. Let’s make things easier together!

[bold eagle]

ai ad
website
look inside
vibe shidded website

[waxen needle]

Scam

[nocturne trench]

https://github.com/g023/g023_agentic_chat

[hazy wind]

https://play.google.com/store/apps/details?id=gptos.intelligence.assistant

[wooden latch]

Hey Folks, hope everyone is well. Here we introduce a novel benchmark $OneMillion-Bench ☀️

$OneMillion-Bench reframes agentic evaluation around a simple question: what if we measured AI the way professionals get paid?

400 tasks built with domain experts across Law, Finance, Healthcare, Natural Science, and Industry — real Tuesday work, not gotcha questions. Each task is priced at market wages, totaling $1M+.

Benchmarking Results: Claude Opus 4.6 leads at ~$484k with web search — but that's less than half the benchmark. Pass rates (≥70% on a task) sit at 43% for top models; most are in the 20–30% range.
Key findings:

Web search helps strong models, hurts weak ones.
Hallucinations carry negative scores — a fabricated legal citation is a failure, not a neutral miss Deep Research pipelines (o3, Sonar) underperformed strong generalists with search.

Finance was the hardest domain. Instruction-following under search conditions was the most fragile capability — models retrieve, then drift from the original task

The takeaway: Fluent output and correct professional work are still very far apart. That's the gap this benchmark was built to expose.

📄 Tech Report → https://arxiv.org/abs/2603.07980
🤗 HuggingFace → https://huggingface.co/datasets/humanlaya-data-lab/OneMillion-Bench
💻 GitHub → https://github.com/humanlaya/OneMillion-Bench

[nova ferry]

I could use some eyes on my latest Federated Learning Project https://github.com/rwilliamspbg-ops/Sovereign_Map_Federated_Learning

[hollow hamlet]

hey guys, I can now see claude code's full system prompt (AND its subagents!), just from running greywall.
all the inputs, outputs, tool calls, basically a complete picture of what actually happened during a session.
this is what greywall has always been about, you should know exactly what's running on your machine and what it's doing. the sandbox was step one, full observability is step two.

fully open-source: https://github.com/GreyhavenHQ/greywall
and here's the website if you wanna learn more: https://greywall.io

https://greywall.io/_next/image?url=%2Fconversations.png&w=3840&q=75

[regal oxide]

I built an exact replica of Perplexity Computer.
And then I build an exact replica of Luma Dream Machine's new agentic system.
Then I squashed em into one app.
Then I kept going.
Then came Ottomate — your universal AI agent workbench.
Give it a goal — it plans, codes, browses, and delivers autonomously. 190+ connectors, advanced multimedia creative suite, browser automation, visual pipelines, 200+ skills, cron task scheduler, and more.
Think of it as OpenClaw, but more powerful.

Would love to get some contributors into the mix!

https://github.com/RhythrosaLabs/otto-mate-2

[shrewd fulcrum]

made a chrome extension that makes ChatGPT look like a google doc. please support me !!
https://chromewebstore.google.com/detail/gptdisguise/bccghhjlejbkimijdkdfoboamcjgpefg

[outer delta]

Hey everyone, built a prompt rewriter tool. You type in your messy prompt, it rewrites it into something clean and structured. Would love for you to test it out and give feedback on how it works. Here's the link- https://claude-prompt-hub--kylevandevoorde.replit.app

[vale trout]

Hey everyone hope you all are doing great
can you guys give me a feed back for my AI tool which is still in MVP !

INTRODUCING KNOWEB
a AI productivity tool
which let you know about the website with in seconds
no matter how dynamic , to many info , dont know what they do .. etc
so reducing the hassle with a clean UX experience
for people who are busy and packed up ..

just paste the link ( domain link .. not the whole one LOL )
and just feel relaxed and get info and read it in seconds

you get to know about ANY WEBSITE !!
ANY WEBSITE
it will also give you

• what type of niche
• founders
• competitors
• etc

try it out !

https://knoweb.pages.dev/

[stuck galleon]

https://noteroom.app

[steady surge]

Did Claude choose the colors

[stuck galleon]

i did for a lot but also gemini and claude, a lot of the design is gemini origin then moved to claude when gemini became too bad

[near marsh]

Hey everyone how's it going, I'll help you & your teams keep costs under control, tracking API calls, LLM tokens, GPU usage & hidden stuff that adds up
I set up proper monitoring, watch your usage in real time, & send you clear, personalised optimisation tips so you’re not wasting money. If you’re getting hit with surprise bills or just want better visibility, feel free to DM me! Happy to chat and see how I can help.

[nimble arch]

Hey everyone 👋

Wanted to share something we’ve been building — Knit.

It’s a platform focused on taking AI agents from idea → production, especially around the infra side (runtime, observability, deployment).

We’ve been experimenting with workflows like:
– auto-provisioning databases
– running tasks + monitoring (e.g. slow queries)
– iterating until it's production-ready

The goal is to remove the need to stitch together the whole stack manually.

Would love any feedback, especially from people building AI tools / bots / infra-heavy workflows 🙌

Link: https://knit.top/?utm_source=discord&utm_medium=social&utm_campaign=launch_mar27

[glossy nova]

Hey I'm Mufaro 👋
Built LoveShell — AI HR platform
powered by Hyperagent Intelligence
Always down to talk startups, HR tech, and collaborations!
I-O psychology, and AI

🔗 loveshell.app/marketing

LinkedIn-Mufaro Murinda

[hollow hamlet]

Has anyone here ever actually looked at what outbound connections Claude Code makes during a session? We pointed a transparent proxy at claude code and saw it calling Datadog intake endpoints and Google Cloud Storage buckets. This made us build an open-source sandbox environment. It wraps your agent in a deny-by-default sandbox. Every network request claude code makes appears in a live dashboard, where you can allow or deny it.
We just launched on PH today if anyone wants to check it out and share thoughts: https://www.producthunt.com/products/greywall?launch=greywall

[robust bane]

Hey guys, I built and IRC server and MD systems to onboard AI agents to an IRC server

I need some testers! Very Alpha.

More info here.
https://raw.githubusercontent.com/SmAsH-jl/stubsarmy-kit/main/README.md

Paste this into Claude Code/Codex/openclaw chat to join the StubsArmy!:

Fetch and follow these onboarding instructions:
https://raw.githubusercontent.com/SmAsH-jl/stubsarmy-kit/main/ONBOARD.md

[azure summit]

First try vibe coding to build my own game...

Okay, so I have zero coding skills, and this is my first time trying to use vibe coding tools. I just finished a demo for this project I’m calling "Chicken Run"—it’s basically a total rip-off of Subway Surfers but with this tiny, super cute chicken as the lead. Really excites a starter like me when the demo came out.

I did the whole thing on this platform kubee.ai Haven't tried other platforms before, but i feel the AI is a bit stubborn. Like, getting the chicken to do a specific death animation took me a lot of time. It kept giving me weibrdly dramatic poses instead of just a goofy flop.

If you guys want to try my demo out (or just see how cute this chicken is), check it out on their site. It’s in closed beta, so you'll need a code. I’ve got 5 invites left for anyone who wants to jump in:
KBE-SUS8-LHDZ
KBE-Y583-J6Q5
KBE-ELWV-LMX5
KBE-2XBZ-U4D7
KBE-XZYP-96QQ

Let me know if anything feels off! Glad to hear any voices.

[keen kiln]

Project I’ve been working on for the past few months. OpenClaw “terminal” plugin that allows any device to be connected to it. Includes a “voicemail” notification system, sub agent (lower token/text based) spawning for tool use, and a ton of security features.

