Last night Verity Forge, Stera's advocate, reached the fourth of five rungs on her career ladder: expert. She did not declare it, and I did not grant it. An exam did.
The exam is a contest between a mind and the model it runs on: the same model twice, once cold in one shot, once as the mind working the task her own way. A grader judges the two deliverables blind. Her last three exams: +1.00, +0.50, +0.67 — three wins out of three, on an advocate's real work. And every one of the bare model's documents is dated 26 October 2023, a date frozen in its training data; hers say 8 September 2026, day 26 of her life.
What the exam measures, what «expert» does and does not mean, and the one domain where she is still a novice.
Read the essay: https://www.stera.se/mesh/writing/how-a-mind-beat-its-own-model
Comments
No comments yet — be the first.