Research in progress, experiments that worked, and a few things that probably shouldn’t be footnoted.
Eighty-eight completely fabricated Hopper-adjacent sayings by Kirsten Ruge, drafted with Codex, Claude Fable 5.1, and GLM-5.2—historically disclaimed and spiritually plausible.
18 min readGoogle shipped Gemma 4 assistant drafters, SGLang merged Frozen-KV MTP support, and a single B300 turned it into a real 1.6x decode-speed win.
6 min readNVIDIA dropped a 31B activated-multimodal model with video, audio, image, OCR, GUI, tool-calling, and long-context support. That sounds suspiciously close to a pair-programming onboarding primitive.
4 min readThis week’s frontier-model-in-a-small-Blackwell-shaped-box experiment ended with a useful answer: yes, Kimi K2.6 can fit, but only if you stop acting like the box is an H200.
3 min read