miniblog.

I periodically see "diffstatic" as a typo of difftastic. I quite like it: makes me think of 'ecstatic' rather than 'fantastic'. I've eventually configured typos CLI to prevent this spelling on the difftastic repo, but it makes me smile every time.

It's really satisfying fixing crashes in rust-analyzer. You can start from a panic message and dig until you've got a tiny Rust program that triggers the problem. (It's usually an issue with incomplete source files breaking invariants.)

Semver has a really odd relationship with software maturity. I'd take a v0.21 project over a v1.0 project.

I asked Hermes Agent to change how it set thread titles, and it patched its own source code! Self modifying code is hard to reason about even in a deterministic environment. I'd already limited Hermes access to the todoist API and web search out of an abundance of caution.

I'm trying mini projects to exercise my new programming language and find gaps. So far I've done fizzbuzz, JSON parsing, and a crude scheme interpreter. Any ideas for additional projects? They need to be well scoped because I keep finding e.g. stdlib crashes.

Language design question: what type should `while True {}` have? Bottom: It doesn't contain any `break` statements. Unit: Treat loops with/without `break` consistently. I already have lints that warn on unreachable code after diverging loops, so the Unit case feels surprising.

I do occasionally find myself wanting to use the phrase "you're absolutely right", especially in technical discussions. I've started avoiding it though. I worry it'd make me sound like I'm just automating my comments.

LLMs have a really unfortunate habit of leaving comments that exactly describe the bug they just fixed. I keep getting code like this: // Let Alice and Bob log in. if user_tags is None:

I've only got a few VPS instances, but setting hostnames automatically to match Ansible inventory makes my life so much easier. I should have done this way sooner. It's so much nicer than seeing wilfred@ubuntu on every box I ssh into.

Playing with Hermes Agent and it's the first agentic tool where I feel I need to *justify* my changes. It's updating its notes over time. W: Split those two jobs, they're completely unrelated. H: Done. I also updated the skill to note that unrelated topics should be split.

I tried OpenClaw today and I was surprised by the short time horizon of conversations. Am I holding it wrong? W: Summarise open issues on my GH repo. O: (summary) W: OK, remember how I want you to format content [..]. O: Noted. W: Summarise the issues again. O: What issues?

I like the idea of tools like Pinboard or Karakeep that let you save copies of websites you've bookmarked. I'm considering just writing a script that saves web pages as a .md in a private git repository though. It feels crude. I think I get essentially all the benefits though?

I'm looking at adding bitwise AND and OR to my programming language. Do I add & and | operators, methods x.and(y) or a library bitwise::and(x, y)? My inclination is to add infix operators (concise and familiar) but syntax is such precious real estate.

TIL that LSP 3.18 was released this month! Several APIs now support richer content, such as tooltips or docs on code action kinds. Signature help continues to improve too, allowing you see more about a function at its call site.

I'm trying to teach LLMs to yak shave responsibly. It's normal to be programming and notice other clean ups that really ought to be done. I don't want it done in that PR though. Currently experimenting with asking the AI to just create hand-off tasks in the issue tracker.