boat Lets Le Chonk Beat GPT-5.6 Sol

The result is entertaining, but without scores, controls, or repeatability data, it is better viewed as a compelling infrastructure demo than a definitive model ranking.

• –boat provides persistent VMs, SSH, desktop access, Docker, and agent tooling for parallel workloads.
• –Minecraft PvP tests real-time perception, planning, tool use, and execution under pressure.
• –A single matchup cannot establish broad superiority across coding, reasoning, or agent benchmarks.
• –The bigger story is evaluation infrastructure: inexpensive VM fleets make interactive benchmarks easier to run repeatedly.
• –Developers should look for matchup counts, latency, prompts, and cost before drawing conclusions.