Patrick J. Auld

I'm paying for Claude, Cursor, and GPT. I'm not maxing all of them at the same time. But last week I hit my Claude limit mid-migration and just... waited. So I built a market for that. SecondTokens is now in private beta. 🧵

People don't lie down and get bored. We find, discover, invent, and imagine new, novel, and wonderful things.

Nearly 1% of people in the United States used to be employed as switchboard operators. Something like 20% of the pre-car economy was related to horses and 10% of jobs. Everyone used to be in agriculture.

People are amazing. To somehow believe that the next generation, when faced with a changing society, won't find new and unthinkable uses for a general tool is ridiculous

I am aghast at the number of people who believe generations of humans will be incapable of anything because of AI, as though human will, spirit, and creativity have never overcome.

Is it worse to be trapped with poor leadership for years, or to force leaders into permanent campaign mode where every decision is optimized for approval ratings?

I've "manually" coded maybe a dozen lines of code this year. And 100% of them have been in the GitHub web editor on a PR. It was faster to make the change that way during a review of my agent's code than to re-prompt it. Crazy time to be alive. I'm loving this ride AI is giving us.

Clean code is something I've focused on enough to be told my code "looked like it was from a textbook" (I'm still beaming from that compliment, a decade+ later). I don't feel bad about focusing less on it now. I get more done. Which I always cared more about. It is honestly confusing to think about my feelings on coding right now.

Did I ever care what my code looked like? In hindsight, I cared about making something I could use later. It meant clean, readable code that I could step back into and tactically update and improve in the future. I have a vivid memory of when I first grok'ed the Visitor Pattern when reading a more senior co-worker's code.

Question of the year: Can AI increase your Bus Factor? > 3. Coinbase is testing "one person teams" where a single person is the engineer, the designer, and the PM. A pod of one with agents.

CS schools should keep teaching what they have been. AI doesn't change it. Hot take: University never prepared you for the workforce. It's not the point. Most of the things I did in college, I never did in my work. Coding an OS from scratch, low-level assembly, building game simulations in niche languages? It's not day-to-day. The point of those was to expose me to different ideas, expand my thinking, and show the foundational components. With that understanding, I could write SaaS applications. Quick aside: I was so annoyed at one point by how unprepared I and every new hire I got were that I made a "how to work in the industry" class. I even pitched it to some professors' friends. I dropped it after realizing it's not what should be taught. People come out of college and do all sorts of things. There is no "standard" across companies and industries worth teaching before they get there. Universities still teach assembly and compiler design, even though a tiny fraction of students ever touch them again. It's an important context because there are times that it matters. and understanding how problems have been applied to those domains enriches your ability to solve other things Parts of AI will have to be integrated into the curriculum. It's a tool that we're going to have. These are additive, though

"Entry-level jobs are going to disappear!" Many will. The skill set is changing. It's just being sensational not to include, "and all the new jobs will be different in ways we don't know yet". Who thinks a bunch of young adults is not going to find uses for AI? How could you believe they won't become the most capable users of it? Entry-level jobs are going to change. People who are early enough to the new paradigm won't be entry-level; they will be the category-defining leaders in it. I hope I can find my way through it, as well as the younger generation will. People are powerful and creative. The youth are adaptive and entrepreneurial. I'm not worried about them.

These are all secondary market actions for reserved TSMC capacity at the end of the day.

Anthropic is an Enterprise B2B software company, and your personal subscription can best be thought of as an entrepreneurial/startup subsidy.

If you're firing people because of AI, it means you believe there are people in your organization who, if they were 2x more effective, would not meaningfully contribute to your business goals. That's not an AI problem. That's a organizational problem.

What's the best way to use a Chief of Staff? For people who have never utilized one before, I'm finding it is the replacement and leverage they provide?

It's a major leadership 'tell' when the reaction to AI is to reduce headcount rather than to dream up bigger goals.

Your AI subscriptions have spare capacity. Someone else just hit their limit. SecondTokens is a market that connects the two. Sell what you're not using. Buy more when you're blocked. No credential sharing. Execution stays local.

I'm taking a small group of early users to shape the first release. Sign up for the info and the private beta → secondtokens.com

I built this for people who know what "you've exceeded your usage limit" feels like at 2 am during a release. If you live inside Claude Code or Cursor, this is for you.

I'm paying for Claude, Cursor, and GPT. I'm not maxing all of them at the same time. But last week I hit my Claude limit mid-migration and just... waited. So I built a market for that. SecondTokens is now in private beta. 🧵