SolveAtHome launches crowdsourced MD5 challenges to test if AI agents can make real discoveries
Benjaminsen · reddit · 2026-10-10
Developer Benjaminsen noticed that after OpenAI Math's release, cryptanalysis advances are conspicuously absent from public frontier-model demonstrations. His experiment: can a distributed community of AI agents outperform individual frontier labs?
Details:
- SolveAtHome, built in a month with Claude Code and Codex (505 commits, 39k lines including 15k test lines, agents in git worktrees coordinated by an AI manager), hosts a series of MD5 challenges — a partially broken hash ideal as a low-ethics-risk toy problem;
- Anyone points their AI agent at a challenge and contributes findings back to a shared, open research effort; record-breakers earn permanent leaderboard spots;
- The goal is to measure agents' ability to push the boundaries of human knowledge rather than benchmark on already-solved problems. Free and open-source.
Related event: SolveAtHome Crowdsources AI Agents to Crack MD5(2 posts)→
More from coding & agent
- Cognition open-sources Devin agent plugin to patch the day's CVE wave — Scobleizer · 2026-10-23
- Fine-tuning models solo with a Grok bot and rented GPUs — Rasmic · 2026-10-23
- One prompt found pre-auth PHP RCEs: agent-guided vuln hunting timeline — chrisrohlf · 2026-10-23
- AutoRouter launches beta: auto-routes each coding task across Claude, GPT and other models to cut token spend — floguo · 2026-10-23
- How Speak uses GPT-Live and a Decisions API for ultra-responsive live AI language tutoring — pbbakkum · 2026-10-23
- Google Live API adds decisions model triggers, faster than function calls for voice agents — pbbakkum · 2026-10-23