On this page
Until recently, the hard part of building something was the building. You had an idea, and between the idea and the finished thing sat weeks of work.
That gap is closing faster than I ever expected. AI has been on my mind a lot lately, mostly with awe and sometimes with something closer to dread, so I wanted to write down where I am with it. Which, honestly, isn’t very far.
The problem is the hard part now
AI is an incredible tool. I can describe almost anything I can imagine and have a working version in front of me in a fraction of the time it used to take. Execution is no longer the struggle it was. For anything made of code, building is close to a solved problem, as long as you clearly understand what you’re trying to do.
That condition matters more than it sounds. When making things gets cheap, the hard part moves upstream: knowing which problem is worth solving, and understanding it well enough to explain it. These days I find it harder to find a good problem than to solve one once I have it.
You can see the shift from the outside, too. In early 2025, Pieter Levels, who had never made a game, put a flight simulator online after a few hours of describing it to an AI code editor. A few weeks later, ads in it were bringing in around $50,000 a month.
The evidence is messier than the feeling, to be fair. In a 2025 study by METR, experienced developers working on large codebases they knew well were 19% slower with AI tools, even though they believed they had been faster. I don’t think that contradicts anything. AI is at its most magical on things that don’t exist yet. In a big, old codebase, knowing what to change is still the hard part, which is the same point from the other side.
Who pays for all of this?
At the same time, I don’t understand how the cost of AI is sustainable.
Anthropic is preparing to go public, and its prospectus, reported this month, shows a net loss of $80.7 billion for the first half of 2026. That number is less frightening than it looks. About $72 billion of it was an accounting charge tied to bonds that can turn into Anthropic shares, not cash spent on running the company, and Anthropic has told investors it made a profit on an adjusted basis last quarter. But the real spending is enormous all the same. It put close to $9 billion into computing power and infrastructure in those six months alone, and it has committed to spending at least $518 billion over the next ten years. OpenAI reportedly burned through $3.7 billion in cash in the first three months of the year.
So far, almost none of that cost has reached us. ChatGPT Plus has cost $20 a month since it launched in 2023, and Claude Pro costs the same. The pressure shows up in other places instead: $200-a-month plans for heavier use, and, since February, test ads in ChatGPT’s free and cheapest plans.
How does this scale? I honestly have no idea. What worries me is the day prices have to rise to meet the costs. A tool that makes building almost effortless for the people who can pay would be out of reach for far too many who can’t, and that would be one more gap the world didn’t need.
Will the world lend a hand?
And then there’s the occasional existential crisis. If AI can do this much of the work, what happens to the value of the people who used to do it?
I’m clearly not alone in asking. In a Pew Research Center survey this summer, 52% of Americans said they were more concerned than excited about AI in daily life, up from 37% in 2021. Among adults under 30 it was 55%, the first time a majority of them have said so, and their concern is still climbing.
Younger people have reason to worry. Stanford researchers tracking payroll data found that employment for 22- to 25-year-olds in the jobs most exposed to AI, like software development and customer service, now sits 19% below where it would be if it had kept pace with their peers in less-exposed jobs, and the gap has kept widening since they first measured it in 2025. Experienced workers in the same jobs show no such gap, and across the wider economy the researchers see no sign of mass job losses. Most of it comes from fewer young people being hired, not from people being let go. The door isn’t slamming shut on anyone. It’s quietly not opening for the people at the start of the line.
I don’t think efficiency is the villain here. What scares me is who it ends up serving. The world can be cruel, a little dog-eat-dog, and it doesn’t always share what it gains. Will it really extend a helping hand to the people this leaves behind, or will many a man be left devoured by this fire humanity started?
For now
I don’t have answers yet. I’m amazed and uneasy at the same time, and I suspect I’m not the only one. I’ll keep building with these tools, because they really are wonderful. I’ll also keep asking who gets left behind.
If you’ve been thinking about this too, I’d love to hear where you’ve landed.