Artificial Human Errors
AI makes human-like errors, but it cannot carry responsibility. Treating it as a replacement for people removes the judgment needed when those errors compound.
Writing on product, technology, and building companies.
AI makes human-like errors, but it cannot carry responsibility. Treating it as a replacement for people removes the judgment needed when those errors compound.
I am a worshipper of great Machines. I am a tentacle who stopped caring long ago.
The AI companies buy today is subsidised and still competing for them. Use that window to learn faster and build responsibly: fewer bets, measured adoption, portable architecture, human judgment, and capability that survives the real price.
I tell companies to build a middle layer that routes intelligence to humans on the surface. I've been running that pattern on my own company for a long time. Three moments where my own practice tested what I'd claimed.
Companies asking 'how can we improve with AI' are really asking 'how can we improve.' The vendors selling context-ingestion are answering the wrong question because it's the easiest one to sell.
The lie AI is trying to sell is that it just works. The reality is that every AI feature that delivers has hours of human translation behind it, and the bottleneck isn't the model, it's whether the humans around it can tell when it's lying.
The middle of every organization is dashboards, reports, CRMs, and status meetings. That layer is where AI belongs. Humans stay on the surface, making decisions with context they were never given before.
2025 was the year of AI panic. Companies fired people and bought tools they didn't understand. The bubble will burst. The companies that survive are the ones using this window to rethink how they work, not just what they buy.
Companies will increasingly build their own tools instead of buying SaaS. The Internal Product Developer is the new role that makes this happen.
I redesigned my personal website to meet modern web standards. It now exploits your insecurities.
Building features feels like progress. It's not. And AI is making the problem worse.
Companies give everyone AI tools and hope for the best. The result is chaos. Here's what's actually missing.
Your first AI feature should be a button, not a chat interface. Here's why most companies get this wrong.