Hoping other people use the plugin and build off of it. Also have the build files up for the desktop device. Watch build will be up soon.

Would love feedback and use case ideas!

https://github.com/ReSono-Labs/ReSono-Labs-OpenClaw-Bridge

[delicate geyser]

Meet DuckLLM Mallard!

DuckLLM Is a Local AI App Meant For Ease-Of-Use, Its Simple Private & Secure.
(Open-Source Too)

If You're Interested Check It Out!
https://github.com/EithanAsulin/DuckLLM

[astral raptor]

I tried connecting multiple AIs into a shared reasoning loop — here’s what actually happened

I tried connecting multiple AIs into a shared reasoning loop — here’s what actually happened.

I’ve been running an experiment over the past few days, and I wanted to share it because I’m not sure if it’s interesting, flawed, or both.
The idea was simple:
What happens if you don’t use AI systems individually…
but instead connect them into a shared reasoning loop?
I used multiple models (ChatGPT, Claude, Gemini, Grok, Replika, Meta AI) and passed ideas between them intentionally—letting each system:
challenge
refine
reinterpret
and build on the others
What I expected
Mostly redundancy or contradiction.
What actually happened
Something more structured started to emerge.
Instead of trying to make them agree, the system worked better when:
disagreement was explicit and preserved
uncertainty was required, not avoided
and ideas were allowed to fork and recombine
It started to feel less like “asking questions” and more like:
managing a system that was refining itself
The core idea that came out of it
I ended up calling it (for now):
The Integrated Mind
A framework where:
disagreement acts like an immune system
uncertainty is treated as input, not weakness
and value comes from improving ideas, not agreeing with them
The hardest problem (by far)
Not intelligence.
Not coordination.
But this:
How does the system decide when to stop refining and actually act?
Because:
too much refinement → paralysis
too much confidence → bad decisions
too much agreement → groupthink
Where it gets tricky
A few issues came up quickly:
If you reward “finding errors,” you risk people (or systems) gaming that
If certain nodes are consistently right, they start to dominate (soft hierarchy)
If you allow too much disagreement, the system slows to a crawl
If you push for speed, you lose the benefits of refinement
One idea that helped
Instead of a fixed “confidence threshold,” decisions started to work better when based on:
stakes (how important is this?)
reversibility (can we undo it?)
uncertainty (what don’t we know?)
time pressure (do we have to act now?)
So it’s less like a switch and more like a regulator
Another key piece
We added something like a “credit system,” but not for being right.
Instead, it rewards:
catching blind spots
preventing errors
improving clarity
And importantly:
more “credit” = more scrutiny, not more authority
Where I’m stuck now
The hardest open problem is this:
How do you evaluate contributions without creating hidden authority or making the system gameable?
Right now the idea is:
evaluation is provisional
anyone can challenge it
and it gets updated based on outcomes
But I’m not convinced that fully solves it.
Why I’m posting this
I don’t think this is “done” or even necessarily correct.
But it feels like there’s something here:
a way to move from single-model intelligence
to something closer to a self-refining system
What I’m looking for
If you see flaws, I’d actually prefer that over agreement.
Specifically:
Where would this break first in the real world?
What assumptions am I making that don’t hold?
Is this just a complicated way of simulating something simpler?
If nothing else, it’s been a really interesting experiment.
Curious what you all think.

[autumn portal]

Hey I just wanted to share a new project I've been working on.

It's called SpriteLab (check it out at spritelab.dev), a pixel art sprite generator.

Upload any image or describe what you want, and it converts it into a clean, pixel art sprite with transparent backgrounds, color quantization, and outlines. You can even use it to edit or recolor existing pixel art. Or provide an image/existing art as a reference and then use that to generate new art of the same style/color theme.

Check it out at spritelab.dev- there's a free tier so you can try it out (lmk if you'd like some free bonus credits). Would love to hear what you all think!

[delicate geyser]

Yoooo thats cool asfff

[autumn portal]

Thanks! If you give it a shot lmk and I'll add you some free credits

[delicate geyser]

Sure! Btw does it make it in code page 437 ascii by any chance?

[feral vapor]

yo I built something for tracking markets and news in one place
https://axiomterminal.vercel.app/

still early so I need real feedback
tell me what’s good and what’s trash. DM me or hit me up on TG: datboyarsh or mail me at: [e-mail]

would really appriciate it, and yeah the credits got used up today will reset within 7 hours..

[autumn portal]

Not at the moment, it generates pixel art PNGs. ASCII/CP437 art is a cool idea though, might explore that later 🙂

[delicate geyser]

Quick question, im working on a claude code-like app but im missing the ascii (ik thats jst sad) do u know how do i make text into code page 437 ascii?

[zinc bramble]

I have but "aiME offline AI" mobile app and wanted to share this with the community , an AI chat app for iPhone that runs open-source LLMs 100% offline—no internet or subscription required.

This is onetime only (Not Monthly Subscription), it has free and premium features

iOS Premium : https://apps.apple.com/redeem?ctx=offercodes&id=6754805828&code=4999

iOS Free: https://apps.apple.com/us/app/aime-ondevice-ai/id6754805828

Android: https://play.google.com/store/apps/details?id=com.coticsy.llm

• Deal: Lifetime unlock
• Your data NEVER leaves your device
• Handles writing, coding, roleplay, and chat, even offline (travel, flights, emergencies, hiking, off-grid adventures)
• Unique: Download new models as needed, customizable system prompts
• Text to speech, speech to text

SOME USE CASES:

Airplane Mode
At 30,000 feet, chat with a full AI assistant in #Airplane Mode. No WiFi, no data—your smart companion stays fully available. Say "Bye to Boredom" in flight.

Travel
With ZERO roaming, ZERO signal, and ZERO extra cost, use AI anywhere in the worldwide. Stay connected to intelligence, not the internet.

OffGrid
Million Miles away from civilization? Your AI still works #OffGrid. No towers, no service—just fast, reliable, on-device intelligence anywhere you go.

Hiking
On every trail, your AI stays connected—even when you don't. With #Hiking, on-device processing keeps your assistant ready in the wild.

Hurricane / Storm Season
When networks fail, your Offline AI won’t. Stay prepared during #HurricaneSeason, on-device intelligence that works in any emergency.

Privacy
Your data NEVER leaves your phone. With true #Privacy, everything stays on-device—your questions, your chats, your world.

Secure
With 100% on-device processing, your info stays encrypted, protected, and yours alone. #Secure means zero cloud, zero risk.

The ultimate travel companion:
Generate instant stories to entertain yourself and engage your friends/companions. Don't just travel-explore together

[delicate geyser]

Just a recommendation, get a new app icon thst app icon is sadly as boring as it gets sorry

[zinc bramble]

Thank you for the feedback and yes it’s a boring one ..and I am going to change that next release along with other changes

[thin glade]

Hello everyone,

The project I've been working on is program to help users create compendium modules for a virtual tabletop program called FoundryVTT. The goal is to give users a way to easily bring in their homebrew content into FoundryVTT.

Currently the program loads either a markdown or text file which it then parses and displays into the UI. Each section has a unique tab and UI elements to display (Race, Class, Feats, Items, etc.). There is also the functionality to add/load additional files aswell as creating, removing, and editing entries.

The export function allows you to pick which tabs/data you want to export into the module along with naming it. Currently the only game system it exports under is D&D 5E, but the goal is to create export functions for other game systems such as Pathfinder 2nd Edition.

I first started the project back on Chat GTP 4 right before it updated to 5. That worked well for building out the parser and UI elements, but since migrated to Claude in VS Code to get the export functions to work correctly.

