GPT-6 trains a tiny 12k-param CNN to play VizDoom at 35 fps realtime
paraschopra · x · 2026-09-28
Paras Chopra shows a clever approach: instead of directly controlling the game in realtime (too slow), GPT-6 Astra is asked to write code training a small CNN with just 12k parameters to play VizDoom, observe failures, and iterate until it wins the level — achieving 35 fps realtime control. The key idea is using the LLM as a coach/trainer while a tiny model handles low-latency inference.
More from coding & agent
- Open-source MCP firewall blocks prompt injection in 2.7ms with deterministic regex and AST parsing — EstablishmentTough18 · 2026-09-28
- Self-Hosted K3 Cluster Lets Autonomous Infra Research Burn Tokens Without Worry — Xianbao_QIAN · 2026-09-28
- GPT-6 Astra builds 3D scenes from video, augmented with SAM3, Pi3X and Blender sim — yshan2u · 2026-09-28
- OpenAI launches first Codex Physical Builds cohort pairing agents with Raspberry Pi — OpenAIDevs · 2026-09-28
- AgentWorld Benchmark: Best LLM Only Hits 52% on Long-Horizon Multi-Agent Collaboration — Raphael Shu · 2026-09-28
- Same Prompt, 4x Difference: Claude Builds Three.js Output in 2 Hours vs 8 for Unreal — nptacek · 2026-09-28