chamomeal

Anecdotally, GPT-3 was super good at creative writing. It didn’t have any of the typical LLM giveaways. It would write super weird, interesting stuff. Especially if fine-tuned on a specific author. Of course it would occasionally descend into saying the same thing over and over. But IMO none of the current models come close!

biff.graph is basically a lightweight version of Pathom. It implements only a subset of Pathom's functionality with the intention of being easier to understand.

I’ve heard of pathom but I’ve never actually dove in and tried it out. It sounds super neato though.

I have a bunch of microservice DB’s (that should really just be on DB, but I think we’re created in the peak of microservices hype). I def need a better way to explore the data. “Easier to understand” sounds sick and I think I’ll check it out this week!

I’ve had people tell me how “fun” it was to build a macro to handle some one-off text-refactoring problem. But when I looked at what they were doing and how long it took, my honest reaction was: I could have done that in Sublime in a minute with multiple cursors, or just written a quick script

I totally agree with the larger point, but there are things you can do with vim macros that are just an absolute PITA to do with the built-in tools in vscode. Or maybe there is a specific tool that can compete (or beat) a specific use case of a vim macro, but macros are a single tool that covers a zillion use cases. So for this specific example I think there’s a tangible difference in capabilities.

Also 99.9% of the time-saving macros that people write on a day to day basis are not being shared with a single other person. It’s just a tool that becomes invisible to people who are comfortable with it. I’d argue that modal editors are particularly good at getting out of your way! Particularly ones with little or no config, like helix (or even vim mode in an IDE)

CL may not but other lisps do. Clojure’s concurrency is just so hawt

I am also pretty stoked about the doom performance improvements lol

Yeah that was the only really surprising part to me. So every time copilot breaks my source code to “fix” its crappy unit tests, does it know what it’s doing?

The whole redfall thing was tragic. It was crazy watching Arkane become a laughing stock. Some of the most talented game devs ever. Prey, dishonored

That is so wicked cool, I’m going to try this later today!!

Sounds like a super cool project. Gonna post the design anywhere?

That happened to me to! A similar moment of weakness: I got tired of being locked out of like half of small business profiles. A local brewery (very small brewery, but top tier beers) posts on instagram if they’ll even be open that day.

Caved, tried to sign up, asked for my face, then rejected me forever.

I mean it’s still worth doing, even if AI can do it. But I definitely empathize with that bit of AI ennui.

Greenville, South Carolina is also totally covered in trees. I think they have a bunch of laws that new developments have to plant trees, cutting a tree down requires planting multiple others nearby. The whole city is just covered in trees. Even in the suburbs nearby. It’s awesome.

MD has some replace-felled-trees law, but it’s kinda crappy cause trees can be planted somewhere else entirely. So a big development can be treeless.

I just check the git diff after claude code writes stuff. Stage things before letting it run wild so I can undo whatevs.

So many of the top comments look like parodies of HN comments

Yeah I’ve been thinking that LLM unit tests are basically snapshot tests. Just sorta ossify things in place. If they break, you just ¯\_(ツ)_/¯ and have the LLM fix them. It’s like they were never there!

So it’s just like the olden days of everybody ignoring tests, but we give anthropic a ton of cash

Anecdotally, coworkers are writing a LOT more unit tests. By which I mean NOBODY is writing unit tests, but Claude is generating a ton of em. We’re talkin 300 lines of unit tests for 20 line changes that are already covered by other, better kinds of tests. Huge JSON objects of test data that we already have generators for, etc

Not sure if OP is actually in here, but anybody interested in this topic should absolutely watch this talk by Matt Might:

https://www.janestreet.com/tech-talks/algorithm-for-precisio...

It’s a fascinating story similar to this one, but using “old” AI. The dude spent years learning genetics, and built a program called “medikanren” with Will Byrd. It’s p much a huge graph database of facts they extracted from academic papers, and query using logic programming. It’s a great story

Super cool but I did get this error while scrolling the timeline on safari/iOS

Application error: a client-side exception has occurred while loading jivx.com (see the browser console for more information).

That is an absolutely crazy story, I hope you have it written down somewhere besides HN comments lol

I think the hack bypassed 2FA. If you can call “asking for account access” a hack lmao

Ok dang. Well. I guess I have no reason not to dive fully into the elixir now.

I’ve toyed around with it a handful of times and I really like it. I like the clojure-ey immutability and threading operators and such. And of course I’ve heard so much about the magic of the BEAM and the phoenix framework. But between typescript and clojure I’ve never felt like I needed anything else.

But if the type system is pretty good, that’s a huge plus over clojure in my book.

That’s a good point! And I suppose you don’t have willy-nilly reassignment of properties like you do in ruby/php/js.

It also takes a long time for the ecosystem to catch up. It can be hard to retrofit static types over something that wasn’t built with them in mind

Websites are broken by default. They used to be functional, fast, and accessible but ugly. Now they're slop, agentic, and on fire — but they get attention, and attention is the only metric left. Nobody's reading and you know it.

I’m upset if an LLM actually wrote this because this is p sick

Are JVM virtual threads not on par with golangs's concurrency? I think core.async even uses virtual threads now

Not the person you're replying to, but have you tried TypedClojure? I've always thought clojure-with-types would literally be the perfect language, but I also read TypedClojure is more of a research project than a real language that you should use in prod.

Raw models aren't as effective with clojure as they are with typescript or python, but clojure has a superpower that most other languages don't have: the REPL! Specifically nREPL and the ecosystem around it.

An LLM is only as good as its feedback loop. If your LLM can actually test the code it writes, it's going to be much more effective. Static types are a form of feedback (if it can use the LSP), unit/integration tests are another.

Clojure has an exceptionally good repl. LLMs can eval any piece of any function. They can test out functions they aren't familiar with. They can fetch data, try out different arguments, try different approaches before committing to one. They can query a database (read-only connection, of course), look at the result, fetch data from an API, and stitch it all together. It can even hook into your running program and debug it from the inside out!!

It makes it so much more effective at using libraries or paradigms that it isn't trained on. In my experience, hooking an LLM up to the clojure repl lets it write WAY more complex stuff. I'm talking like 10x more complex programs with zero errors, cause it can literally try it out every little piece before putting it together. It's like watching a human programming. But like, really fast.

Sorry I get a little ranty when clojure + LLMs come up, because I don't think most people realize what they're missing out on. It's crazy stuff. It's also easy peasy if you use vscode. There's an extension called calva-backseat-driver that just hooks it all up for you. Gives copilot access to the repl, and I think it exposes an mcp if you want to give claude access too.

Aw man I love babashka. I will say the lack of static types in clojure is pretty brutal for me. Especially when combined with the obtuse error messages. But I still love babashka and the whole REPL driven world.

What did you end up rewriting your bb scripts in?

Is fennel clojure-ish? I’m vaguely aware of it being a lisp, but I didn’t know it was clojure-ey. Pretty neat sounding!

It is absolutely, positively mind boggling that you have to sit through those animations. And key presses don’t even take effect if your new desktop until the animation is done. It’s just lunacy.

How does a company with infinite resources and talented designers come up with shit like that??