// dev.blog  —  jltdpxl
user@jltdpxl:~$ ls -la ./posts

The Permutation Engine: On What AI Actually Does When It Solves Hard Problems

The critique that LLMs are derivative is correct. It's also incomplete. The permutation space of human knowledge is enormous, and humans have barely scratched it. AI doesn't think novel thoughts — it runs more paths in less time using the tools we already built. That's not nothing. But it is backward-looking. And that distinction matters.

cat post >>

The Token Ledger: Why Efficiency Is the Next Engineering Metric

Tokens used to feel free. They aren't anymore. On token maxing, the coming stratification of AI tooling by job function, and why the engineer who can do in five thousand dollars what their peer does in ten is demonstrating something real: not frugality, but judgment.

cat post >>

The Jevons Moment: Why AI Making Software Cheaper Will Make It More Valuable

Cloudflare had their best quarter ever and laid off 1,100 people. IKEA retrained their support staff as interior designers and opened a billion-dollar business line. One is a number; the other is a strategy. On Jevons' Paradox, forward deploy engineers, and why software demand is about to get larger than it's ever been.

cat post >>