- Wrote about the future I’m thinking about and leaning more into day-by-day: self-driving software through gradual agent integration and adoption https://twkrr.com/2026/01/24/path-to-self-driving-software/
- Understanding something new means changing how you already think. But first you have to notice your current pattern, like fixing a habit. You can’t change how you tie your shoes until you realize how you’ve been doing it all along https://x.com/phokarlsson/status/2012112342239379676
- The algorithmic world doesn’t just commodify interaction, it standardises imagination. Creators optimize for algorithmic performance, producing functionally identical outputs. Nonconformity now carries financial costs most can’t afford https://om.co/2026/01/16/our-algorithmic-grey-beige-world
- Ghostty’s creator banned AI-generated code from outside contributors after low-effort PRs flooded the project. His clarification: “This is not an anti-AI stance. This is an anti-idiot stance.” Only maintainers can use AI for approved work https://github.com/ghostty-org/ghostty/pull/10412
- Your code is a strategy memo to strangers maintaining it at 2am during an outage. Optimise for their comprehension, not your elegance https://addyosmani.com/blog/21-lessons/
- Stateful sandboxing for AI agents is coming fast. Fly.io’s Sprites use object storage as the durable layer with local NVMe as cache, enabling rapid checkpoint/restore, critical for agent state management. https://fly.io/blog/design-and-implementation/
What I’ve been reading (W4, 2026)
Self-driving software paths, the cost of algorithmic sameness, and why clarity beats cleverness