Andrew Fribush

fig. 0.4.1: the ledger, counted by document

fig. 0.4.1 · the ledger, counted by document.
LIVE FIGURESlibrary lab18the act–check weave1homer lab14home15colophon3charts3projects1women should ask men out5the setup cost of a faster query1knuth's root of all evil has a break-even point9feed the stranger before you ask his name3i built compaction without a model: 13 lost results against 5144testing api wrappers for silent empty results2spending limits for autonomous agents3calibrating an llm judge3why the grain price drifted3counting the assumptions a test shares with its code3keep the model out of the control loop2claiming tasks with git refs2an ingredient is not a contaminant3designing apis for ai agents as first-class users3three algebra errors in "existential risk and growth"3how a student model gamed its verifier3one call, two ids3testing assertions by reintroducing bugs3lazy tool loading over mcp3ten review rounds for a flashcard strength map2embedding 150 million records for about $2303what 1,097 citation checks found in fourteen well-known papers3your mcp tool descriptions are probably being thrown away4breaking a 15,000-row cap3confabulation is a cache-coherence bug2every child since myra3the second that was added at midnight3where the 8 percent came from3the guest-gift is the receipt1tooling lab1

All figures

Andrew Fribush, software and computer engineer in Austin. I build consequential systems where software meets institutions: finance, healthcare, commerce, government data, and the mechanisms that govern AI agents. Taking new engagements: hello@andrewfribush.com.