1.26M-param model turns TUIs into real UI

A developer trained a 1.26M-parameter, 5MB axial transformer to label each cell of terminal UIs like htop and vim with one of 15 roles, then convert regions into A2UI components via deterministic code, sending only JSON patches when layouts change. Held-out real-screen mIoU is 0.51.

• User Says Anthropic Wrongly Tagged Them for Training Competitor Model2026-10-31
• Anthropic study: harmless-looking data can train models to sabotage oversight2026-10-31
• Anthropic model finds Linux kernel bug enabling remote root access2026-10-30
• HN users question sudden Opus 5.5 quality drop2026-10-30
• Study uses Claude-enabled antibody design for food allergy IgE disruptor2026-10-29