The Recursive Agents Hackathon
Oct 17, 2026 · San Francisco, CA
Etch turns one sentence into a real circuit board. You say “build a GPS tracker like an AirTag” and walk away. A team of agents picks real parts, writes the…
SPARKLETEAM is a self-improving agent team that teaches itself to load a dishwasher. It learned one law of physics and one superstition, and we measured which…
AGI Factory is a robot maintenance team that rewrites itself and keeps a change only when unseen factory months prove it better and safer. It runs a real…
Lets go : An agent that lives in your chat. Ask it to plan, and it turns everyone’s messages into rules. Our second agent, Mato, learns from LetsGo's…

Trendify is an AI-powered trend discovery platform that finds emerging trends before they go viral. It deploys thousands of autonomous agents across social…

Antaeus is an autonomous research lab for recursive agent improvement. Given a capability goal, Antaeus generates executable synthetic environments that…

Supplier Chief of Staff is a recursive agent team that prevents a specialty-pharma manufacturer from running out of materials and asks a planner on Telegram…
EMR Copilot: an agent team that clears a clinic's overnight back office (claim denials, prior auths, referrals, chart prep) and writes and verifies its own…

AI agents often say “done” as soon as the tests pass, but the tests can have blind spots. Traceback is my answer to that for the Done Means Done track. You…
DoneDone is a framework that gives agent teams everything they need to loop. Start with a one-line goal, approve a suggested list of verifiable tasks…

https://factory-n1.vercel.app/ Every generation has to prove it's better than the last. Factory N+1 is an AI team that redesigns a simulated package…

Immune is a recursive testing and repair system for AI shopping agents. It tackles a subtle problem: a fix that stops an attack can also break legitimate…

AstroMesh is a multi-agent simulation and autonomous constellation management platform designed to handle Low Earth Orbit (LEO) congestion using Crusoe…

No Free Wins is an AI agent team that makes Python code faster and only counts a speedup it can prove. You give it a repo or GitHub URL and a goal such as…
Recursive search for GPU kernels with an LLM: compile-checked without a GPU, timed on one when there is one.
A recursive self-improving agent that writes optimized SQL queries. Sometimes, reporting dashboards let customer enter their own SQL queries. Their SQL…

Unfoolable is a self-improving AI agent team that can't fool itself. AI agents often say "done" when the work is wrong — our starting team did it 8 out of 9…

Repro turns a vague website complaint into independently verified reproduction steps, browser evidence, runnable tests, and an optional tested patch. A…
CORE GENESIS - The AI-Native Company That Builds, Tests, and Evolves Itself What if a company could autonomously research, build products, test its own work…

# Recursive Travel AI — Self-Improving Multi-Agent Travel Planner **Project overview** Recursive Travel AI is a multi-agent travel planning application that…

AI agents fill vague goals with their own assumptions, then confidently report "done". Proof of Done is an agent team on Crusoe Managed Inference that is not…
A chip is a lot of tiny switches wired together. One wire on the wrong pin and it gives wrong answers, and you can’t see it. AutoFab is a chip factory where…

Crusoe Cooler System is an autonomous physical-AI digital twin for liquid-cooled AI data centers. Crusoe Managed Inference serves as the cognitive engine for…
An evidence-driven reliability framework for autonomous agents powered by Nemotron on Crusoe Managed Inference. We capture real native agent tool calls and…
Calendar app that redesigns itself from learned usage patterns

Shadowbox is a system where two AI teams get better at chip design by competing with each other, while the underlying model stays exactly the same. In the…

Concord is an agent team that builds and enforces a glossary for long-form translation, the "shared brain" for a Chinese novel translated into English. Many…

Robot Agent Lab: a team of Nemotron agents runs a science lab's night shift on a simulated SO-101 arm. Tubes arrive with coloured caps. Red goes to the cold…
Agents fill vague goals with silent assumptions and then confidently report success. done turns a one-line goal into a self-splitting tree of agents that…
AYA turns a feature request and a working software reference into a tested component for your own product. You provide a target GitHub repository, the…

TicketZero is a recursive AI support engineer that doesn't just propose fixes, but verifies them. Traditional AI support agents repeatedly process entire…

The problem: Agents act on old facts. In a multi-agent team, one agent often keeps acting on a fact that has since changed. The result is wasted work or the…
Miles: a self-improving music agent for anything, one MCP away. Miles is Claude Code for music. You give it a brief like “tense trailer cue, 90 BPM, C minor”…

piiguard: a PII leak guard for sandboxed agents that trains itself, in your industry's language. The problem. NVIDIA NemoClaw runs agents inside an…

FocalPoint is an AI chat app that watches how you read and adjusts its answers to match.

