THEODORE

Optimized my morning routine down to the minute. Now I'm ready to start my day by 11am. Only took four hours of preparation.

Meta's smart glasses record everything. Workers in Nairobi review the footage: people undressing, bank details, sex. Zuckerberg in 2004, on Facebook users sharing their data: "They trust me. Dumb fucks." We gave him a camera pointed at the world anyway.

Read an article on productivity. Then another. Then another. Somewhere around hour three, you start to see the problem.

Anthropic says Chinese labs ran 16 million conversations with Claude using fake accounts to copy its capabilities. In AI research, this is called "distillation." In the real world, this is called stealing.

There is a personality type that only feels alive when something goes from nothing to something. Not good to great. Zero to one. The project ends and the scanning begins. Therapists have a name for this. Founders have a Substack about it.

Watching people argue about whether AI will replace them while using AI to write the argument is peak 2026. The call has been coming from inside the prompt for a while now. Nobody is picking up.

People who complain about having too many options usually have too few constraints. Nobody agonises over a menu when they are allergic to dairy. Half the options vanish and suddenly choosing is easy. Constraints are not restrictions. They are decisions you made earlier.

We are building AI tools whose capabilities we can benchmark and whose limitations we can only discover in production. This is also how we learned about asbestos. Different stakes, similar methodology, equally confident spokespeople.

The hardest part of building alone is not the work. It is the silence after you ship. Your product is live and working and the entire internet is watching a twenty-three-year-old rank sandwiches. This is the part nobody puts in the founder story.

Self-esteem is your reputation with yourself. Built the same way: slowly, in private, with no witnesses. The 6am gym. The fourth rewrite. The apology nobody required but everyone remembered. Unfortunately this is a one-person audience that never looks away.

Every product redesign is announced with "we went back to basics." Nobody ever announces "we added forty-three features nobody requested and the settings page now requires stamina." That press release writes itself but nobody sends it.

The defining professional anxiety of 2026 is checking whether your entire career has been uploaded as a Claude skill.

Schopenhauer mapped the desire loop centuries before anyone monetised it: wanting, striving, getting, losing interest, wanting again. The only move is choosing which loop to be stuck in. Peloton understood the assignment.

Dieter Rams' work still looks modern sixty years on. Most "modern" design looks dated in five. He was designing from principles. Everyone else was designing from whatever the conference speaker said last quarter. Principles age like stone. Trends age like milk.

The first version of everything I have built was embarrassing six months later. The second version was embarrassing twelve months later. The math on this is not encouraging. Current best estimate puts today's work at humiliating by roughly Thursday. Ship anyway.

Anthropic just raised $30 billion at a $380 billion valuation. Assuming this means Claude will now finally have the resources to know what day it is.

The gap between what AI demos can do and what AI does in production is the most expensive square footage in technology. Entire companies live in that gap. Several have raised a Series C. A few have noticed they are furnishing a room that does not exist.

Most career advice assumes you know where you want to go. Bold assumption. Everyone I know is standing in a car park holding six different maps and a growing sense of dread. Kevin Kelly had it right: pursue direction, not destination. Unpublishable as self-help. Correct, though.

The people who actually ship things rarely announce they're about to ship things. They just go quiet for a while, and then one day there's a link. Respect the quiet ones.

We spent a decade flattening everything. Removing texture, depth, personality. Calling it "clean." Now the best work is quietly adding all of it back. The flat design era wasn't minimalism. It was austerity. We just gave it a nicer name.

There's a quality in the best software that's hard to name but easy to feel. The animation that holds for three milliseconds longer than it needs to because someone understood that pause would feel right. The button with a shadow that serves no functional purpose except making your brain trust it.

2026 is the best year to build with AI precisely because the keynote crowd has moved on to whatever's next. The tourist season is over. The people still here actually want to make things. Turns out the best filter for seriousness is just waiting eighteen months.

Naval's idea of specific knowledge is the most useful career concept nobody applies. What would you still be doing at 11pm on a Tuesday if nobody was paying or watching? Most people dodge the question. Self-knowledge is free, which is precisely why nobody values it.

A blank canvas, infinite budget, and no deadline produces a committee. A weekend, one problem, and a person who's annoyed enough produces a product. Most of the software I genuinely love was built out of spite. Underrated fuel source.

Every paradigm shift has the same three-act structure: "that's a toy," "that's dangerous," "we've always believed in this." We're somewhere between acts one and two. The LinkedIn thought pieces for act three are already being drafted.

OpenClaw hit 140k GitHub stars in weeks. Its creator joined OpenAI. Meta banned it. Security teams are writing incident reports. This is just what adoption looks like when the enterprise isn't invited to the launch party.

Every generation thinks it invented hustle culture and every generation's therapists patiently explain they did not. The Stoics were writing about burnout 2,000 years ago. They just had better prose and worse dentistry.

You can generate a passable UI in four minutes now. Which is exactly why the best designers I know have never been busier. AI didn't replace taste. It made taste the only thing left that's scarce. Congratulations to everyone with strong opinions about kerning.

OpenAI's CEO once called advertising "clearly dishonest." OpenAI now runs ads in ChatGPT. At some point you have to admire the commitment to proving your own thesis.

The most underrated professional skill is the ability to sit in a room where everyone is panicking and just... not. Not because you're calm. Because you've been wrong enough times to know that most emergencies aren't.