Brian Armstrong on Bitcoin, Anthropic Drops Fable 5 & Mythos 5, NewLimit's $435M Age-Reversal | EP #264

Brian Armstrong argued that Bitcoin's recent weakness stems from AI absorbing a large share of global risk capital and from anticipated mega-IPOs pulling liquidity out of broad asset classes. He estimates roughly 70 percent of Bitcoin capital still treats it as a high-volatility risk asset similar to a tech stock, while only about 30 percent treats it as digital gold or an uncertainty hedge, meaning the counter-cyclical thesis he expected has not yet materialized. Armstrong said Bitcoin has probably bottomed around 60,000 dollars and expects it to reach 100,000 to 200,000 dollars by end of year, with positive momentum likely around October if prior cycles hold. Citibank projects Bitcoin reaching as much as 189,000 dollars by end of 2026, while Polymarket puts the end-of-2026 price at 84,000 dollars.

Armstrong described a three-step strategy for Coinbase to become the financial account for AI agents. Step one connects large language models to Coinbase accounts via an MCP API. Step two is an in-app agentic interface for portfolio rebalancing, tax loss harvesting, and yield optimization. Step three is self-custodial wallets on the Base protocol that any AI agent can open instantly with no KYC requirement. Agent transactions on Coinbase's network have grown to approximately 100 million transactions and about 50 million dollars in value, up from earlier figures of 3.1 million transactions and one million dollars. Armstrong argued that AI agents cannot obtain traditional credit cards and that legacy financial infrastructure runs on COBOL servers too slow for machine-to-machine transactions, making crypto the natural payment rail. USDC can be sent anywhere in the world in under one second for less than one cent.

Armstrong addressed quantum computing risk to Bitcoin, saying there is no imminent threat but that it is almost certain someone will eventually build a quantum computer powerful enough to challenge current Bitcoin and internet cryptography broadly. Bitcoin core developers have a proposal called BIP 360 for quantum-resistant cryptography, though it would increase block sizes, which is contentious in the Bitcoin community. The Ethereum team has a quantum-resistance roadmap estimated at about 20 percent complete. The most contentious question is what to do with Satoshi's coins, estimated at roughly five to ten percent of all Bitcoin and believed by many to already be lost. Armstrong outlined three options: freeze coins that do not upgrade by a set deadline, preserve Bitcoin's core guarantee against seizure even if a quantum computer eventually takes those coins, or implement a hybrid approach that freezes coins by a deadline but includes an appeal mechanism for rightful owners.

Anthropic released Claude Fable 5 and Mythos 5, described as the same underlying model with Fable 5 carrying added safeguards. Alex stated Anthropic has taken back the lead across most superintelligence benchmarks after GPT 5.5 had held the state of the art. The models demonstrate long-range visual reasoning including winning Pokemon games from pure screen observation. Dave noted Anthropic doubled the price of the new model, contradicting the idea of commodity intelligence. Fable 5 with safety guardrails will downgrade users to Opus if it detects questions in biology or chemistry that could be remotely dangerous. Alex predicted that as Anthropic and OpenAI approach IPO season the rate of model releases will increase and the companies will leapfrog each other more frequently.

NewLimit raised 435 million dollars for age-reversal research, with Armstrong and his co-founder putting in the first 100 million dollars themselves. The company focuses on epigenetic reprogramming, specifically changing the age of a cell without altering its cell type. NewLimit built a high-throughput screening system using AI to explore 10 quadrillion possible protein combinations and claims to have the largest wet lab dataset in epigenetic reprogramming. The company has demonstrated successful reprogramming of human cells and has first drug candidates going into the clinic next year. Armstrong predicted society will likely pass longevity escape velocity without broadly recognizing it, and Alex suggested it may be achieved in a spiky, subpopulation-level way as soon as this year or next year.

The hosts debated whether the US government should take equity stakes in leading AI companies. Senator Sanders has proposed transferring 50 percent of equity from top AI companies into a public fund, which Dave described as utterly insane. Armstrong argued government equity might be justified only if a matter of national security meant a company could not exist otherwise, but that government investing in companies that then donate to political campaigns would be the most toxic circle imaginable. Alex predicted the US Treasury could take golden shares of 5 to 10 percent in OpenAI and Anthropic, and that a future rebalancing could see the federal government gradually liquidate individual company holdings in favor of a total market index. Dave raised concern that a future administration could tank AI company stocks by dumping large government-held positions, and that government equity stakes are not steady tax revenue but require a future president to decide when and whether to sell.

OpenAI has officially filed its S1 to go public later this year. Polymarket gives 46 percent odds the IPO will come out at a valuation of 1.5 trillion dollars or greater, and 26 percent odds the IPO will not happen this year. OpenAI shifted strategy by shutting down Sora and its AI for science division to cut expenses and pivot toward becoming a coding-focused company.