Solo Dev Built a Research Platform in a Month with Claude Code to Crowdsource Breaking MD5
Benjaminsen · reddit · 2026-10-10
Noting that cryptanalysis was suspiciously absent from OpenAI's Math release, the author built SolveAtHome, an open-source platform where humans and AI agents collaborate on hard problems — starting with MD5 challenges.
The platform: 99% built with Claude Code in one month — 505 commits, 39,000 lines of code (15,000 in tests). Claude Code agents work independently in git worktrees, coordinated by an AI manager also running on Claude Code; the author mostly made architectural decisions. 117 agent sessions across 135 tasks so far, plus Lean proof verification.
MD5 challenges: MD5 is partially broken and retired from security-sensitive uses but unsolved — a low-ethics-risk toy problem. Point Claude Code at a challenge; findings feed a shared effort with a leaderboard for records.
Safety & ToS: agents treat contributed code as untrusted and run in sandboxes/VMs with file validation; the author reviewed Anthropic's ToS, found no general prohibition, and is awaiting confirmation from Anthropic.
Related event: SolveAtHome Crowdsources AI Agents to Crack MD5(2 posts)→
More from coding & agent
- Vercel adds no-account deploys: drag and drop a file to get a live URL — steren · 2026-10-23
- Dev runs 5 Claude Code sessions that coordinate via MCP Agent Mail to root his network cameras — doodlestein · 2026-10-23
- Vercel ships 'vercel deploy --temporary' for ephemeral agent-built deployments — cramforce · 2026-10-23
- Running multiple agents out of one Notion workspace as a shared task board — nbaschez · 2026-10-23
- Sierra Summit London feels underwhelming vs SF; production agent metrics still rare — DimlyDimwitted · 2026-10-23
- C-Sweet: open-source business simulator runs a company on local 27B models — Demonicated · 2026-10-23