Claude Code has finally started building his own AI team, and this isn't just adding new features.

Until now, Claude Code is No matter how smart we are, we are basically humans. I thought about the job and gave instructions to the AI. But this time,

くまらぼAug 1510

GLM 5.3 just dropped - and it's sitting at the top of the benchmarks.

But benchmarks are a lab result, not a reality check. So we ran it against every frontier model, head to head: 1. GPT-5.6 Sol vs GLM 5.3 2. Opus 5 vs GLM 5.3 3. Gemini 3.7 Flash vs GLM 5.3 4. Grok 4.6 vs GLM

Build Fast with AIAug 159

Every model we've tested answers the surface question.

Claude Opus 5 was the first to tell us the question was self-contradictory - then hand us the working envelope instead of a compromise. Two requirements. Only one was survivable. New clip from the review series 👇

SmokeTestAug 15

Town number 3 - a Wild West Frontier town.

This one was a lot more work (though also a lot farther along) than the previous 2, definitely not a one shot. A lot of lessons learned though, so I feel more optimistic the pipeline will do better with Town number 4. The goal with each

Chris RuppAug 15