Rob Schoening

@mitchellh I can’t give this up. No real interest in utility. Just joy riding.

Here's the ATC audio clip of my final landing ever. From humble beginnings in a DA40 to ~600 hours in a Vision Jet at 31,000 feet, I love to fly. Due to my growing personal and professional obligations, my final and best ADM decision as a PIC was to stop. This was my toughest landing. The skies were clear. The wind was straight down the runway. It's my home field I've landed at hundreds of times. But I knew it was my last. You can't hear it in the clip but my voice felt shaky. I'm doing my best to focus on a safe landing but I knew that was the last "cleared to land" readback I'd ever give. The landing was smooth. I parked the jet, locked her up, and gave her one last pat on the nose. We had good times. I'm no longer a pilot. Locking in for what's next. 🫡

Something I told 14 yo: There's a kind of politician who tells people "Your life is bad because <outgroup> stole what's rightfully yours. Vote for me and I'll get it back for you." They do it on both the left (Lenin) and right (Hitler), and they're invariably bad news.

I've had more "I can't believe it's this good" moments with GPT5.5 than any other model since Opus 4.5. It's shockingly, scarily capable. Days and days of amazing progress. All steering, no handwriting. Yet utterly delightful to conduct its coding. So, so good.

Burnout isn’t caused by hard work. Instead, it’s a decoupling of effort and outcome. You push as hard as you can and nothing moves. Your company has become a supertanker with its own inertia.

Marc Andreessen (@pmarca) on why Starlink may be the most misunderstood success story in tech right now: “Elon’s not the first guy who said we’re going to do satellite-based internet access. There was Bill Gates, Craig McCaw. Complete catastrophe, total bankruptcy, complete disaster. Elon’s like, ‘I know, I’m going to do another three of those. We’re starting as a side project at the rocket ship company.’ If the rockets are reusable, we’re going to be launching them all the time. What’s going to go in the rockets? I could wait for the customers to come to me, or I could just put up my own satellites. Anybody who knew anything about the history of satellites knew that was the craziest idea in the world. And of course it’s this giant success. It’s the side project. It’s clearly the least studied and understood thing I know of in the world right now.”

@aadilbuilds @rileybrown Neon is a phenomenal product.

@rileybrown neon is overly underrated, when you work at scale supabase cries on connections

Supabase will likely be worth more than Lovable Replit and Bolt combined when it’s all said and done.

The reason agents are so good at Linux is that all 40 million lines of kernel code was part of the pre training. Along with every other open source dependency. This really does make every obscure error message shallow, and the system completely malleable.

Pattern Recognition is also the form of intelligence that causes the most stress. You will see things that others do not. You'll feel crazy. Things will be *so obvious* to you, and others will just deny it.

A 6-person team is building task-specific AI models that are 4-8x faster than anything from OpenAI or Anthropic. 500K downloads on HuggingFace. No hype. Just better engineering winning on the merits. This is what "make something people want" looks like in the model layer. zeroentropy.dev

Trolley problem, California high speed rail edition

Nothing will shape your worldview as much as simply reading how humans from the past thought. You realize very quickly that our time is a radical anomaly in human history.

The right way to use AI in schools is to have sharply different policies about how much you allow it. Using AI should be encouraged in some situations and absolutely banned in others.

Been limiting my showers to 1 minute to save more water for the data centers.

Bull case: people are thinking before they write and the production becomes easier Bear case: the thinking is "write me a tweet about this data" OpenAI data on ChatGPT usage shows that across knowledge work, AI is most used where artifacts get produced. When lawyers use ChatGPT, drafting documents (45%) beats asking for advice (42%). AI is used more to produce artifact rather than make the judgment. For most expert work, the artifact is writing. For engineering, it's code. For IT, it's procedure. In 9 of 12 business functions, the dominant artifact is writing. When using AI, Finance writes memos more than it builds models. Sales drafts proposals more than it closes deals. Researchers write papers more than they run experiments. As artifact production collapses in cost, thinking before producing becomes asymmetrically valuable. h/t @RonnieChatterji and the @OpenAI Signals data *disclaimer, this tweet was written by hand

@typesfast I don’t know why you don’t notice, but it is definitely improving.

With all these AI coding improvements why isn't the software I use everyday getting better?

@allenholub Communication and logic.

Programming has nothing much to do with math. It is primarily about communication. It's about us communicating with customers and us communicating with the machine. Programming is a language skill, not a mathematical one. Consequently, I get tired of people claiming that programming has anything to do with math. Beyond a little set theory and statistics, I've never used any real math in my decades of programming work. Logic is a subdiscipline of rhetoric, not mathematics. Analytical thinking (and problem solving) is part of almost every human activity, from carpentry to oil painting. Math and programming both use those skills, but so do a million other disciplines. Some programs implement problems in mathematical domains, but in those cases, the math is part of the domain, not the programming process. Some programs involve dating profiles, but that doesn't mean that dating profiles are an integral part of the programming process. I think this false equivalence dates back to the very early days of computing, when computers were used almost entirely to solve mathematical problems. In the present, mathematical problems are only a tiny fraction of the problems we solve. Of course, computer science, which is that branch of mathematics concerned with the analysis of computer programs and algorithms, uses some math. (Though I never used the required 1.5 years of calculus and differential equations in a single CS class.) I'm talking about programming, not computer science, however. So, returning to my original claim, if you want to be a good programmer, focus on developing communication skills. Math is irrelevant to the vast majority of us.

@ivanburazin You have a cool platform.

Building agents on hyperscaler infrastructure is like building a sports car on a truck chassis. Hyperscalers were built for stateless workloads. You don't want your web app to change on the fly. You just want it to be stable, predictable, and immutable. Their entire architecture was designed around that assumption. Agents are the opposite. They need state. They need to pause and resume. And they need to run until the job is done (not until a timeout kills them)

prediction: by the time most enterprise buyers say they want an "ai strategy," they really mean they want permission to cut cycle time without looking reckless. founders who pitch magic get curiosity. founders who pitch compressed headcount, fewer handoffs, and auditability get budget.

Everyone on the How I AI team is brainstorming how to get Reese on the podcast, because this is SUCH an important message, especially for the ladies to hear (btw, my youtube channel subscribers are still EIGHTY SIX PERCENT MALE) If you have an in, please connect us!

we're going to hit 1M daily active users in the next few weeks the whole way here almost all the thought leaders kept explaining how what we were doing was wrong, bad taste, wouldn't work, etc none of them were curious enough to ask us what they got wrong