2026-04 - W4
A good AGENTS.md is a model upgrade. A bad one is worse than no docs at all. | Augment Code
Some tips to improve the quality of the AGENTS.md. I should seriously spend some time to improve those files in my projects…
GitHub - justrach/kuri: Browser automation and web crawling for AI agents. Zig-native, token-efficient CDP snapshots, HAR recording, and a standalone fetcher. · GitHub
Could be a good alternative to playwright.
Mike Acton’s Expectations of Professional Software Engineers - Adam Johnson
Really good expectations. Should be the industry standard, but I know humans are lazy, so people tend to be lax at some point, myself included.
Richard Hamming: You and Your Research
Really long essay, but good read and lots of insightful stuff (yes, I know, I lack vocabulary).
One Developer, Two Dozen Agents, Zero Alignment:
From a staff at GitHub (I knew her from her blog post to maintain her digital garden). She is a really good writer. This blog post to talk about agentic engineering, by having shared AI env with her teammates instead of having one env per engineer. Interesting approach. But the end of the day, it’s a promotion to their GitHub’s new product Ace.
most of the context you need for alignment isn’t in the codebase.
It’s in people’s heads. … Many people are now realising that in a world of fast, cheap software, quality becomes the new differentiator.
The bar is being set much higher Craftsmanship is what will set you apart from the vibe-coded slop.
Twenty Years of Stacking Commits — jd:/dev/blog: Another statement about code review being the bottleneck, and stack PRs not necessarily the answer.
Stacks aren’t free and they aren’t universal. Exploratory work that changes shape mid-stream wants a single branch, not seven. Tightly coupled changes where commit five only makes sense once you’ve read commit four don’t decompose neatly. Bug fixes are usually one PR. The skill is knowing which kind of work you’re doing before you start. If your team can’t write atomic commits or doesn’t trust interactive rebase, stacks make things worse, not better. The model demands discipline. The bet is that AI removes enough of the writing tax that the discipline becomes affordable.
Illegal vs Unwanted States • Buttondown
Interesting distinction about 2 states.
An illegal state is a state we never want our system to be in. An unwanted state is a state we don’t want to stay in. Many states that we wish were illegal are actually unwanted.
I benchmarked caveman against two words | Max Taylor
Caveman plugin finally not that great compared to 2 words “be brief”.
Boring code is an organizational tell — Simon Aronsson
Technical debt we are all familiar with. It’s a debt you can measure and pay down. The other two you find in the incident.
A clever culture that ships clever code slowly is recoverable. The same culture with agents is not - because intent debt doesn’t accumulate, it evaporates.
BEWARE SOFTWARE BRAIN | The Verge
Well written article. Interesting idea of “software brain”.
software brain, and it’s a particular way of seeing the world that fits everything into algorithms, databases and loops — software. … The entire human experience cannot be captured in a database. That’s the limit of software brain. That’s why people hate AI. It flattens them. … it is a failure when you ask people to adapt to computers. Computers should adapt to people. Asking people to make themselves more legible to software — to turn themselves into a database — is a doomed idea.
Laws of UX
Interesting laws. Could be interesting to put them into skills.
Side note: if you’re asking your LLM to write a skill, after it does first draft, prompt it with something like “what are the goals you want to achieve? does it achieve these goals?” It will make suggestions, ask it to apply these corrections, and repeat the process 3~5 times. This process of revision-through-criticism works for blog posts, documentation, code, and more.
A Couple Million Lines of Haskell: Production Engineering at Mercury | The Haskell Programming Language’s blog
Very very good and pragmatic feedback over the usage of Haskell in a fintech.