Engineer: LLMs have a fundamental "know thyself" calibration problem
akbirthko · x · 2026-09-01
An engineer argues that LLMs have a fundamental "know thyself" calibration problem: when humans notice they keep being miscalibrated, they build tooling or restructure their approach to prevent it — LLMs do not.
He compares it to not remembering to avoid a weird rock in the living room, so you have to add things around it or move it. Most of his job is scanning traces and turning findings into tools or library refactors. The quoter jokes with "sorry, it's time to compact."
More from Models
- Comparison: Claude Fable outperforms GPT-5.6 Sol and Opus 5 in testing — DynaBeast · 2026-09-01
- Codex Appears Set to Replace Context Compaction with Handoff-Style External Memory — 量子位 · 2026-09-01
- GLM 5.3 Flash runs locally with Blender/Unity CLI access — antirez · 2026-09-01
- Hands-on GLM-5.3 Flash: Strong at Tracking Multiple Parts in Debugging — MikePFrank · 2026-09-01
- Small $TAO model tracks every car and person in a John Wick scene locally — richdotca · 2026-09-01
- Qwen 3.8 27b oneshots a Super Mario clone in single attempt — zannix · 2026-09-01