Vibeleaderboard
← All Intel
Intel / video

AI-Native Engineering: 100% adoption, 5x search throughput, unlimited tokens — Mikhail Parakhin

Source
youtube.com
Author
Latent Space
Date

Context

Mikhail Parakhin, Shopify's CTO, says on Latent Space that spending on AI , the chunks of text models and bill by, is a poor goal by itself. Shopify funds effectively unlimited tokens for employees, and Parakhin calls NVIDIA CEO Jensen Huang's push for heavy token use directionally correct, but says how tokens are spent matters. Running many agents in parallel that do not communicate is nearly useless, Parakhin says. Fewer agents with a critique loop work better: a second agent, ideally from a different , reviews the first agent's work and sends it back.

Critique loops add and people dislike the wait, but Parakhin says code quality is higher. Because AI writes far more code, review must be strict, and Parakhin tracks the ratio of budget spent generating code to budget spent on expensive models checking pull requests. A good model now writes code with fewer bugs than the average human, Parakhin claims, but more code reaches production, so bugs can still rise. These are management observations, not a study.

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.
  • Process — A running instance of a program, with its own working memory, open files and identity.
  • Model — A computational system that uses learned patterns to produce an output from information supplied to it.
  • Latency — The time between asking for a result and having it, including the model's generation time, tool calls, retries and any queueing.
Read the source www.youtube.com
More from Latent Space
Recommended reads
Comments

Checking sign-in…

Loading comments…