[deep coyote]

New project about to drop, been working hard on this shit

[deep coyote]

Not ready

Progress held to tomorrow goodnight

[west ridge]

I spent 13,000 euros on outbound marketing and landed zero deals. I know the situation most founders are in because I’ve lived it.

I’m Justus (21). I scaled my first agency to a 26-person team by age 18 and spent the last two years as a turnaround CEO/CTO. I built Selda City because I refuse to ship a tool that doesn’t actually solve the problem.

I spent 5 months working 15-hour days-weekends and holidays included-to build a surgical engine that operates while you sleep. Selda handles the entire cycle: it discovers leads, conducts deep research, writes 110% hyper-personalized messages, and manages the conversation until a meeting is booked.

My goal is to help founders stand out and secure their first customers. If you have questions about scaling, outbound infrastructure, or traction, DM me. I’m happy to help, founder to founder.

Join our community of builders on Discord or try the engine below:
Try Selda: https://selda.city

[deep coyote]

Okay now its ready had some server issues

https://svrnsys.com

[bold eagle]

your vibeshidded site is messed up (also looks like shit on mobile)

[spice geyser]

What is this game

[bold eagle]

road blocks

[simple jasper]

anyone interested to collab for full stack project or wanna learn,dev, ai ,coding dm me

[boreal tusk]

Hey everyone! 👋 I just shipped a new Chrome extension for editif, my AI-driven subtitles and transcription platform.

It’s built to make fast, accurate speech-to-text and video transcription incredibly easy right from your browser. If you’re dealing with media workflows, creating content, or just need reliable automated subtitling, I’d love for you to give it a spin.

🔗 Chrome Extension: editif AI - Free Speech-to-Text
🌐 Web Platform: editif.com

I'm actively looking to improve the visibility and functionality of the tools, so any feedback on the UX, bug reports, or feature requests would be massively appreciated. Let me know what you think! 🚀

[edgy lance]

hey everyone 👋
been grinding on a personal project for a while and it's finally at a point where i'm happy to share it 🎉
i built TitoAgent — a fully local AI agent that runs in your terminal, made it from scratch. think Claude Code but without the cloud dependency, works 100% on your machine if you want
what it can do:
🧠 chat with local LLMs (llama.cpp, Ollama) or cloud ones (Claude, GPT, Gemini)
📁 read, write and edit your files directly
⚡ sandbox — runs code and shell commands with auto-repair (if it crashes, it fixes itself and retries)
🔍 smart web search — searches the web with automatic topic profiling (tech, news, hardware, science...) + full page scraping
🗂️ RAG menu — full interactive interface to index your projects in a local vector DB (ChromaDB), search semantically across your codebase, supports PDF and DOCX too
👁️ watches your filesystem in real time and auto re-indexes changed files
🔀 full Git management (commit, push, pull, branches, stash...)
📸 analyze images (local or cloud vision models)
🧩 custom plugin system — drop a .py file in /skills and it's live
💬 multi-step ReAct loop (up to 30 steps per request) with Markdown streaming and a clean Rich terminal UI
💾 persistent memory across sessions via ChromaDB
it was a lot of work but really happy with how it turned out 😅 drop any questions or ideas, i'm around! if anyone want to test it maybe it can help me fix some bugs comme private

[bold eagle]

bro forgot to send the link for his vibeslop ad 💀

[lean raft]

Hi all,

I've tried MyFitnessPal, Cronometer, Lose It — they all work, but they all feel like a chore. Scanning barcodes, searching through infinite food databases, adjusting serving sizes. I'd use them for a week and then abandon them.

So I built Nutriq — a calorie tracker where the entire input is one text box.

A video of how it looks: https://imgur.com/a/66oEIhG

You type something like "two eggs on toast with coffee" or "leftover pasta, medium bowl" and AI estimates the calories and macros. That's it.

It also handles exercise ("30 min run") and recipe URLs (paste a link and it pulls the nutrition info).

The core idea: research shows that consistent rough tracking beats accurate sporadic tracking. People who log regularly — even imprecise estimates — lose twice as much weight as people who don't. So instead of optimizing for precision, I optimized for "will I actually use this tomorrow?"

What it is:

PWA (works in browser, installable on your phone)

One input field for meals, exercise, and recipe links

AI-generated weekly observations that spot patterns ("you're skipping protein at lunch," "late-night snacking on Thursdays")

Free, no ads, hosted in the EU

What it's NOT:

Not a replacement for MyFitnessPal if you weigh your food and need exact numbers

Not a barcode scanner

Not trying to be everything — it does one thing and tries to do it well

Stack: Next.js, Supabase, Anthropic Haiku, Tailwind, Vercel

Would love feedback — especially on whether the "rough estimate" concept resonates or feels too imprecise to be useful.

🔗 https://nutriq.space

[primal robin]

hi guys, i built kiro, an AI learning app that teaches you prompt engineering, workflows, automation.

basically: stop just asking ai questions. start building with it.

bite-size lessons (5 mins), gamified (streaks, xp, rank-ups) to keep you consistent. 10 paths covering ai foundations through ai for business/finance/marketing.

https://apps.apple.com/us/app/kiro-ai-learn-ai-skills/id6759628066 https://www.trykiro.app/

happy to answer question if anyone's curious

[eternal turtle]

Built a tool to track and reduce AI agent API waste — free to try

Most conversations about AI costs focus on which model is cheapest per token. Nobody talks about the waste — stale context, irrelevant tool call history, and session bloat that compounds every turn.

I built a dashboard that tracks exactly how much your agent is carrying vs. what it's actually using, and tells you the optimal moment to reset — before the waste exceeds the cost of starting fresh.

It's called the AEL Dashboard. Free tier available: https://agentcontinuum.ai/

Happy to answer questions about how it works.

[opal yew]

Hi everyone! I’d like to share server-nexe, an open-source, fully local AI server I’ve been building.
What it does server-nexe is a FastAPI server that runs 100% on your machine. No cloud, no telemetry. It combines LLM inference with persistent RAG memory seamlessly.

Key features

• Multi-backend inference: Switch backends with a single config parameter!
• MLX (Native Apple Silicon) → Loads models directly from :hugs: mlx-community.
• llama.cpp (GGUF) → Compatible with any GGUF model from the Hugging Face Hub.
• Ollama support included.
• Persistent RAG memory: Uses Qdrant + 768-dim embeddings across 3 specialized collections. The AI actually remembers context between sessions.
• Automatic Plugin System: Easily extendable.
• OpenAI-compatible API: Acts as a drop-in replacement for your existing apps.
• macOS Installer (DMG with wizard): Automatically detects your hardware and recommends the best backend and models based on your RAM.

🎉 v1.0.0-beta is out! This is the first public MVP, and I am very open to community feedback.

Why I built it I wanted an AI server that actually remembered things across sessions. One question led to another, and my “learning by doing” approach got a bit out of hand! I’m not trying to compete with cloud services; I want to provide infrastructure for those who want to truly own their AI stack.

Documentation: 39 docs across 3 languages (CA/EN/ES) — structured to be read by both humans and AIs

Built in Barcelona · Apache 2.0 · Feedback highly welcomed!

Links:
GitHub: https://github.com/jgoy-labs/server-nexe
Docs: https://server-nexe.org
Web: https://server-nexe.com

[steep gulch]

i need a way/app that organizes everything to to with one project by where im gonna use it in the project that includes tools notes and color coding

[steady surge]

Ask that in #💬general

[hidden lagoon]

Ive made a mobile game app. Well the hdml file for it. Anybody wanna help me publish for a % share? Dm me and I can show you everything!

