- The future advantage will come from a company owning the learning loop between its people, workflows, data, and AI systems - not the model it chose to use. Human judgment sets direction; AI captures, scales, and improves that judgment over time. The companies that win will turn their institutional knowledge into compounding intelligence they control, instead of letting it leak into generic models that make everyone’s expertise look the same. https://x.com/satyanadella/status/2066182223213293753
- Software now takes shape in the conversation, not the commit. Pull requests are limited because they attach conversation to code too late. The real collaboration often happens while the code is being written, not after it has been committed and pushed. As AI agents write more code, the conversation that led to the code becomes part of the source itself. The better model is to keep code, edits, and discussion connected as they evolve, so teammates and agents can understand the history and reasoning behind the work without waiting for a pull request. https://zed.dev/blog/introducing-deltadb
- Fast is better than slow because speed gives you more chances to learn. When you start quickly, share early, ask for help, avoid pointless debates, and do only what is required, you get feedback sooner and make better decisions. To move faster, don’t waste time doing things that nobody asked for. https://dubroy.com/blog/fast-is-better-than-slow
- Emotional intelligence is realising that most things are not about you. People are usually reacting to their own fears, wounds, and memories. When you take something personally, you are often reacting to yours. The goal is not to grow thicker skin. It is to see clearly. https://x.com/FU_joehudson/status/2067623632407531755
- Yak shaving (doing a chain of small, related tasks that slowly pull you away from your original goal) is fun because every detour teaches you something. You start with one simple task, run into problems, and end up learning how tools work, why systems are built the way they are, and how different pieces fit together. It can be wasteful when you just need to ship something, but for learning it is deeply satisfying: even if you do not finish the original task, you understand more than when you started. https://parksb.github.io/en/article/32.html
What I’ve been reading (W25, 2026)
Owning your learning loop, code in the conversation, shipping fast, and detours that teach