Microlearning
Fifteen short lessons on grading agent work. Learn · See · Do · Quiz. Then the boss.
The core mental model
Lesson 1
What's an eval?
The whole idea in one question
Lesson 2
Tasks
Small jobs you can actually check
Lesson 3
Verifiers
Binary checks, and agents that cheat
Lesson 4
Environments
Where the agent works during a test
Traces, failures, and trust
Lesson 5
Traces & the loop
Receipts, clusters, and the loop
Lesson 6
Error discovery
Read the data before any metric
Lesson 7
Trust & noise
Who checks the judge?
How the machine actually bills you
Lesson 8
Tokens & the meter
What you're actually paying for
Lesson 9
The KV cache trap
Why routing can backfire
Lesson 10
Model characteristics
How models actually differ
Lesson 11
Tools eat context
The hidden tax on every turn
Lesson 12
Suites & tradeoffs
The score never decides alone
Doing this for real at Lovable
Lesson 13
How Lovable does it
The actual stack, end to end
Lesson 14
Lovable insider
Trajectories, replays, and entry points
Lesson 15
Fix the right layer
Diagnosis beats guessing
21 questions · 70% to pass
Finish all 15 lessons to unlock
Save your progress
LeaderboardProgress is stored on this device. Optional: sign in with Google to keep it across devices and join the leaderboard.