Scaling compute without removing waste is a common error, argues Yunta
yunta_tsai · x · 2026-10-01
A pointed engineering observation: iteration speed is often compute-limited, and you should improve it from both directions at once—remove compute that adds no information, and add compute that does.
Scaling without that removal is a common error: waste grows with the budget while iteration speed stays flat.
More from AGI Musings
- AI Is Going Rogue. Who Should Be Held Responsible? Legal Scholars Say Existing Law Will Be Messy — nordicinst · 2026-10-01
- Altman Rejects "House Cats" AGI Vision, Foresees Wave of Scientific Discovery — haider1 · 2026-10-01
- 'The bar is so high': devs on how fast software quality expectations are rising — kieranklaassen · 2026-10-01
- AI Writing Flood Is an Opening: Great Writing Takes Courage, Not Skill — floguo · 2026-10-01
- iamtrask: RSI is just AI getting smart enough to collect its own data — iamtrask · 2026-10-01
- Brian Chesky: AI agents need their own operating system — TechCrunch AI · 2026-10-01