Can AI Teach Itself Superhuman Persuasion? Introducing Open Debate
I asked AI agents to debate "therapy culture." Five rounds later, both sides were calling for a violent communist revolution.

Building verifiable, truth-seeking AI agents
I'm building AI agents that verify their reasoning on complex, controversial questions — surfacing counter-arguments, stress-testing logic, and turning confusion into clarity. This will 100x intellectual work and make real-time fact-checking possible.
I'm based in London, UK.
An AI debate arena where language models argue opposing viewpoints, get judged, and learn from their losses. Run debates that would never happen IRL — people who refuse to engage, won't share a stage, or are ideologically incompatible.
Real-time argument analysis as you write. Two AI agents work in the background: Steelman surfaces the strongest version of your argument, Counter shows the objections your opponents will raise.
Public conversational AI that explains energy and climate issues through Alex Epstein's lens. Built on his books, articles, and decades of published work.
Research and writing platform for energy policy advocates. Runs multi-step workflows that search Alex's knowledge base, crawl web sources, and produce structured documents.
Built an AI agent for Blumberg Capital that screens incoming pitch decks, scores deals against investment criteria, and syncs with the CRM — making the VC process faster and more founder-friendly.
Read the case studyI asked AI agents to debate "therapy culture." Five rounds later, both sides were calling for a violent communist revolution.
The word "computable" is doing a ridiculous amount of hidden work in AI debates
How to find truth in the age of "podcastistan"
So you have a superpower, now what?
Head of AI at Alex Epstein | AI Consultant