Jido – Run 10k agents at 25KB each (Elixir)
Hi HN! I'm Mike Hostetler and I built Jido, an Agent SDK in Elixir that lets you run thousands of agents without heavy infrastructure. Repo: https://github.com/agentjido/jido Getting Started: https://hexdocs.pm/jido/getting-started.html Why another framework? After using several popular Agent frameworks and platforms, I had two key challenges: - Running multiple agents required process-heavy infrastructure like Docker or K8s. Running 50,000 agents in parallel was costly and diminished the benefits of agentic programming. - Today's agents require too…
In plain words
Jido is an Agent SDK built in Elixir that enables developers to run thousands of autonomous agents with minimal infrastructure overhead, using only about 25KB per agent. It addresses the challenge of scaling multi-agent systems by eliminating the need for heavy containerization like Docker or Kubernetes, making large-scale parallel agent deployment more cost-effective. The framework is designed to support agents managing their own workflows with less human intervention than traditional agent frameworks.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
Hi HN! I'm Mike Hostetler and I built Jido, an Agent SDK in Elixir that lets you run thousands of agents without heavy infrastructure. Repo: https://github.com/agentjido/jido Getting Started: https://hexdocs.pm/jido/getting-started.html Why another framework? After using several popular Agent frameworks and platforms, I had two key challenges: - Running multiple agents required process-heavy infrastructure like Docker or K8s. Running 50,000 agents in parallel was costly and diminished the benefits of agentic programming. - Today's agents require too much human intervention when building workflows. Why couldn't agents manage their own WDLC (Workflow Design Life Cycle)? This felt like a major missing piece. Agentic frameworks were written for humans. LLMs working with this code were constantly working around human work-style assumptions. So, I wrote a framework specifically for LLMs to code and operate their own agentic flows. Elixir was a natural choice because of it's functional nature, rock-solid concurrency primitives and "let-it-crash" philosophy with dynamic error compensation. Hot code reloading was a bonus. Agents in Jido use 25Kb of memory at rest and can easily serialize then hibernate for long-lived access. Agents possess the APIs to dynamically start and manage their own sub-agents or any other Elixir process utilizing Elixir's OTP architecture. Jido Actions are functional primitives that Agents can dynamically orchestrate into workflows. Generated code can either run in a separate process in the current VM or in another BEAM VM that's linked and hardened before introduction into the Agent VM. I'm excited to help enable a world where thousands of agents work seamlessly on behalf of their human operators. Thanks!
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 · 26d ago · cactuscompute.com
