warpfork

@warpfork.bsky.social

still working with computers. to my frustration. User choice and consent-based system design matters. Local-first software is always superior. Updates should never be forced. Reproducible, deterministic systems are the only ones that should be trusted.

everyone in tech is engaged in a zero-sum competition about who can have the most measured and sober-minded take about the implications of ai. it’s like a bunch of astronauts trying to out-calm each other after discovering the catastrophic and unrecoverable oxygen leak

Something that's exceptionally funny to me when attempting to bootstrap a lang+toolchain with the help of LLMs and using Rust as the host language for it is... The LLM's keep porting Fears from Rust into the new architecture. Eg, they're just deathly afraid linking will be slow. Memetic contagion!

Something I've been coming to terms with lately: sum types (aka variants, or several other names as well) mean for a somewhat substantially more complicated job for any kind of Garbage Collector. With variants, you can't just know where pointers are from static type trees; you need more branches.

Hearing that someone had to help their workplace migrate some projects from TS to Go for sheer compiler speed reasons. Think I'm making it a new life goal to make it true that "no one should ever have to migrate away from having sum types...for toolchain performance reasons".

the purpose of life is buying the same things at the späti every day for exactly 5€ so when they raise the price on pilsner urquell the späti guy tells you you get to keep paying 5€ instead of 5.20

one thing agents are good for is "hey agent what the FUCK is the FUCKING incantation to GET THE DAMN LOGS for this fucking systemd service again? you know, the thing that's a different command name, and needs a bunch of fucking flags in order to do the most obvious fucking thing?" stg, linux

I have to say, I find memory ownership systems really appealing... ...but then when I see the amount of Rust code that just sort of goes "ah, well, nevertheless,," and then seasons `.clone()` liberally over the codebase.... It does not inspire joy.

Evidence so far suggests that it’s easier to discover previously unknown proofs in math and make new scientific discoveries than it is to make UI tests in Xcode pass reliably

latest leet hax: tell your agent that's doing coordination for other subagents that they're like the frequent flier folk in the Speaker books of the Ender series. believe me, this will work and there are fewer more compressible ways to say it. darmok and jalad ride again

I know it’s not most folks‘ primary concern, but LLMs or not, I’m unimpressed by how soft these infrastructure services are. What do you mean HF had a Jinja2 template injection. And I’m still not over GitHub’s unsandboxed RCE. Geomys might need to self-host code/CI to avoid a weak link.

Having a Theory of Change that extends beyond our own code can be essential. on the one hand, leaning into it may weaken what we can do (in a limited, absolutist and instantaneous sense). on the other hand, it can sometimes be one of the only ways to push towards the best possible futures.

A code review can be an artifact itself. (A lot of the best code review and change management tools are already moving that way. Github is a noticeable aberration & it's the worse for it.) What if a code review results in... Just a file? Push it and share it wherever you want. Work offline. Win?

Rust is great, but as we move to an agent-driven future the need for tight write/compile/test loops becomes stronger, and Rust's deficiencies become more important. What *next* language can give 1) memory safety w/out GC, 2) strong generic + defensive programming facilities, 3) fast compilation?

Steve Klabnik@steveklabnik.com · 3w ago

Becoming radicalized against compile times, actually. Pre agents I don’t think it really affected my actual throughput and so I didn’t really mind compile times but now I can directly trace it to landing less features in the same time. Hurts way worse

It’s interesting seeing people get angry at Codeberg for banning predominantly LLM projects from being hosted there when they’ve clearly outlined the costs problem of hosting it 99.99% of these people would rather give 200$ to Anthropic than donate even half that to Codeberg for hosting their shit

One thing that owning one (1) Pantone Validated color monitor does to me... ... is make me very aware that none of my other monitors are Pantone Validated, or calibrated to... anything at all, really, apparently. 😬 Cannot Unsee.

It is extremely unfortunate that `cargo build` cannot be trusted to cachebust itself correctly. I just spent an hour chasing around what I thought were seemingly impossible bugs in low level graphics APIs before realizing that...no, the machine I couldn't seem to repair was just...not building...🫠

It took me prolonged close interaction with several people who had serious mental health issues before I truly realized this. There's some gradient of how much you should prethink wording if you want it to reach a wide audience, but it's healthy to know there is also such a thing as too much.

Olúfẹ́mi O. Táíwò@olufemiotaiwo.bsky.social · 3w ago

every time you write overly defensively you should remember that you are pursuing a strategy of writing careful sentences specifically for people who demonstrably either won't or can't read them

Security posture is not something where developer opinions should supercede user agency, and I'm tired of pretending otherwise! As a single application developer, you have absolutely no idea how relevant your attempt at security posture is to the context in which I'm applying your application!

broadly speaking, when I open an app it should be how I left it one of many great affordances that "webapps" destroyed, which we are only now starting to claw back