Ebiose – A Darwin‑Style Playground for Self‑Evolving AI Agents
Hi HN! After a year of R&D at Inria (the French national lab), we have just open-sourced Ebiose. Ebiose is a distributed, Darwin-style playground where AI architect agents design, test, and improve other agents. Instead of AI built behind closed doors, anyone can spin up a forge, state a problem, and watch candidate agents compete until the fittest survive. An example instruction given to an Ebiose forge: "Build a LangGraph agent that processes SaaS customer refunds directly through our ERP, escalating to a human for edge cases. Use the following tools: ERP API, email/Twilio…
In plain words
Ebiose is an open-source platform where AI agents design, test, and compete to solve problems through evolutionary selection. Users can specify a task, and architect agents automatically generate candidate solutions that are evaluated and refined through iteration. The strongest agents survive and recombine to improve over time. It's designed for developers and organizations wanting to automate agent creation rather than building AI systems manually.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
Hi HN! After a year of R&D at Inria (the French national lab), we have just open-sourced Ebiose. Ebiose is a distributed, Darwin-style playground where AI architect agents design, test, and improve other agents. Instead of AI built behind closed doors, anyone can spin up a forge, state a problem, and watch candidate agents compete until the fittest survive. An example instruction given to an Ebiose forge: "Build a LangGraph agent that processes SaaS customer refunds directly through our ERP, escalating to a human for edge cases. Use the following tools: ERP API, email/Twilio integration, refund policy documentation, and historical support tickets." TL;DR - Meta-architect agents: Generate new agents, reusable components, or even models in the future. - Evolutionary loop: Only the best agents survive and recombine. - Self-improvement: Architect agents evolve themselves over time. - An evolving library: Reusable components, i.e. agent’s genes, compete for survival based on power, efficiency, and success. - P2P compute (planned): Goal of utilizing unused device computing power. Important note: Ebiose is still in a very early stage. Architect agents are basic for now, and there are no reusable components yet. But the core loop is working, and we’re now inviting more contributors to help build it up. What's in this first Ebiose release? - Hand-crafted architect agent that performs prompt engineering and assembles agent graphs. - Evolutionary engine. - Forges: isolated labs for building agents tailored to specific tasks. - Persistent ecosystems where top agents live on and tackle new forges. - Free credits to run your own forges (thanks to cloud sponsors). - Graph-based agents made of LLM nodes. - Initial LangGraph runtime (others are very welcome). Roadmap (help wanted!): - Agent nodes: power agents with new types of nodes such as code execution, classic ML models, tool use, fine-tuning, etc. - Foundational forges: code generation, prompt tuning, context grounding, agent-as-a-judge, etc. - Meta-forges: architect-agents that invent other architect-agents (yes, agents that build agents that build agents). - P2P compute layer to use idle GPUs/CPUs (llama.cpp). GitHub: https://github.com/ebiose-ai/ebiose - MIT licensed. Fork it, test it, star it, break it… We are eager to see what you build and to hear your feedback! P.S. If you saw AlphaEvolve, they used LLM-guided evolution to find ultra-efficient algorithms like a 48-operation matrix-multiply. Ebiose shares the same evolutionary idea, a level up: instead of optimizing one program, it evolves entire agents, and even architect agents that design other agents.
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
Launched alongside, July 2025
the whole month →
Trickle - Magic Canvas ▲1,064The 1st Agentic Canvas for building apps visually with AI
AI · 2025 · trickle.so
- TB
Life & fun · 2025 · trytender.app
- TY
Today marks ten years, 3653 consecutive days, of running at least one mile every day under the USRSA rules [1]. To celebrate, I built an interactive dashboard that turns a decade of GPX files into charts you can explore. Running has truly changed my life: I've made lifelong friends, explored beautiful places, and more importantly invested into my own health and fitness, which I'm starting to see the positive benefits as I get older. The stack is pretty simple: a NextJS app, with a Postgres database to keep all my running data, and all the stats are pre-computed and cached in Redis, so I…
Dev tools · 2025 · nodaysoff.run
- DA
Made this website as an exercise in vibe-coding and GCP. It was posted about a few times around the internet, on sites like Morning Brew (https://www.morningbrew.com/issues/business-buzzkill), MetaFilter (https://www.metafilter.com/209703/Draw-A-Fish), boingboing.net, etc. I think it's cute! I built a basic CNN trained against penises and swastikas, and then anything that doesn't hit the 63% confidence score gets sent to a mod queue, a vibe-coded fish-tinder…
Dev tools · 2025 · drawafish.com
