Christian Ulstrup

I actually don't understand the discourse around "hurr durr AI is hitting a wall," especially now.

I felt like it was plateauing maybe 3 months ago and started planning accordingly.

Then o1 comes out, and I'm telling you, IT IS A HUGE DEAL FOR APPLIED AI.

I don't care what the benchmarks say. I don't care what the experts say. All I know is that across clients across industries, I have personally been able to tackle new problems faster and generate more free cash flow with o1-preview than without.

My eval is whether or not I get paid, and that's continued to work pretty well as a means of developing an intuition in re which tool to reach for, when...

Going through the experience of building a piece of software (or at least some kind of scalable, reliable, and valuable digital content) that at least one person pays for should be a core part of standard education.

It's how you build leverage (along with figuring out your vocation) without having to go into debt.

"But AI will kill SaaS..."

Not exactly — the invention of construction equipment led to grander, more complex buildings that couldn't have existed prior.

Reality (and, thus, latent space) is infinite, so the number of problems to be solved is also and infinite; therefore the amount of software that can be built to bring a little bit more order to the world, whether for an enterprise or single neighbor, is nigh-infinite.

Big problem with "Product" professions per se over the past 20 years is that we assume:

• experimentation (by expert operators to who have a deep love of a particular "game" in service of some trusted network of clientele),

• discovery (and the word-people who have a knack for it), and

• development (by highly specialized software engineers et al)...

...need to be decoupled. This assumption is, now heading into 2025, simply untrue.

Sure, at certain levels of scope, scale, and profitability, it may make sense to invite specialists to collaborate, but—and I don't think people want to hear this—you can bootstrap a project to that point with AI assistance and sufficient desire.

When I was 10/11, I learned a decent amount about "entrepreneurship" going door-to-door with other neighborhood kids after a hurricane and selling yard cleanup jobs. Figuring out how to efficiently block and tackle delivery of marketing calendars for a realtor in a way that would both help them reach the most valuable prospects but also ensure it was worth our time. Cold-approaching adults at a local Starbucks to help raise funds for a school program.

Similar lessons can be learned in cyberspace (and there's more leverage to be gained) — I look forward to arming my own kids with what they need (eg Claude, Chat, @Replit) to figure it all out, and it would be great if our educational institutions were set up to support these kinds of initiatives: i.e., p/f classes where the outcome is contingent upon nps and p&l, not whether you were able to memorize some set of rote facts.