OpenServ's Tim Hafner on making agents reliable enough to trust with money

Tim Hafner, founder of OpenServ, on what has to change before an AI agent can be trusted to move money.

Tim founded OpenServ at the end of 2023 and builds reasoning infrastructure for AI agents. He bought his first Bitcoin in 2014 through LocalBitcoins, posting an envelope of cash to an address and hoping something came back. The conversation covers BRAID, the framework his CTO built to compile a vague English prompt into a decision graph the model has to run inside, so it cannot drift; why that lets a much smaller model do work that used to need a frontier one; why he thinks the moat in AI moved to data and distribution rather than disappearing; why he expects automation to collapse four roles into one person rather than delete four jobs; and what an AI with a controlling stake in Ethereum could do to the chain.

Humanode: X https://x.com/humanode_io | Telegram https://t.me/humanode | Discord https://link.humanode.io/chat | LinkedIn https://www.linkedin.com/company/human-node/

Tim Hafner: X https://x.com/open_founder | OpenServ https://www.openserv.ai | OpenServ on X https://x.com/openservai

00:00 Who is Tim Hafner

00:30 Buying Bitcoin in 2014 through LocalBitcoins

04:01 What was left for people without capital

05:57 What OpenServ builds on top of the models

08:09 Why AI agents fail at critical work

12:17 How long it takes to plug in

13:17 Frontier results out of a smaller model

16:32 Keeping customer data off US servers

19:07 Automating government work

23:14 AI accelerating institutions that are already broken

26:07 What happens when the electricity goes

28:24 Is there a moat in AI

34:20 Which jobs survive

38:15 An AI with a controlling stake in Ethereum

43:45 Close