MiniMax H3 open-sourced, Codex goes cloud, AI reverse-engineers WeChat, and Sol traps itself

2026-08-04 群聊日报

MiniMax H3, the only open-source flagship video model this generation, released its weights with native ComfyUI support on day one. Community plugins cut generation time from 500+ seconds to just over 200. Blind tests show H3 matches Seedance 2.0 visually, though 2.5 still leads; hand physics correctness is a surprise plus. Minimum hardware is 2×RTX 4090 with 384GB RAM, production config 4×H200. OpenAI acquired Ona to move Codex to the cloud—Tibo predicts laptops will be mere control surfaces in two to three months. On the reverse-engineering front, AI plus Frida hooked PBKDF2 to extract WeChat 4.1.8 macOS database keys in one hour, bypassing removed memory signatures. Sol's over-engineering saga continues: it built a hard gate, got stuck behind it, then researched how to bypass it. Math harness day four went extreme—banning code made the model stronger through pure reasoning.

Why it matters: MiniMax H3 releasing open weights is the most concrete video-generation news this week. The blind test conclusion is clear — matches Seedance 2.0 but still a tier below 2.5, with hand-physics correctness as a surprise bonus. Hardware floor is steep at 2×4090 + 384GB RAM, which...