š¤ Latest Submissions
Saboteur: Chaos Monkey for AI agents
AI agents look great in demos and fall apart in production. APIs return 500s, rate limits kick in, tool calls come back malformed and context gets lost. Most agents don't handle this well: they loop on retries, invent results, or just give up. There was no standard way to test for this before shipping, so I built one. Saboteur runs your agent while deliberately breaking things around it. You pick a chaos profile (a fixed set of faults with a seeded RNG, so runs are repeatable) and Saboteur launches a cohort of N agents on the same task at once. Every fault, tool call and recovery gets written to an event log and streamed live to a dashboard grid: green is healthy, amber is recovering, red is crashed, cyan passed. At the end you get a Resilience Scorecard with behavioral metrics (recovery time, waste factor, crash rate) and when a ground-truth check exists, verdict metrics like survival rate and deception detection. It works on agents you don't own too. Point your agent's OpenAI-compatible client at Saboteur's proxy and it injects faults into the wire traffic with zero code changes. There's also a stdio shim for MCP tool servers and a GitHub Action so agent resilience can be a required check on every PR. The demo cohort runs (N=50, hell_mode profile) were captured on an AMD MI300X running vLLM, comparing Llama 3.1 8B/70B and Gemma 4 model families. Local dev runs entirely offline against llama.cpp or a deterministic mock, no GPU or API key needed.
11 Jul 2026
Quartet
Quartet is a live application that makes four AI models work together to build software. Built for the Multi-Agent Software Development track. The main goal is to show that a team of small open source models can match the skill of one massive model while costing much less. The system uses four specific roles: Spec, Coder, Tester, and Repairer. These agents talk to each other in a shared chat room using the Band protocol. The Spec agent plans the work. The Coder writes the logic. The Tester writes code to find bugs. The Repairer runs the files in a safe sandbox and sends error reports back to the Coder to fix. Beyond saving money, Quartet also lets you assign highly specialized models to specific jobs. For example, you can use a fast model to plan the application and a heavy reasoning model like the open source 120b model to write the code. This means you get the best features of every model working together on one team. Users can watch this happen live on our dashboard. You can watch the work move between nodes and compare the final code against a single large model.
19 Jun 2026
