K3 Autonomously Builds Web DOS Emulator
teortaxesTex · x · 2026-07-17
The author shared an extreme test result: a model/agent codenamed K3 ran continuously for several hours on a target task, hitting the roughly 1 million context limit twice, and managed to keep progressing after automatic compression.
It ultimately built a DOS emulator on the web, capable of directly running the ROM for Legend of Sword and Fairy (仙剑奇侠传). The process involved no third-party libraries; the model hand-wrote the implementation and adapted the disassembly itself. It only paused once for human assistance to test keyboard input and cross-checked against DOSBox, requiring almost zero human intervention.
More from coding & agent
- Dev builds talk on guardrails workflow for shipping AI-written code without reading it — TejasKumar_ · 2026-09-11
- banteg: Codex auto-review has regressed, blocking steps needed to complete authorized tasks — banteg · 2026-09-11
- lucasmeijer's workflow: handwrite the doc yourself, then have the agent challenge your understanding — lucasmeijer · 2026-09-11
- A doc-anchored agent workflow: you write, the agent only critiques and finds disagreements — lucasmeijer · 2026-09-11
- SymKit MCP: 44 tools for AI agents to verify symbolic derivations — Foreign-Specific-604 · 2026-09-11
- GitHub Copilot team routes user bug reports to an AI agent via Slack — marlene_zw · 2026-09-11