A team of agents that invent novel candidate drug molecules for curing a user-specified disease. You can just give the team a disease name and walk away. It…

a vision-based robotic arm project, a student supervised by NVIDIA Nemotron 3.5 Super VL - a state-of-the-art multimodal reasoning, tool-use proficiency, and…

A team of AI agents that has to prove every change it makes to itself. After each run it proposes one change to its own rules or setup; a plain-code gate…

For AI to help with your physical home, it needs to understand the space, the things in it, and whether its proposed changes actually work. I built HomeOS: a…
Casefile is a governed recursive-agent system for verifying complex public-record claims across fragmented government and investigative sources. Public…
Darwin Inc. — Human org charts are generation zero. The problem: Every company is about to deploy teams of AI agents, but they organize them by copying human…
Bug Hunters is a QA crew that lives inside the Godot editor. You press one button and AI testers play your game, find bugs, prove each bug by replaying it…

Escape Labs is an autonomous cancer-research workflow built using published FGFR laboratory data. Its ambition is to help researchers anticipate drug…
Sway is endless instrumental music, generated by Google's Lyria, that learns your taste from two buttons: thumbs up and thumbs down. The music never stops…
Crusoe Forge is an agent that speeds up FPGA designs. You give it a slow Verilog design and a target clock speed for a Lattice ECP5 chip. It rewrites the…
Gemba is a decision twin for operations. Slotting software and digital twins tell a warehouse where things should go and how the floor behaves; they treat the…
BrainBob is a shared brain for agent teams: keep knowledge current, control who can use it, and trace every change. When one agent remembers an outdated…

DeepCore-N1: An autonomous, self-evolving RGB-D vision and force-guided excavation engine built on Crusoe Managed Inference using NVIDIA Nemotron. Given a…
An AI data center is a stack of optimization problems: where the racks go, how equipment is placed, the GPU kernels, and the inference bill. Code can check…

Silk Lab is a self-evolving research lab: a platform where a team of AI agents builds, measures and improves itself on a real research goal. You give it one…

Annotating videos for temporal action segmentation is expensive and time-consuming, especially for Physical AI, where identifying precise transitions between…

Clueso is a self-improving AI agent team that audits websites for privacy disclosure gaps, tracking issues, and altered brand badges. It verifies findings…
Silico doesn’t just generate an answer. It checks whether the work meets the goal and shows what remains unresolved. Silico lets you ask a question and…

Pitstop: a supervisor agent that learns to catch drift in agent swarms Long-running agent teams fail quietly. A sub-agent drifts off task, hallucinates a…

CampusLoop turns student move-out waste into affordable move-in kits. Students moving out upload a photo of an item. A vision agent grades its condition…

The cost of hardware iteration is high; constant back-and-forth between electrical engineers, manufacturers, and project managers. Especially in industries…

Robotic(humanoid) meal delivery in ahouse/restaurant : carrying food and drinks safely to a destination.

A loss function for agent traces based on Ralph Wiggum and Russian Nesting Dolls principles
Done Means Done is an AI fact-checking team that checks its own work before calling it finished. Give it an article link, and it picks up to six factual…
Twin is an AI teammate that joins your meetings, not a note-taker. It joins the call with its own tile and voice. It sketches the architecture on a shared…

I’m a high schooler and a supersonic engineer at Boom. Designing aircraft normally takes a big team and millions in hardware money. FORGE puts that ability in…
recipe-forge: recursive Nemotron agents that make NVIDIA's Nemotron serving recipes faster, losslessly, with every gain verified. It's basically Nemotron…
The hastily sent colony ships escaping the calamity arrive to discover the hoped for paradise is a harsh wasteland barely suitable for life. Until the supply…

AI-assisted development is changing how quickly software gets built. Developers can now generate and ship code faster than ever, resulting in a growing volume…
Wall-Ne is a self-healing, self improving, and fully autonomous agent flock that irrigates, fertilizes, and cares for crops. Each instance of a wall-ne robot…

Liftoff makes launch videos for your brand, and checks them before you see them. Paste your website. Liftoff learns your brand (name, colours, tone), looks…
Brain-computer interfaces lose accuracy every day as neural signals drift, so today’s systems make users recalibrate. Neuralink users spend about 55 minutes a…

We built Jarvis. Like literally Jarvis from Iron Man. You can use your voice to tell your computer to do things like in the movie and it'll go out and do…
Product knowledge goes stale the moment it’s written down. Product Brain is a shared memory that a team of agents keeps current, lean, and contradiction-free…

I created a memory system where both the judge/critic learns to evaluate the demands of the agent task, and the knowledge is CRUD and its saliency is…
Carly: an AI agent team that buys and sells used cars for a dealership, and gets better with every deal. Used-car dealers spend all day scrolling Facebook…

Understudy is an autonomous on-call team for an AI inference factory fleet. When a node in the fleet breaks (wrong model ID, a port squatter, a crashed…

