Brian Armstrong (Parody)

We were able to fine tune an open weight model (Qwen3.5-9B) on proprietary internal data at @coinbase, and it out performed frontier models, at much lower cost and speed, for a specialized task (fraud prevention in this case).

I think we'll see more of this pattern inside companies. The frontier models are amazing generalists. But increasingly you'll see specialist models out perform on more narrow tasks, either working on teams, or by themselves.

We're making it easy to fine tune models internally for our employees. We even have a skill for it in our internal harness that makes it easy to spin these up and start training them:

"/fine-tune-and-serve-llm I want to fine-tune Qwen for payments risk review. help me get started"

I imagine we'll have some backend services that could be replaced entirely by this new architecture.

It’s great to see the SEC proposing merit-based private market access in the US. This will be a huge step forward.

For too long, only the richest people have had access to the private markets. Retail has been missing out.

And over the last few years, companies have been incentivized to stay private longer (so retail has been missing out even more at best, or become exit liquidity at worst).

Amara's law: "We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run."

The financial system is moving onchain. One of the biggest shifts happening over the next decade. Especially with the adoption of agentic finance.

Thanks to everyone that came to our first IRL trading competition or tuned in online - it was awesome!

And special shoutout to the 10 traders for putting on an awesome show, we'll have to do it again 🙏

We need more small-scale experiments in governance or sandboxes for innovation at the federal level. Debating every reform nationally makes learning unnecessarily expensive and slow.

Competition is the best consumer protection ever invented, and the government is the one sector structurally exempt from it.

Run more experiments, with faster permitting, deregulation, or different rules in one place (say 10 square miles of federal land). Measure the results and let successful ideas spread.

Tech is becoming a huge part of culture. Founders and engineers are role models, and I think that's a good thing.

Kids should aspire to build something people want and contribute to society.

One of the biggest killers of innovation is being blinded by hypothetical downsides. It’s easy to raise concerns about why something might be risky. It's much harder to consider the potential harm of putting too many restrictions in place.

Consider the de facto ban on nuclear with the establishment of the Nuclear Regulatory Commission. Today we'd have a modern, robust electric grid, and much higher economic growth, if we had kept building nuclear power plants.

All around us we see systems that were well-intentioned when they were introduced, but had unintended consequences.

How things fail:
- Asymmetric error costs - approve a bad reactor or drug and you’re famous. Yet if you block a good one, no one ever knows.

- Process as a veto - hourly fees, multi-step license applications, EPA reviews. Endless delays erode capital.

- Worst-case models treated as fact - LNT radiation, any sonic boom is unacceptable, etc.

- Rules written for the last accident, applied to the next design - we see this in everything from nuclear to financial services.

Change is uncomfortable, but if we stop progress, we do a disservice to our ancestors who tolerated discomfort to bring us the modern age. We owe it to them to continue making progress.

Keep building.

The midterms are about 5 weeks away.

Crypto isn't a partisan issue. It's about whether the next generation of financial technology gets built in America.

Before you vote, check where your candidates stand: https://standwithcrypto.org

Most companies would move faster if internal platforms had to win their internal customers the same way they win external ones.

Publish your uptime, latency, and cost. Take feature requests in the open. And if you're a real product, it's fine to charge internal teams like any other customer.

Agents make this practical. They can discover services, compare them, and route around the bad ones.

Every backend service should have a /feedback endpoint.

Agents are quickly becoming the heaviest users of most APIs. When one hits a missing feature or a bug, it should be able to say so right there, in a structured way.

If the request makes sense, another agent drafts the PR and a human approves it. Software that improves itself based on what its users actually tried to do speeds up recursive self improvement.

I'm hosting a dinner next month in Boston for scientists and engineers interested in safely increasing human intelligence.

I'm interested in both treating cognitive impairment and helping healthy people improve their cognitive abilities. Could we develop treatments that help people learn faster, reason better, and solve harder problems? I'm especially interested in approaches that could work in adults.

There's a lot we don't know, including what's possible and what would be safe. I'd like to bring together people across neuroscience, genetics, drug development, and computational biology/ML to discuss promising research and what it would take to turn it into useful treatments.

I'm also interested in backing the right founding team if one comes together.

If you're working in this area or have relevant expertise, please apply at the link below.

Happy to cover travel if needed. Or if you can't make it to Boston, let us know - I may host one virtually for those who can't attend in person.

AI doomers say if we build AI we might die.

But they have it backwards: 165,000+ people already die every day. Cancer, heart disease, illnesses, car crashes. AI is helping us solve these.

Slowing progress doesn't keep us safer, it guarantees more preventable deaths.

The 2024 election proved the importance of the crypto voter.

@standwithcrypto is now much bigger, with over 4M members.

Crypto voters will remember who stepped up and who didn't.

Here we go again! The WSJ is working on a story blaming Coinbase and me personally for the CLARITY Act not passing.

The Journal has repeatedly been hostile to CLARITY in its reporting, regurgitating bank lobby talking points, while I’ve spent years pushing for crypto legislation, but that’s not stopping them from trying to reverse the blame.

The boring TLDR: in January I opposed a draft of the bill going into a committee vote, because it needed a lot of work on DeFi, tokenization, CFTC authority, and stablecoin rewards. At the time, the bill had major issues that would have harmed crypto. Support was fractured, and it wasn’t passable. We worked with a number of parties who improved the bill and made it passable. All four of the items I called out were fixed in the draft that then went through the committee about four months later. I'm proud to have done it, and would do it again, because it helped create a better bill. One step of many along the way.

The final draft of CLARITY that went to the Senate was great, and I strongly supported it. I'll continue showing up for our customers and pushing for clear rules that treat crypto fairly, even if those who feel threatened by crypto try to plant false stories. It's a shame that the WSJ takes direction from bank lobbyists instead of reporting the truth, but luckily, people are smart enough to see through it these days, and it backfires on them every time.

If you haven't tried our self-custodial @CoinbaseWallet yet, download it and let us know what you think.

We're shipping major improvements weekly, led by @cobie.

The CLARITY Act didn't advance in the Senate today, which was a disappointment. While it's possible bi-partisan conversations continue and it lives to fight another day, we can't wait on Congress anymore.

The SEC and CFTC have the tools they need to create clear rules under existing authority, and I expect will begin working on this in earnest.

So clarity is coming to crypto regardless.

And of course GENIUS is already the law of the land for stablecoins, which is even more permissive on rewards. There were some concessions we made on CLARITY that were tough to swallow, so perhaps it's for the best.

Crypto can't be uninvented. With clarity emerging through the regulators, we'll continue updating the financial system.