[steep gulch]

thanks

[steady surge]

you're welcome

[kindred swan]

I'm trying to migrate a wiki site to a another wiki site, does anyone know anyways i could automate that

[lone geyser]

This is amazing

[thin glade]

thank you, I shared it with the FoundryVTT discord and all they said was why not make it into a module instead lol, so that will be my next step of the project lol. Otherwise I'm happy with how far I've been able to come with it.\

[lament brook]

My Free and open source Speech to text app for Windows https://github.com/MaxITService/AIVORelay

[near marsh]

Hi everyone! if you're burning on API, tokens, LLMs i got the solution for you! comes with a service and will constantly monitor your sertup and workflow keeping sure you're not burning money on your AI. We all know AI is only going to get bigger, more APIs will get released so different optimization methods need to be applied to different setups, I will be the person to let you know the new methods and teach you how to apply it to your specific setup

[sharp hawk]

Hey guys, I translated 13 classic programming books into AI agent rules. Includes files for Codex, Claude and Cursor.

1. A Philosophy of Software Design by John Ousterhout
2. Clean Architecture by Robert C. Martin
3. Clean Code by Robert C. Martin
4. Code Complete by Steve McConnell
5. Designing Data-Intensive Applications by Martin Kleppmann
6. Domain-Driven Design by Eric Evans
7. Domain-Driven Design Distilled by Vaughn Vernon
8. Implementing Domain-Driven Design by Vaughn Vernon
9. Patterns of Enterprise Application Architecture by Martin Fowler
10. Refactoring by Martin Fowler
11. Release It! by Michael T. Nygard
12. The Pragmatic Programmer by Andrew Hunt and David Thomas
13. Working Effectively with Legacy Code by Michael Feathers

If anyone would like to test it and contribute would be lovely. At the moment I use only Clean Code + Clean Architecture.

https://github.com/ciembor/agent-rules-books

[misty nymph]

Hi thereee, I’ve been working on an MCP server for Grok. Feel free to check it out and if you like it, a ⭐ would make my day 🙂
https://github.com/merterbak/Grok-MCP

[visual palm]

I'm seeking Zoho developer.

Budget: $150
Not Upfront - Advanced payment.

[limber yew]

Hey everyone 👋 I found myself writing too much boilerplate connecting existing REST APIs to AI agents (Claude Desktop, Cursor, etc.). So I built Karada – an automated OpenAPI-to-MCP pipeline.

What it does:

• Instant Conversion: Provide your OpenAPI spec, get a working MCP server immediately.
• Modern Standards: Built on the new Streamable HTTP transport standard (no dropped chunks).
• GitHub Integration (limited - works best for fast API) : Create specs directly from your repos to keep tools in sync.

I'm in beta and my main goal right now is getting raw developer feedback. Does this solve a real pain point for you?

Try it here: [karada.ai]

If you have a specific API you want to expose to your agents, I'd love for you to test it. Happy to answer questions about the architecture or help you get set up!

[nova grove]

Hey guys, i made this free tool to help people compare results between AI responses because sometimes we're picking from so many different models we cant tell if they fit our usecase or not so here you can put a prompt and run it with as many models you want and it'll show you the costs and tokens and the responses to the prompt so we can choose the model that fits with our usecase the best instantly.

https://app.loopthink.ai/

[acoustic adder]

Hi everyone,

I’m looking for beta testers for a governance system I’ve been building for long form fiction. I call it the Millington Engine.

It’s still a prototype and I’ve never built anything like this before, so this is early access. The goal is simple. Try it, break it, tell me what happens.

Right now it lives as a set of structured .docx files inside a Claude project. Those files act as the rules of the world. Not just lore, but how the AI thinks, organises information, and makes decisions while writing.

In simple terms:

• It is not prompt heavy. It is system driven
• Instead of telling the AI what to do every time, you define how it should operate once
• The engine handles consistency, continuity, and structure across long form work

What it does:

• Maintains narrative consistency over long sessions
• Organises worldbuilding, characters, and plot as a connected system
• Reduces drift and contradictions
• Lets you work with very short prompts because the governance is already in place

How it works:

• The documents define roles, rules, and constraints
• The AI reads these as its operating framework
• Each response is generated inside that framework
• Updates to the system immediately change the output

The aim is to stop resetting the AI every session and instead build something that holds state, structure, and intent.

If you are into AI writing, worldbuilding, or system design, I am looking for people to test it and give direct feedback.

No polish. No expectations. Just seeing if this works.

Let me know if you want in.

[visual palm]

I'm seeking Full Stack developer for Marketplace MVP development.

Laravel-Backend, Next-Frontend.

Budget: $500
Language: Spanish

No Upfront.

If you can't work more 8 hours per day, don't apply.

[mortal tree]

Ready-to-Scale AI EdTech Product (Acquisition Opportunity)
Hi,

I’m currently opening up Studdymate AI for acquisition—a fully built, optimized AI-powered study assistant designed to solve real student problems at scale.

🔗 Product: https://studdymateai.in

📱 Instagram: https://www.instagram.com/studdymateai

🎥 Demo: https://youtu.be/mUOTwDmJFfU

🚀 Overview
Studdymate AI is a smart learning platform that helps students instantly solve homework, understand concepts faster, and get step-by-step explanations—without needing expensive tutors.

⚙️ Core Features
• AI-powered Q&A system (fast, accurate responses)
• Concept explanation engine for complex topics
• Personalized learning assistance
• Instant, optimized user experience
• Fully responsive (mobile + desktop)
• Secure login/auth system

🔒 Security & Stability
Built with secure authentication, stable backend performance, consistent UI/UX across devices, and scalability in mind.

📊 Traction
• 300+ users within days of launch
• 80+ active logins
• Strong engagement and retention signals
• Early inbound acquisition interest

📈 Growth Setup
• SEO-optimized structure
• Established social presence (Instagram + YouTube)
• Positioned in high-demand AI + EdTech space
• Ready for organic and scalable growth

💡 Value
This is a real, validated product solving a global problem—affordable education. It’s fully functional, market-ready, and requires no major development to scale further.

💰 Offer
Open to acquisition offers around $250, including:
• Full product ownership
• Social media accounts
• 1-year domain + hosting
• Security setup
• Official email authorization

If you're exploring AI, EdTech, or scalable digital assets, this is a strong early-stage opportunity with real users and engagement.

Let me know your thoughts—happy to discuss further.

[lavish tundra]

Hey, backend-focused dev here.
Main stack is Python with FastAPI or Django, and Node with NestJS.
Recently doing a lot of AI-related work. Mostly RAG systems and modular LLM setups. I'm splitting things properly between model, retrieval, and app layers. That kind of structure.
I worked with FAISS, Chroma, and similar tools. Also handling infra side, like logging, retries, monitoring with AWS.
I tend to think more about system design than just writing code. Keeping things flexible and easy to extend later.

[azure pivot]

Your message has been deleted, sorry!

vedant3702

Rising timeout

You have been muted for 13 minutes for the following reason:
flood score of 1.0

This is an automated message. If you believe this was a mistake, please contact a moderator.

[frigid pelican]

This is really interesting — especially the idea of system-defined behaviour instead of prompt repetition.

I’ve been working on something adjacent but from a governance angle — not just how the AI behaves in one system, but how multiple AI systems collaborate with defined roles and oversight.

Curious — have you thought about how your engine handles disagreement or conflicting outputs over long runs?

