Vibeleaderboard
← All Intel
Intel / video

Tokenmaxxing: How Top Builders Use AI To Do The Work Of 400 Engineers

Source
youtube.com
Author
Y Combinator
Date

Context

Garry Tan, who runs Y Combinator and returned to programming after years away, describes maxing in this Light Cone episode as spending more calls and whenever extra work makes a result more complete. A token is the unit of text a model reads or writes, so this means paying for more of it. The example is Garry's List, a California political advocacy site where an agent researches issues and drafts sourced articles. Rather than settle for one source, Tan says, it can gather about 20, note that 13 say one thing and seven disagree, and feed all of that into the .

Tan says the site took about five days and roughly $200 of a Claude Code Max subscription from Anthropic to build, and that an article costs about $5 to $10 of Opus calls for work Tan compares to a human researcher going through dozens of articles and books. Tan also turned instructions retyped often, such as asking Claude for diagrams of data flows and error cases before coding, into a toolkit called G stack. These time and cost figures are Tan's account, and they say nothing independent about article quality.

Terms in this piece · Glossary
  • Token — A piece of text, such as a word, part of a word or a punctuation mark, that a language model represents as numbers and handles as a unit.
  • Model — A computational system that uses learned patterns to produce an output from information supplied to it.
  • Context and context window — Context is the information available for a generation; the context window is the capacity available to hold it.
  • Prompt — Information or instructions supplied to a model; in a chat, the request you type is often called your prompt.
Read the source www.youtube.com
More from Y Combinator
Recommended reads
Comments

Checking sign-in…

Loading comments…