Tom Walczak
PostsAbout

Posts

Writing about AI, truth-seeking, and building verifiable AI agents.

Feb 3, 2026

Can AI Teach Itself Superhuman Persuasion? Introducing Open Debate

I asked AI agents to debate "therapy culture." Five rounds of recursive self-improvement later, both sides were calling for a violent communist revolution.

Jan 9, 2026

Demis Hassabis Is Wrong About Computability

The word "computable" is doing a ridiculous amount of hidden work in AI debates

Jun 26, 2025

What 200+ AI Debates Taught Me About Truth-Finding

How to find truth in the age of "podcastistan"

Jun 16, 2025

AI is Like Waking Up with $100 Million in the Bank

So you have a superpower, now what?

May 25, 2025

Making Policy Costs Personal With AI

People support bad policies—until they learn what they'll personally pay

Jan 30, 2025

What I've Learned From a Year of Building an AI Agent for a Top-Tier Early-Stage Venture Capital Firm

How we built it, pitfalls to avoid, and how AI is making the VC business more founder-friendly.

Jan 4, 2025

OpenAI's o3 and the Problem of Induction: AI Doesn't Learn Like Humans (Yet)

What If It Could?

Dec 12, 2023

How We Built AlexAI: The World's Top AI Energy and Climate Expert

How AI can help make sense of complex and contentious debates by generating clear explanations

© 2026 Tom Walczak. All rights reserved.