12 entries
Lab Notes
Each entry is dated the day it happened and reports a real case or a real run: my own reads, my own numbers. Short by design, so the record keeps pace with the work.
What the best general model read on a specialist's ruler
The strongest general model read three findings in ten on my cases and invented three in four. One accuracy number hides that. Three questions to ask.
"Normal" is not a verdict
A radiograph can say not detectable on this image, never healthy. Three lines that make a negative read honest, for humans and for models.
Agreement with past cases is not a check
A model turned my rulebook into a decision table that agreed with 99 of 100 past lines. A reviewer briefed to disagree found 19 errors, then 20 more.
What does not depreciate when the model changes
Weights get swapped by a config line. Graded truth, written rules and your own corrections compound. What does not depreciate when the model changes.
Saying less is not seeing more
The newest general model invented half as much on 128 of my own radiographs and missed more bone loss and caries. A single safety number cannot tell a safer reader from a quieter one. The two-direction test, in three steps.
It counts the teeth and misses the disease
Blind on 128 of my own radiographs, the strongest general model I could reach counted every implant exactly and never once called the jaw joint. The parts list is solved, the call is not. The check, in three steps.
Your own labels drift as you learn
My facts per read nearly doubled over a reading campaign. A sparse answer key scores true findings as inventions. Measure your drift before you grade anyone.
Read a paper for the experiments it lets you skip
One paper read twice struck a planned experiment and handed me a step. How to read a paper for the experiments it lets you skip.
Know the resolution of the instrument you trust
A vision model attends in tiles of about three millimetres on a panoramic radiograph. Most of what a radiologist calls is smaller. The one-minute arithmetic.
Three training runs, one lesson
Three fine-tunings on my own radiology reads learned my report and not the image, and no training dial showed it. The three checks that catch it.
Who owns the pixels
598 patient radiographs changed licence four times with nobody careless. Re-derive provenance from the files, not the licence field.
Write the missing page
The first Unsloth fine-tune of a vision mixture-of-experts model: the thirty-second gradient check and the two silent defaults that would have ruined the run.
The newsletter
Get the essays as they’re written.
Imaging AI, evaluation, and building alone, from a radiologist who reads every day and tests before he trusts. A few essays a month, nothing else.