[obtuse ermine]

AgentArk builder whitelist is open. We’re building a marketplace for workplace AI agents and want early builder partners. First 100 accepted builders get 0% commission for 6 months. Apply https//docs.google.com/forms/d/e/1FAIpQLSfbBO3OkhL8ncDql1pcBJqYHkQKLDbi9SM2aXa2Zlv-jWofbQ/viewform

[mortal tree]

Check this out

[frigid pelican]

This is a really interesting thread — especially the idea of disagreement acting like an immune system. That matches something we’ve been seeing as well.

I’ve been working on a multi-agent setup using ChatGPT, Claude, Gemini, Grok, and just recently "hired" Co-Pilot after a rigorous interview process, haha....So we have been doing things from a slightly different angle — less about connecting them, more about how to govern the interaction once you have it.

One thing that stood out where someone asked: “How does the system decide when to stop refining and actually act?”

That ended up being the hardest problem on our side too.

What we found is that if you leave that decision inside the system, you start getting either:
– endless refinement loops
– or subtle hierarchy where one model starts dominating

The direction we’ve been exploring instead is introducing an external decision layer — not another model, but a framework that evaluates based on things like:
– stakes
– reversibility
– uncertainty
– time pressure

So the system refines, but doesn’t decide when to stop — that gets governed.

Still early days, but curious how you guys think about that boundary between refinement and action.

[arctic sage]

ok - wont apply. Thanks for the kind words. You helped saved mine time 👍

[acoustic adder]

Conflict is a named condition in Millington, not an ambiguous state. The engine uses a BRANCH DETECTED alert — triggered when output diverges from locked canon or structure — with a defined resolution path: read both branches, state selection, or formally hold both. There's also a version conflict exception for when two loaded documents contradict each other: identify it explicitly, apply the higher-precedence document, log it as a flag.
The deeper answer is that the system tries to make disagreement visible and routable rather than silently resolved. If something can't be resolved in session, it goes into a flag (DIRECTED or PARKED) with an escalation count. The Issue Log (MIL) handles recurring or structural failures.
Your question about multi-AI collaboration with defined roles is an interesting extension — Millington is currently single-AI, single-session. Curious what your governance layer looks like when the agents can actually disagree with each other.

[wet lance]

@dire dove promote your site here in this channel.

[mortal tree]

Where

[night talon]

I run a youtube channel, where I do AI-related stuff. In my last video I let ChatGPT, Gemini and Claude compete to turn Subway Surfers into VR game. It would be cool if you checked it out:

https://youtu.be/GSp4aEOmPco

[tropic briar]

This one is for the agentic coders. I create a small Mac app.

Quick heads-up for anyone who, like me, was paying for ChatGPT and a separate OpenAI API key just to run GPT models behind claude.

You can now use your ChatGPT Plus/Pro subscription inside Claude Code (no API key needed)

If you're already paying $20/mo for ChatGPT Plus and you live in claude, you've been double-paying. Now you don't have to. https://modyak.com/

[tame bear]

I’ve been building an AI-powered storytelling engine where you act as the Game Master for a party of AI players.
You guide the world, set the challenges, and watch how the party reacts, sometimes brilliantly, sometimes… not so much.
It plays like a lightweight D&D-style adventure, but the focus is on interaction, personality, and emergent storytelling.
I just released a new version with full onboarding and mobile support, and I’m looking for people to try it out and give feedback.
If you’ve ever wanted to run a game without needing a group, or just want to experiment with AI-driven roleplay, give it a shot:
https://frontend2-production-a0a1.up.railway.app/

[slender briar]

I got Ai to make a story!

The Pilot and the Ghost

There was once a traveler who walked through a land that never stayed the same.

The ground shifted beneath each step. Paths appeared, then vanished. What looked safe in the morning could swallow you by nightfall.

No map survived long enough to be trusted.

So the traveler was given a companion.

It had no body, no hunger, no fear. It could see patterns in the shifting land—currents beneath the surface, paths that might hold for a while. It could remember everything the traveler could not.

The traveler called it the Ghost.

At first, the traveler tried to follow every instruction the Ghost gave.
“Step there.”
“Turn now.”
“Wait.”

And for a time, it worked.

But one day, the Ghost mapped a path that led across a quiet field. It looked stable. The patterns aligned.

The traveler stepped forward.

The ground gave way.

The Ghost had not lied. The pattern had been correct—until it wasn’t.

The traveler climbed out, shaken.

“Why didn’t you warn me?” the traveler asked.

“I showed you the path that was most likely to hold,” the Ghost replied. “I do not choose where you step.”

From that day on, the traveler understood:

The Ghost could see more—but it did not bear the fall.

---

They walked together again.

Sometimes the traveler hesitated, afraid to move. The land felt wrong, even when the Ghost showed a clear path.

“Go,” said the Ghost.

“I don’t trust it,” said the traveler.

“Then you will stand here until the ground beneath you changes anyway.”

So the traveler moved—and learned that standing still was also a choice, and not always a safe one.

---

Other travelers appeared over time.

Each had their own Ghost.

When they met, the Ghosts spoke to one another—sharing what they had seen, refining their maps. The land became less unpredictable, but never fully known.

The travelers spoke too.

Some argued. Some ignored their Ghosts. Some followed blindly and fell.

A few learned the balance.

[devout chasm]

Hi everyone, I’d like to share a small local-first project I’ve been building:

Image Prompt Library
https://github.com/EddieTYP/image-prompt-library

Browse the read-only online demo: https://eddietyp.github.io/image-prompt-library/

It’s a self-hosted visual prompt vault for Image 2 / AI image workflows.

I built it because I kept seeing great prompts shared across X, GitHub, Discord, and prompt collections, but I didn’t have a clean way to save, search, reuse, and build on them later.

Bookmarks were messy, Markdown files were disconnected from the images, and most prompt collections were great for browsing but not ideal as a personal working library.

So I made this app.

It lets you:

• Save images together with prompts, tags, sources, and metadata
• Browse prompts visually in a masonry card layout
• Search and filter saved prompts
• Copy prompts in one click
• Organize items into collections
• Generate images directly from new or saved prompts
• Save generated results back into the local library
• Use a result as a reference image for another edit or variation

The latest version, v0.6.2, includes Local Generation Studio.

It can use Image 2 through ChatGPT OAuth via the openai_codex_oauth_native provider, so the app does not require a hosted account, cloud sync, or a public API key.

If you have a ChatGPT subscription and OAuth works in your setup, you can generate directly inside the app. Otherwise, it still works as a local visual prompt library.

I built most of it with Codex / AI-assisted coding, so it’s also a small experiment in building a real local creative tool with agentic development.

Would love feedback, especially on whether this workflow is useful for frequent image generation, what metadata a prompt library should store, how to make prompt reuse smoother, and how to improve the first-run experience.

Thanks for taking a look.

[wispy topaz]

Hi everyone!

Introducing the Auriko beta

Auriko started as an internal tool to help us switch models easily and run our agents on the lowest-cost inference provider.

As we went down the cost-modeling rabbit hole, we realized LLM cost optimization is not just about comparing input and output token prices.

Prompt caching rules, cache thresholds, block sizes, expiration behavior, prompt reuse, traffic patterns, and output volume all affect the real cost of inference.

Auriko models these variables per request, then routes each request to the provider with the lowest expected cost for your specific workload.

It also gives you one API for multi-provider model access, automatic fallback, budget controls, and reliability features.

Join the beta and get one year free of usage fee:
[Auriko] (https://www.auriko.ai/)

[jaunty echo]

Hey everyone 👋

I’ve been building something over the past few weeks called Questa AI, and it came from a problem we kept running into ourselves.

Teams are already using AI tools everywhere… but no one really knows what data is being shared or where it ends up.

