data driven orgs/people are interesting ,
corporate prefers them cause you're better off making the wrong decisions with the right data ,
plus the consequences for intuition/gut /trust are reckless if you're wrong and surprisingly ignored/reckless if you're right because it should still come from data.
Irish writing features prominently on bookshelves all over the world.
This year we celebrate Bloomsday & Beyond with booksellers who explain why writing from Ireland, the island of writers, resonates so strongly with their readers.
LLMs still aren't good enough to work autonomously on large codebases.
Statements like: "You shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents." feel more like marketing for the next evolution of coding than a reflection of reality.
I absolutely see value in autonomous loops for things like finding bugs introduced in the last 24 hours, reviewing code for security or performance issues, generating test cases, investigating flaky CI failures, stress testing features, etc
But these are supporting workflows. They are not the product.
The closer a task gets to core product logic, architecture, or long-term maintainability, the more important it is for a human to stay in the loop.
In my experience, LLMs perform best when making small, well scoped changes. Once you let them autonomously modify large parts of a codebase, the probability of introducing complexity, technical debt, and subtle bugs starts compounding quickly.
The people selling fully autonomous software engineering are usually the same people with effectively unlimited token budgets and very different incentives than the teams that have to maintain the code six months later.
Stay in the loop.
"Books are the closest thing you’ll ever come to finding cheat codes for real life. You can access the entire learnings of someone else's career in a few hours." —@tobi