Skip to content
cloud-codex / ~ % _
hometagsgithubresumeabout rss
~/codex / series / ai-assessment

// series · 7 parts

Building an AI Assessment System

A field build of AI-graded assessment — grading written answers, testing the judge, red-teaming it, and measuring what actually moved the needle for students. Read in order.

  1. [ 01 ] The AI Tutor Everyone Builds Is the One Students Ignore Jul 06, 2026 · 6 min read
  2. [ 02 ] Grading Written Answers with an LLM, Properly Jul 07, 2026 · 5 min read
  3. [ 03 ] Your LLM Judge Needs a Test Suite Jul 08, 2026 · 4 min read
  4. [ 04 ] Students Are Adversaries: Red-Teaming an LLM Grader Jul 09, 2026 · 5 min read
  5. [ 05 ] The Grader Was Right and the Students Quit Anyway Jul 10, 2026 · 4 min read
  6. [ 06 ] The Biggest Effect Size Was the Boring Feature Jul 11, 2026 · 4 min read
  7. [ 07 ] Instrument Like a Learning Scientist Jul 12, 2026 · 4 min read
cloud-codex

Field notes on building with AI agents — the harness, the tokens, and the platform engineering that keeps it production-grade.

explore home work series tags archive about
elsewhere github x bluesky linkedin resume rss feed
© 2026 michael tuszynski privacy built in public
Press ESC to close.

This site uses cookies to measure traffic via Google Analytics. Our other analytics are cookieless. See our Privacy Policy.