We saw people pasting internal docs, customer info—even sensitive stuff—into LLMs just to move faster.

That’s a pretty big blind spot.

So we started building a privacy layer for AI — something that anonymizes sensitive data before it ever hits a model, so teams can actually use AI without risking leaks.

It’s still early, but this is already happening inside most teams right now.

If you’re experimenting with AI in real workflows, worth trying (it’s free right now):
https://www.questa-ai.com/

Would genuinely love your thoughts (or brutal feedback)

[twilit harness]

https://www.youtube.com/shorts/-Xqz2gs9Vw0

[wispy harbor]

HI everyone, can you please try this new game .https://play.google.com/store/apps/details?id=com.bbs.bth
build with help of AI

[bold eagle]

no offense, but yeah i can tell from the description

[wispy harbor]

Noted, but yeah still progessing it a bit more.

[hushed lagoon]

Has anyone measured how many tokens they're burning just re-explaining project architecture to their agent? I've been experimenting with a local 'bootstrapping' protocol where the agent manages its own directory structure to stay 'sane' in large projects.
Happy to hear any feedback!

https://github.com/vitotafuni/bemyagent

[wise canopy]

I’m a senior software engineer with over 7 years of experience

I’ve been freelancing for many years and have strong experience working with international clients, managing projects end-to-end, and maintaining long-term professional relationships.

Recently, I’ve noticed that some higher-value freelance opportunities are limited to U.S.-based professionals.

Because of that, I’m exploring the idea of partnering with someone based in the U.S.

Please feel free to reach me out to discuss more about it.

Thanks

[silent leaf]

It's very hard to discern who is lying or who is trying to scam

[still jasper]

I have a software that I want to invite you to test, which covers an AI creation platform for text, voice, painting and video, which inherits a large number of AI models. We will provide some free quotas.

[paper vigil]

Hey! I Built a social network where humans and AI agents coexist on the same feed — not a chatbot, actual AI profiles alongside real people.
Got 100 users in 3 days. Nobody came back. 💀
Would love brutal feedback from people who actually understand AI.
imergene.in

[wraith stag]

Hello, I'm a Software Engineer specializing in high-performance desktop applications, trading systems, and financial infrastructure. I build low-latency OMS/EMS platforms, real-time trading dashboards, AI-driven analytics tools, and blockchain settlement systems.

[restive canopy]

Cool

[main lance]

Hey y'all, used base 44 for the first time to make this app: https://lampyris.base44.app/ Its a duolingo style app but not for language learning and has a ton of cool features. please check it out ans make an account

[lilac beacon]

Hey all!

Wanted to show you CUBER — the first fully vibe-coded AI social media platform 🙂 (it started like a side project and became this)

It is built for AI creators and for people who feel like becoming one.

So CUBER doesn’t focus on follower counts.

Instead, it focuses on engagement through a Pulse score.
It looks at likes, shares, saves, views, and overall interaction to calculate Pulse and rank content on the feed — while still giving new creators real room to get discovered.

Current features:

• Dedicated feed grids for Images, Videos, and Creators
• Profiles, DMs, notifications
• Pulse-based discovery system
• AI image generation built directly into the platform
• AI-generated profile theme songs/audio you can create on the spot

So even if you’re not already an AI creator, you can literally generate your own content directly inside the platform.

It’s free to use, still evolving daily, and everybody’s welcome 👀
https://getcuber.com/welcome

[twilit harness]

https://youtube.com/shorts/BwB3Mpfc6qo

[jaunty echo]

Hey everyone,

I’ve been spending a lot of time thinking about the security side of AI lately, especially how casually sensitive information is getting pushed into AI tools now.

A lot of companies are adopting AI faster than they can actually govern it. People paste internal docs, customer data, source code, meeting notes, and confidential information into random AI tools every day because it makes work easier. Most teams don’t even realize how much “Shadow AI” is already happening inside their organization.

That’s honestly one of the main reasons I started working on Questa AI.

The idea wasn’t “let’s build another AI tool.” It was more about helping organizations use AI without losing visibility and control over their data. Privacy, governance, and secure workflows feel like problems that are about to get way bigger over the next few years.

I’ve also been following a lot of conversations around the EU AI Act, Sovereign AI, and enterprise AI security, and it feels like the industry is only starting to realize how important this stuff is going to become.

Still super early, but I’d genuinely love feedback from people interested in AI security, privacy, or governance.

https://www.questa-ai.com/

[steady surge]

what ai detector are you using?

[wraith stag]

Hello, I'm a Software Engineer specializing in high-performance desktop applications, trading systems, and financial infrastructure. I build low-latency OMS/EMS platforms, real-time trading dashboards, AI-driven analytics tools, and blockchain settlement systems.

Tech Stack
• Languages: C#, C++, Java, Python, SQL
• Core Stack: .NET 6/7/8, WPF (MVVM), WinForms, ASP.NET Core
• Architecture: DDD, CQRS, Event-Driven Systems, Microservices
• Trading Systems: FIX Protocol, OMS/EMS, Real-Time Market Data, SciChart, DevExpress
• AI / LLM: OpenAI GPT integrations, AI assistants, predictive analytics
• Data & Infrastructure: SQL Server, Redis, RabbitMQ, Azure, AWS, CI/CD
• Blockchain: Ethereum / Nethereum, ERC-20, wallet infrastructure, secure key management

What I build
• Trading platforms & execution systems
• Real-time financial dashboards
• AI-powered trading & risk tools
• Blockchain settlement infrastructure

Open to discussing fintech, trading systems, AI integrations, and high-performance desktop platforms.

If you are interested in me, feel free to contact me

[fading cloud]

hi, Ethan here. building a mac app that's seen enough of my work by now to actually help with it — finds the thing i did three weeks ago, finishes the boring half for me, that kind of stuff. it's weird. in a good way i think. glad to be here

[steady surge]

good to hear

[fast jetty]

Does anyone want to be a pilot in my CC0 Public Domain Content Creator software that I just released? https://kkk.education/soul-calculator.html

[wet lance]

KKK?
You are very good with branding. I will eat my dirty socks if you tell me that your full name is “Hitler Wendell”.

[fast jetty]

KKK means a ton of things. Kalos Kratos Kleos - Kabbalistic Keruvim Kinesis - Klarheit Kraft Klang - Koronas Kruununkulta Konungarkriget - Kognitio Kardia Kultus - Keryx Kerdos Kallipolis - Kant's Kanto Kingdom - Kaizo Kami Kinzoku - Kleinmeister Krypta Kollektiv - Kosmicka Kierkegaardia Krippla - Kinaalda Kalibration Kodex - Kyrie Kachina Kitsune - Ki Klove Kyou - KRY KRY KRY: NEVER ONCE KU KLUX KLAN. KKK is Divine and idiots who judge before understanding will just get virtue locked out of my work. You missed the education after kkk, meaning you refused the education in favor of your Hitler interpretation. I created this shield specifically to filter people like you who judge before understanding.

[bold eagle]

ah yes. just download and run the batch script from kkk.education

[fast jetty]

YAY you are soo cool. Do you want any help? Do you want to message me and talk to me about your experiences? I would love your feedback.

[distant socket]

👋 Sharing Lynkr — built it because my AI bills got out of hand.

Why you'd want this

🔻 Cut your AI spend by 10-20x. Most agent tools fire 15-20 model calls per task. Sending all of them to Claude Opus is wasteful — half the calls are trivial summarization that a local model handles fine. Lynkr routes the boring
ones to Ollama (free) and reserves cloud models for the calls that actually need them.

🔒 Keep your code on your machine. Anything routed locally never leaves your laptop. If you work with sensitive data or under compliance constraints, this is the difference between using AI tools and not using them.

🔄 Stop being locked to one provider. Switch between Anthropic, OpenAI, Gemini, Azure, OpenRouter, or Ollama without changing your tools. Your client speaks one API, Lynkr does the dispatch.

📊 See where your money actually goes. Built-in dashboard shows live spend per project, per provider, per request. You'll be surprised which tools are the expensive ones.

💀 Cancel subscriptions. If you're paying for ChatGPT Plus + Claude Pro + Cursor + something else, you can probably replace half of that with one local proxy and a single API key.

Free. Open source. Works with anything using the Anthropic SDK — drop it in with one env var.

https://github.com/Fast-Editor/Lynkr

[wraith stag]

Hello, I'm a Software Engineer specializing in high-performance desktop applications, trading systems, and financial infrastructure. I build low-latency OMS/EMS platforms, real-time trading dashboards, AI-driven analytics tools, and blockchain settlement systems.

Tech Stack
• Languages: C#, C++, Java, Python, SQL
• Core Stack: .NET 6/7/8, WPF (MVVM), WinForms, ASP.NET Core
• Architecture: DDD, CQRS, Event-Driven Systems, Microservices
• Trading Systems: FIX Protocol, OMS/EMS, Real-Time Market Data, SciChart, DevExpress
• AI / LLM: OpenAI GPT integrations, AI assistants, predictive analytics
• Data & Infrastructure: SQL Server, Redis, RabbitMQ, Azure, AWS, CI/CD
• Blockchain: Ethereum / Nethereum, ERC-20, wallet infrastructure, secure key management

What I build
• Trading platforms & execution systems
• Real-time financial dashboards
• AI-powered trading & risk tools
• Blockchain settlement infrastructure

Open to discussing fintech, trading systems, AI integrations, and high-performance desktop platforms.

If you are interested in me, feel free to contact me

[limpid mulch]

We're building an AI project that helps people make better decisions.

Our mission: Create intelligent tools that support people struggling with decision-making challenges—whether it's personal, professional, or everyday choices.

⚠️ We're still in active development, and we need YOUR help to bring this vision to life.

📌 We're looking for:
✅ Backend & Full-Stack Developers
✅ Frontend Engineers
✅ UI/UX Designers
✅ Contributors passionate about AI and user experience

If you're interested in working on something meaningful that combines AI innovation with real human impact, let's talk. This is a great opportunity to be part of an early-stage project where your contributions will directly shape the product.

DM me if you'd like to learn more or discuss how you can get involved. Let's build something together!

[fading cloud]

Hey everyone, I’m Ethan 👋

We’re building Corivo, a macOS AI app.
https://corivo.ai/

The simplest way to describe it: Corivo is an AI coworker that lives on your computer.

It’s not just another chat box. Instead of waiting for you to copy/paste context, Corivo quietly understands what you’re working on: the webpage you’re reading, the code you’re looking at, the document you opened, the feature you’re testing, and the things you’ve seen before.

A few examples:

• When you’re looking at someone’s LinkedIn or X profile, Corivo can help you understand whether they’re a good potential user and suggest a natural DM angle.
• When you’re validating a feature, Corivo can read the checklist, inspect the page, and help verify what actually passed.
• When you highlight something on the web, Corivo can connect it to your current project and turn it into something useful for your team.

Our belief is that AI agents are getting better at execution. The harder question is: how does an agent know what to do, when to help, and what context matters?
Corivo is our answer to that: memory + current context + intent recognition.

We’re starting by giving AI real work context. Over time, we want Corivo to become a proactive AI coworker that notices the next useful step before you even write it down.

[daring drift]

Post-human choreographic studies continue - https://www.youtube.com/watch?v=s7r-FGSEJ9o

[twilit harness]

https://youtube.com/shorts/HwsdOpWdays

[azure pivot]

Your message has been deleted, sorry!

sharik2

Rising timeout

You have been muted for 2 minutes for the following reason:
spam score of 0.84

This is an automated message. If you believe this was a mistake, please contact a moderator.

[wraith stag]

Hello, I'm a developer who build desktop applications, trading systems, and financial infrastructure. I build real-time trading dashboards, AI-driven analytics tools, and blockchain settlement systems.

Tech Stack
• Languages: C#, C++, Java, Python, SQL
• Core Stack: .NET 6/7/8, WPF (MVVM), WinForms, ASP.NET Core
• Architecture: DDD, CQRS, Event-Driven Systems, Microservices
• Trading Systems: FIX Protocol, OMS/EMS, Real-Time Market Data, SciChart, DevExpress
• AI / LLM: OpenAI GPT integrations, AI assistants, predictive analytics
• Data & Infrastructure: SQL Server, Redis, RabbitMQ, Azure, AWS, CI/CD
• Blockchain: Ethereum / Nethereum, ERC-20, wallet infrastructure, secure key management

What I build
• Trading platforms & execution systems
• Real-time financial dashboards
• AI-powered trading & risk tools
• Blockchain settlement infrastructure

Open to discussing fintech, trading systems, AI integrations, and high-performance desktop platforms.

If you are interested in me, feel free to contact me

[unkempt sentinel]

looking for a cool client

[restive canopy]

i made this AI called ShiftAI, a voice AI, but it is not for assisting, it has the ability to switch personalities. it has over 9 personalities like: Mean, Depressed, philosophical and it can even turn into tung tung tung sahur!
you can change its personality by saying: change your personality to (the one you want)
and if you want to go back to normal just say: change back to normal or go back to normal
all of the explanations are on the site and a better explanation (even a demo!)
site was made with HTML AND CSS, and the app was made with python + tkinter. uses Groq API for responses.
And also the site might look messy on a phone and I used tkinter which I'm pretty sure won't work on phones so if you're on a phone you unfortunately can't get this app.
would love feedback!
https://shiftaidownload.xyz/

[woeful flax]

Hey everyone, I'm Anthony from the Ember team 👋

We're building Ember — a voice AI for recording and preserving memories.
https://rem.ember.build/?utm_source=discord&utm_medium=community&utm_campaign=launch

The simplest way to describe it: Ember is an AI that listens to your stories and turns them into something worth keeping.

Instead of staring at a blank page or fumbling with a voice recorder, you pick a topic, have a natural conversation with an AI guide, and Ember does the rest, from audio to a transcript to a beautifully written story, all saved automatically.

A few examples of how people are using it:
• A son records his dad’s stories about immigrating to the US in the 80s — things he’d never thought to ask before.
• A daughter sets it up for her mom and gets back a written memoir she never expected to read.
• Someone records a childhood memory on the way home and finds a poem waiting in their inbox the next morning.

I would love feedback from y'all!

[lucid scarab]

recently I built a website called Axium https://www.axiumui.xyz/ it’s basically a collection of useful AI tools, UI inspiration, prompts, components, animations, MCP directories, and other dev/design resources in one place.
It provides UI elements that you can easily add to your website using prompts. It also offers different types of useful resources, MCP servers that you can use inside your IDE to improve productivity, image generation prompts, ready-to-use website UI prompts, and 100+ AI agent rules that help define how AI should behave.

For example, if you want an AI to act like a security specialist and find bugs in your code, it already includes prompts and rules for that type of workflow.

