Urgent.News

What's breaking now, across thousands of outlets.

AI

Y Combinator CEO has a message for founders worried about burning AI tokens

Y Combinator CEO Garry Tan encourages founders to spend significantly on AI tokens. He believes this approach grants access to future AI capabilities today. Tan suggests using AI agents for tasks and then documenting successful workflows. This method turns instructions into repeatable automated processes for businesses. While some debate this "tokenmaxxing" approach, Tan sees it as an investment.

Y Combinator CEO has a message for founders worried about burning AI tokens

Y Combinator CEO Garry Tan has advised founders to "burn, baby, burn" when it comes to using large numbers of AI tokens. Speaking on the A16z podcast, Tan argued that founders should not hesitate to spend heavily on AI agents, even if it results in high annual costs. He encouraged them to give themselves permission to use large token budgets when working with AI agents, stating that it can provide access to AI capabilities that will otherwise become available only in the future.

Tan emphasized that by fully loading AI agents with a high number of tokens, founders can effectively "live in 2028." While he acknowledged the expense - which could reach tens of thousands of dollars per year for full-capacity AI agents - Tan argued that for founders and CEOs experimenting with AI-driven workflows, this spending can be justified.

He suggested using AI agents for tasks first, then documenting successful workflows as instructions that can be automated in the future. However, Tan's approach contrasts with criticism of "tokenmaxxing," where companies deliberately allow AI systems to consume large numbers of tokens instead of focusing on lower usage and costs.

Written by urgent.news from Times of India's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at timesofindia.indiatimes.com →

More in AI

Life On Earth is 100% AI Generated Slop.

Hey everyone 👋, Most arguments I make in these posts start from code, from systems I built, or from some technical thing that broke in an interesting way.

  • Earth's behavior resembles AI-generated slop
  • Output plausible but not accurate
  • Lack of reliable grounding mechanisms

The Complete AI Software Development Lifecycle Explained

If you've shipped anything in the last year, you've probably noticed your workflow doesn't look the way it did in 2022. You're not just writing code anymore.

  • Requirement gathering uses AI to draft user stories from notes
  • Design generated by AI with human refinement for organizational context
  • Development sees AI drafting code which developers review critically

Problem Engineering: Why Defining the Problem Matters More Than Your Prompt

Why prompt tricks fail in production: an architectural guide to replacing prompt engineering with formal problem specifications.

  • Prompt engineering techniques face limitations in production environments.
  • Problem Engineering addresses specification errors, not model capabilities.
  • Structuring system invariants, data contracts, and observability hooks reduces production failures.

More from Saturday 22 August →