Jonathan Mills Patrick

PayPal's NFL deal isn't a sponsorship — it's distribution. Embedding into an audience instead of buying ads to build one. Community banks have a smaller version of this move sitting unused: local leagues, employer payroll, chambers of commerce.

@ML_ITM @aryanlabde 🔥Matt. Spot on.

@jmillspatrick @aryanlabde This is the key distinction. Strategy first, always. Automation is a force multiplier for a good strategy, but it's a disaster multiplier for a bad one. The most exciting things being built now are tools that help execute on a sharp strategy, not just 'automate all the things'.

Why is nobody building a SaaS that helps with marketing? that’s what most people here would pay for.

@briannekimmel Agree with a caveat. Founder-led growth works until a point in time. I've not seen any metrics around this, but FLG is not scalable past a point. Likely the moment the founder becomes a bottleneck.

I hate to be the bearer of bad news, but a founder’s inevitable role in GTM is consistently posting on LinkedIn.

Love this GTM play from Stripe. It's community-led growth at its finest. No fees between listed business is 1) an amazing perk; 2) super sticky; 3) creates referencability.

Three federal agencies (Fed, OCC, FDIC) just rewrote the rulebook for bank audits around risk—specifically AI risk. Most FIs hear the words "audit", "compliance," and "regulation" and go 😱. Not me—I see a unfair GTM advantage waiting to leveraged. Tomorrow my latest article covers the new rule and what to do about it.

@garyvee I’ve been saying for a while that we will see a push back towards humanity.

As we move toward 2050, I think we're going back to 1850. More live events, more real world interaction. As AI grows, so will analog.

@jakefenner_ @AmericanAir Zero shock (if true). Airlines are ripe for disruption from someone who can optimize for expeditious boarding and on-time flights.

I would like to thank @AmericanAir, the official airline of the World Cup, for being horrible at their jobs. Supposed to land in Dallas at 10:30 to watch Lionel Messi make history. It’s 12:44 and thanks to hours worth of delays at LGA and DFW, I missed the record-setting goal.

@MarcBaselga Conway's Law.

I was looking through the product roles OpenAI and Anthropic have open, and the contrast is pretty interesting. OpenAI seems to be hiring PMs to turn model capability into a portfolio of products. Anthropic seems to be hiring PMs to make Claude easier to adopt, sell, operate, and scale. This is a hiring-map read, not a leaked strategy memo. Roles change, titles are imperfect, and this is not definitive strategy. But still, hiring pages show where companies think the messy product work is. In the snapshot I reviewed, OpenAI had 14 roles I focused on. Anthropic had 9 PM-titled roles in the same filter. OpenAI's map feels broad. Some are obvious product-business lanes: ChatGPT Healthcare, Families, Shopping, Premium Subscriptions. Some are platform, infrastructure, or monetization work: API Agents, API Infrastructure, Financial Engineering, Self-Serve Business Growth. And a lot are safety or trust-related: Bio Safety, Cyber Safety, Integrity, Private Computing & Encryption, Safety Measurement, Sensitive Deployments. When I squint at that list, OpenAI looks like it is turning frontier model capability into a bunch of separate product lines: consumer surfaces, vertical products, developer infrastructure, monetization, privacy, safety systems, deployment controls. Very portfolio-company energy. Anthropic's map feels more concentrated. The PM-titled roles cluster around API Growth, Claude Code, Compute Platform, Developer Productivity, Enterprise, GTM Experiences, Multi-Cloud Growth, Consumer, and Safeguards. That feels like a company hardening the operating system around Claude: developer workflows, enterprise adoption, compute, distribution, GTM, trust. OpenAI's safety work looks more explicitly productized in this snapshot. Six of the 14 OpenAI roles are safety or trust-adjacent, which suggests safety is becoming PM-owned work: measurement, controls, launch gates, deployment decisions, user-facing boundaries. Anthropic shows safety differently in this dataset. Fewer explicit safety PM lanes, but safeguards and trust are tied into the Claude platform and enterprise adoption story: controls, model performance, customer workflows, and the operating discipline around how Claude gets used. For candidates, the takeaway is team specificity, not "which company is better." OpenAI Healthcare, Shopping, API Agents, Private Computing, Safety Measurement, and Subscriptions are very different PM jobs. Anthropic API Growth, Claude Code, Enterprise, Compute, GTM, Safeguards, and Consumer are different versions of the Claude platform loop. Moving forward, I am going to start paying more attention to hiring pages as strategy docs. Sure, they are noisy and imperfect, but they show where vague capability is becoming owned product work. P.S. If you're interviewing for a product role at OpenAI or Anthropic, this is exactly the kind of company-specific prep we built Insider Loops for. The public job title is just the first clue. insiderloops.com

@FreddyLA7 I love this. I never thought about turning the in-between time into a massive road trip. I pray you enjoy it!

This is our plan for the coming days. The next game we’ll attend is Germany’s Round of 32 match in Boston on the 29th. Since we have 7 days to get there, we decided not to take the direct route and instead make a huge detour through Michigan, Ohio, Pennsylvania, and a few other states. Any recommendations for places or cities to visit along the way are very welcome!

@defyneric Tempting. Say more.

if i hosted a poker event in nyc who would want to attend?

@omarvvvr Market validation.

I am a founder scare me with 3 words

@thesamparr Network = net worth. Even in an age of technology, it will ALWAYS be true.

Amjad wanted to get on Joe Rogan. So he just started visualizing it. What happened next is nuts: - A few weeks later Marc Andreessen AND Lex Fridman both message him the same day. - Both asking the exact same thing: can you help a friend's daughter build an app? - Turns out it's Joe Rogan's daughter. - Amjad builds it with her, she wins her competition. - Rogan calls to say thanks. That's how he got on the show.

Stop trying to out-spend the megabanks on AI. Pick one use case — fraud alerts, loan doc review — and get the data right first. That's the actual unfair advantage.

That's not a smaller version of BofA's data problem. It's a fundamentally easier one to solve — which means community banks can deploy trustworthy AI in narrow use cases faster, not slower.

Bank of America just spent $250M on AI this year. Most community banks read that number and feel outgunned. They're reading it wrong.