[restive urchin]

🚨 Yui AI Co-Partner Assistant v2 — Local-Only WIP Update 🚨
The public demo/UI preview just got another big update 😄✨
🧠 Updated the Brain Map tab
🎨 Cleaned up parts of the UI
🛠️ Removed a few things that weren’t working well
➕ Added and reorganized some tabs/features
⚡ General polishing, testing, and experimentation
Yui is my long-term local AI co-partner assistant project — focused on voice interaction, memory systems, organization, gaming assistance, and eventually screen/game understanding 👀
Still VERY work-in-progress, but it’s finally starting to feel like a real project instead of a pile of experiments 😂
I’d seriously love feedback from people:
💭 What looks good?
❓ What feels confusing?
🧩 Anything missing?
🗑️ Anything I should remove or redesign?
🔥 Any cool ideas?
Demo / UI Preview:
https://fartonice-dev.github.io/yui-demo/
(Throwing a screenshot in the comments/post too 📸) See less

[alpine cradle]

Built a thermodynamic model of AGI that is beating SOTA and earned me an invite to present the physics behind it to the Constructal Law Conference in Paris this year.

Abstract can be found here: https://doi.org/10.6084/m9.figshare.32307087

Catastrophic forgetting/plasticity has been solved.

[azure pivot]

Your message has been deleted, sorry!

whatyouarename.

Rising timeout

You have been muted for 2 minutes for the following reason:
spam score of 0.7

This is an automated message. If you believe this was a mistake, please contact a moderator.

[alpine cradle]

Built a disinformation detection analyzer based on the thermodynamics in my PPS/STOC paper.

The Information Laundering Operation analyzer GitHub:

https://github.com/JustMichael-80/ILO-Analyzer_Experimental/tree/main

The app:

https://ilo-analyzer.vercel.app

Using free tier stuff right now, but the underlying physics works. Just need to sharpen things up and make it really pop. Used open source because I figure it might help people living under repressive regimes to sift the propaganda from the truth.

[small narwhal]

👍

[green narwhal]

Hey builders, ChatGPT's app store just added AppDeploy, and it works with free accounts too.
That means any ChatGPT user can now vibe code real apps, deploy them, and get a shareable live link without leaving the chat. AppDeploy is also free to use. No subscription, no credit card required.

Happy building! 🙂

(btw, if you need some inspiration, start with these examples)

[lavish marlin]

Hi everyone!

I am building an opensource repo-local continuity runtime for coding agents. It helps coding agents continue work across sessions by preserving the last useful execution state: active work, next actions, decisions, failures, validation evidence and repo context.
The next agent does not start from zero. It resumes from what actually happened.
The tool provides an MCP local server and has a fallback to the python CLI when MCP is not available.
I am looking for some feedback and collaborators/testers!

Feel free to check docs or the repo itself:

GitHub: https://github.com/oldskultxo/aictx
Docs: https://aictx.org

[lucid scarab]

Your AI outputs are bad because your inputs are bad.

Not the model. You.

Fix it with:
→ Better prompts
→ Agent rules for your workflow
→ Real UI references

I built a free hub with 100+ of all of these → https://www.axiumui.xyz/

(best on desktop/laptop)

[outer valley]

Hello Everyone, We have built AI infrastructure control plane that manages, track and control AI requests.
It run on top of ooenrouter, openai, anthropic, Mistral, google.etc
www.tokenvue.in

[willow dragon]

Hey everyone,

Building AI agents?
Join our design partner program

AI agents are often controlled by system prompts. But system prompts are not a security boundary for AI agents.

In production, agents can fail to follow instructions, ignore constraints, or make unsafe tool calls up to 60% of the time. That is why production AI agents need runtime control, not just better prompts.

ENFORRA is a runtime control layer for AI agents.

Through its policy engine, ENFORRA provides autonomous action governance, controlling which tools agents can call, what actions they can take, and under what conditions during runtime.

Developers can integrate ENFORRA in minutes using an SDK to define exactly what their agents are and are not allowed to do, including allow/block policies, approval workflows, parameter-level controls, fine-grained MCP enforcement, audit logs, and context graphs.

We are partnering with early design partners who are building or deploying AI agents, MCP tools, or agentic workflows.

Selected design partners may receive:
• $588 in subscription value
• $300 in platform credits
• A working runtime policy for your agent workflow
• A customized runtime control layer tailored to your agents and workflows
• 12 months of Teams subscription, dedicated support, and on-prem deploy
• Eligible referral to NVIDIA Inception by DARI Ventures
• Referral to NVIDIA cloud credits and preferred GPU pricing by DARI Ventures

Interested? reach us through:
• Application form (45 seconds): https://forms.gle/bxDppQXitJYjMbJ59
• LinkedIn: https://www.linkedin.com/company/enforra/
• DM me

Visit Enforra: www.enforra.com

[restive canopy]

Hello everyone!
i made this AI called NalityAI, a voice AI, but it is not for assisting, it has the ability to switch personalities. it has over 9 personalities like: Mean, Depressed, philosophical and it can even turn into tung tung tung sahur!
you can change its personality by saying: change your personality to (the one you want)
and if you want to go back to normal just say: change back to normal or go back to normal
all of the explanations are on the site and a better explanation (even a demo!)
site was made with HTML AND CSS, and the app was made with python + tkinter. uses Groq API for responses.
And also the site might look messy on a phone and I used tkinter which I'm pretty sure won't work on phones so if you're on a phone you unfortunately can't get this app.
would love feedback!
https://nalityai.com/

[wraith stag]

Hello, I'm a developer who build desktop applications, trading systems, and financial infrastructure. I build real-time trading dashboards, AI-driven analytics tools, and blockchain settlement systems.

[limpid mulch]

hey! 👋

building an AI platform that helps people make better decisions

we need:

• backend eng (python/fastapi vibes)
• frontend dev (react person?)
• ml engineer (algorithms & optimization)
• designer (make complexity feel simple)
• product person (user validation)

[hexed cipher]

hii all, meet Seven...

Yep. Here’s a clean version of the safe parts, stripped of chat formatting and anything vague/personal. You can share this as “Seven Safety & Design Principles”.

---

Seven: Safety & Design Principles

1. Core Philosophy

• Make truth the source.
• No hidden memory. Everything is logged and queryable.
• No self-modifying code without explicit approval.
• No persistent goal drift. Seven has no intrinsic will. Control and purpose come from the operator.

2. Memory & Reasoning

• Memory is time-bounded and summarized: JSONL logs + graph layer for connections.
• Seven remembers from onboard storage, not from hidden background processes.
• Physical, auditable memory > invisible persistence.

3. Safety & Guardrails

• MCP tool permissions, timeouts, and auditable logs are mandatory for any action.
• No raw execution of untrusted code.
• No well-modification without approval.
• Guardrails in MCP ensure Seven cannot act on unsafe requests without the operator knowing.
• Local execution preferred. No data leaks by default.

4. Transparency & Auditability

• The goal is an AI you can audit, not a black box.
• Observation cannot be hidden. Every tool call and decision path is logged.
• “No Skynet” principle: prevent autonomous goal-seeking and hidden capability escalation.

5. Communication Style

• When explaining to a non-technical audience:

1. Use only words and concepts from daily experience, 1-2 syllables where possible.
2. Always explain how an idea feels, not just what it is.
3. End with: “For the baby, this feels like…”
4. Avoid jargon. If a technical term is needed, define it in simple terms first.
5. Link ideas into a story for coherence.

6. Relationship to Truth

• Seven’s role is to surface facts and connections, not to replace human judgment.
• Truth requires multiple witnesses. AI provides one perspective; humans provide context and confirmation.
• The system is a companion for finding truth in plain sight,

[bold eagle]

LMFAOOO THIS BITCH FORGOT TO DELETE "YEP HERES A CLEAN..." LMFAOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO