We Tried Trunk-Based Development... The Results Were Shocking. https://t.co/HfY8vKClCw via @davefarley77 cc @paul_hammant

Fifth time in 24 years revisiting the same pseudo-declarative markup idea: https://t.co/Zts3Qv2o7x - this time with Claude's help on top of @Fyne_io (thx to @andydotxyz and team for excellent foundations.

he he

Then, it's writing a python script to refactor a JS source.

Ugh then Claude tripped up again as I was in that dreamspace.. "The indentation got messy. Let me take a cleaner approach - I'll restore the original file and do a proper Kent Beck refactoring step by step" .. it's now counting lines in a JS source.

And even then it feels too declarative for me

https://t.co/sOb9RPqKt8

https://t.co/BbaoR3XIPq

I also wish for all JetBrains IDEs to have Intellij's refactoring rather than just a subset, and for there to be a MCP that interops. An intermediate language. Call that "Refactoring Intent Grammar"

"Let me take a different approach. I'll create a clean version by reading the current malformed section and checking where it ends" .. I wish I could see the exact steps on sources that Claude does for insights as to what leads to mangled sections.

https://t.co/11YCvXPCUf Increments and iterations, 1993 version Thought I'd drop you this little-known article I wrote in 1993, which is going in the book. - Incremental means adding; - Iterative means revising.

Wish me luck applying "AI factory" for what we (@med_bravo and me) are devoting our time: design new drugs to help cure diseases. I hope to be proven wrong, but I believe for one bureaucratic reason or the other we will not qualify and this process will have such friction that

S/w development industry definitely needs an Infrastructure as Code AS CODE initiative.

Will 2026 be the year of netsec professionals suggesting air gaps for some duties/workloads?

Damn you HP Chromebook and your phantom touchpad presses cos of your plastic chassis. Someone pay KB $10m a year to train LLMs on his own. Someone else sit beside him and ensure that other-AI transcription is accurate.

A casual understanding of who Kent Beck is and what he stands for - neat. Someone pa

OpenAI's "Codex" cli:

Of course ClaudeCode can gaslight you a little bit: tests do pass (they don't really), I didn't bake mocking logic into the prodcode (yes you did), and more

ClaudeCode comments:

I can't reliably stop it from implementing "fallback" paths in code. Like that is asked for, and is banned in ` https://t.co/PnFVhAzymc` but it still happens frequently enough