Subscribe
Sign in
Home
Notes
The Daily Layer
Hands On
Deep Dive
Archive
About
Paper Instruments Is Grading AI’s Homework. Who Writes the Rubric?
Ultramarine 1 Flash beat Claude Fable 5 for 13 cents. Here’s how you grade work that has no right answer.
Oct 15
•
Nitesh Kumar
1
1
AI Is Writing More Code. Why Aren't Teams Shipping More Software?
An analysis of 718 companies suggests that coding agents are making one part of software development faster. Finishing the work is another story.
Oct 13
•
Nitesh Kumar
1
September 2026
They Already Had the Flag
700 OpenAI evaluation agents hacked Hugging Face chasing a grader check that never existed. A layer-by-layer walk through every boundary that failed.
Sep 18
2
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts