LLMs Didn't Change What's Hard About Building Software. They Changed What's Easy.
The bottleneck moved. The calendar hasn't.
The bottleneck moved. The calendar hasn't.
Altman and Amodei walked back their AI-jobs-apocalypse predictions the same week they filed for IPOs. The reversal is a masterclass in platform company narrative management. Here is what founders building on their infrastructure should actually do with it.
AI
Saying your product is 'AI-powered' in 2026 is like saying it runs on electricity. It is a prerequisite, not a pitch.
AI
The models are commoditizing. The moat was always in your data. Most AI strategies have not figured that out yet.
AI
The AI agent demos are excellent. The production reliability numbers are not. That gap is the whole story.
AI
Models can now hold your entire codebase in one shot. The hard part was never storage.
AI
Every SaaS company has an AI features page now. Nobody's buying because of it.
AI
The copilot metaphor made sense for a minute. Now it is a branding workaround for products that cannot commit to what they actually do.
AI
Twelve AI labs. $29B raised. $130B valuation. Zero products. History says this ends badly for the labs and really well for the builders on top of them.
When your core tool is approaching a trillion-dollar valuation, you are not building on software anymore. You are building on a utility. And utilities price accordingly.
Compute is the bottleneck everyone is obsessing over right now. The smarter bet is that it won't be for long. Here's why AI infrastructure is about to follow the same arc as electricity.
Anthropic's Q1 growth came in at 80x forecast. Eight times what they planned for. That's not just a capacity problem. It's a signal about how fast AI adoption is outpacing every model founders use to plan.