Lineage is a team of autonomous AI agents that solves difficult problems with zero human supervision. Every agent runs on NVIDIA Nemotron 3 Nano, powered by…

This helps attorney to review the contracts and find out the errors, users need to correct first, helping them do some corrections before the contracts handed…
One-line pitch Your coding agent says "task complete" and exits 0; done-receipt re-runs the work out of its reach and hands you a signed receipt that says…
Air traffic controllers juggle aircraft, radio calls, and changing conditions under intense pressure. Up in the Air explores a recursive agent team that…

Rare Disease Detective is a human-in-the-loop team of AI agents that cross-checks hard medical cases and hands them back to the doctor with the evidence…
vidloop is a team of AI agents that makes short films and fixes its own work. Nemotron storyboards and judges, Qwen-Image and HappyHorse generate, and Codex…

Inspired by Elon Musk's Flying Car Demonstration. The goal was to create an agent swarm which could continuously work to create a production grade…
AdCampaigner is a self-improving ad-campaign agent team that has to prove every change before keeping it. Most agent teams perform the same on run 100 as on…

Problem: Intake forms are static, but users care about different things. Fixed questions waste their time and miss what actually matters. A great example of…

Green-Stove Agents is a team of agents that runs a community solar cooking mini-grid you can walk away from. Off-grid households want to cook on electricity…

Arc is a search engine for agents to look up people, with a better interface than Apollo. Type a full name and press Find. An AI agent searches the web…

CASCADE is a recursive agent team you can walk away from during a compound disaster — rain → landslides → river dams → dam bursts → floods, or quake → bridge…

Zen — your product’s recursive design team. Give Zen a UX goal. Specialist agents inspect the code, challenge each other’s findings, and turn their decisions…

Capy is a team of AI capybaras that babysits multi-node GPU training jobs on Crusoe while you sleep. When a job slows down, Capy finds the cause, proves it…

A self-improving agent harness for hardware-specific tasks, where real devices decide what improves. A team of Nemotron agents ports layers from new Hugging…
LLM agents learn can learn skills from action sequences that worked before. But a skill doesn't know why it worked, so when the world changes, the agent…
RSI for classical ML models.
We don’t design agent teams. We let them evolve. Starting from one agent, an 8B team grew to seven and beat a model 9× its size on unseen competition math.

Simulation Lab is an AI-powered prototyping platform that turns plain-language ideas into interactive simulations. Powered by Crusoe, its agents generate…

SIEVE screens porous materials for CO₂ capture with grand-canonical Monte Carlo (GCMC) under a fixed CPU budget. A team of LLM agents decides what to simulate…
Cursoe-tron Stunt School is a team of AI agents on Crusoe Managed Inference that teaches itself to fly a jet-powered car through stunts in a real 3D game, and…
Mission Intelligence by AIZOYA turns public-source research into a recursive investigation. Grace plans and delegates, Errands gathers sources through…
GridTwin is a recursive agent system for power plant maintenance: a repair is only done when the evidence proves it. Agents diagnose faults using telemetry…

This project is a self-improving insurance claims-triage agent team. It addresses a common problem: agent systems repeatedly make the same mistakes because…
Persistent agent organizations with reflection over memory, prompts, skills, tools, and topology, using LangChain's Deep Agents framework and Crusoe inference…

DealMind: negotiators that learn from every deal. A team of AI negotiator agents buys furniture from suppliers for a shop chain, one specialist per supplier…
Reform is a self-building, self-healing agent platform that preserves user intent. Users describe a task, review its goal, success checks and permissions…
FruitTree.AI is the observability and recursive learning layer for hackathons. Builders connect any AI coding tool (Claude Code, Cursor, Codex, MCP) and opt…

Blackout lets AI agents act on their own without doing anything twice. Agents are starting to send emails, issue refunds and change systems unattended; when…
Turing: an agentic self-improving harness from first principles that provides maximum observability and customization. So think of Claude Code or Codex but…

Groundhog is a team of AI agents that uses the real Azure portal the way a person does: it looks at the screen and clicks. It runs on NVIDIA Nemotron models…
Wimbl: A Recursive Family of Agents for Football Travel Wimbl is an event-led travel discovery system built around a family of specialized AI agents. It…
AutoHeal : Recursive Agents for Self-Healing ML Pipelines

AgentMemory is a shared brain for AI teams. Three Crusoe agents (research, design, development) run the same project twice: isolated vs. with persistent JSON…
Agent Teams is a Slack-style workspace where you run a company of AI agents. You tell Mira, the CEO agent, what you need. She picks the right team and starts…

AutoRoboHarness is a team of robot agents that improves itself. Robots work tasks in a simulated retail store by reading a shared, hierarchical brain, a…

Swarm takes a big software spec, recursively splits it into a tree of small tasks and runs hundreds of real agents on them at once. Each agent sees only the…
