winwang

Really interesting. I was thinking about something similar regarding the shape of code. I have no qualms recommending my agents take static analysis to the extreme, though it would cumbersome for most people.

What about someone inexperienced but skeptical, using AI to learn + fix their own code before opening the PR?

Linear walkthrough: I ask my agents to give me a numbered tree. Controlling tree size specifies granularity. Numbering means it's simple to refer to points for discussion.

Other things that I feel are useful:

- Very strict typing/static analysis

- Denying tool usage with a hook telling the agent why+what they should do (instead of simple denial, or dangerously accepting everything)

- Using different models for code review

If you have enough cores, you could pool the L1 together for makeshift RAM!

Kind of agreed. I like vibe coding as "just" another tool. It's nice to review code in IDE (well, VSCode), make changes without fully refactoring, and have the AI "autocomplete". Interesting, sometimes way faster + easier to refactor by hand because of IDE tooling.

The ways that agents actually make me "faster" are typically: 1. more fun to slog through tedious/annoying parts 2. fast code review iterations 3. parallel agents

Not sure, but https://status.claude.com/ uptime is pretty spotty. Funnily, the latest bar is still green despite there being an incident (and the messages even acknowledge this)

Interesting! I actually split up larger goals into two plan files: one detailed plan for design, and one "exec plan" which is effectively a build graph but the nodes are individual agents and what they should do. I throw the two-plan-file thing into a protocol md file along with a code/review loop.

Awesome stuff. I have a 'root' cli that i namespace stuff into so to remove the need to pass around paths, e.g: `./cli <cmd> ...`

Fun fact, it can miss! I've seen it miss almost half my messages, including some which were actually important, haha.

For manual prompting, I use a "macro"-like system where I can just add `[@mymacro]` in the prompt itself and Claude will know to `./lookup.sh mymacro` to load its definition. Can easily chain multiple together. `[@code-review:3][@pycode]` -> 3x parallel code review, initialize subagents with python-code-guide.md or something. ...Also wrote a parser so it gets reminded by additionalContext in hooks.

Interestingly, I've seen Claude do `./lookup.sh relevant-macro` without any prompting by me. Probably due it being mentioned in the compaction summary.

Apparently LLM quality is sensitive to emotional stimuli?

"Large Language Models Understand and Can be Enhanced by Emotional Stimuli": https://arxiv.org/abs/2307.11760

Is it the performance benefits? Or being able to write concurrent code much more expressively? Though I suppose the latter might imply the former.

Forgot exactly when, but at least ~4 days ago. Or maybe I'm tweaking...

In any case, this is likely the next computing revolution.

(in addition to other replies) I believe there was a study on brainrot where, acrosss a few different platforms, TikTok was significantly worse than e.g. Youtube. (sry, on mobile or I'd ref. hopefully later...)

Dang... it's back...

Honestly, I was hoping to take a nap sometime soon, lmao! As good a time as ever (was working on some claude code scaffolding anyway.)

I think it depends on what you mean by "is a function". You can think of a constant, `x` as `x_: () -> {x}` (i.e. everything can be indirected). It could be argued that this is "philosophically" "useful" since getting (using) the value of `x`, even as an actual constant, requires at the least loading it as an immediate into the ALU (or whatever execution unit).

Even non-functional relations can be turned into functions (domain has to change). Like a circle, which is not a function of the x-axis, can be parameterized by an angle theta (... `0 <= theta < 2pi`)

I thought I was going to see something crazy like using RT cores in parallel with tensor cores. Like compiling matmul into triangle intersections.

I'm not too caught up with politics, but a (presumably) political shooting has the issue of being disruptive to the government and therefore the nation as a whole, since the USA is built on democratic ideals. And since it's a(/the) global superpower, its issues result in serious international problems as well.

Just want to agree with everyone who is thanking you for owning up (and so quickly). Got phished once while drunk in college (a long time ago), could have been anyone. NPM being slowish to get back to you is a bit surprising, though. Seems like that would only make attacks more lucrative.

I hated precalc, but I loved proof-based linear algebra, my first university math course (linalg + multivariate calc).

The simplicity(/beauty) of matrix multiplication still irks me though, in the sense of "wow, seriously? when you work it out, it really looks that simple?"

This is an interesting take. I have a somewhat orthogonal viewpoint -- rather than "heavy cognitive load", I think that going somewhat off-mainstream is good for attracting, on average, better devs. For example, it's likely that the average Haskell dev spends more time honing their craft than the average Java dev. The article kind of touches on this (e.g. FP "vs" the more popular OOP) with familiarity vs simplicity though.

Ironically, the attempt to prevent illegal states may create "complex" code (quoted since it may be due to perceived or actual complexity).

Does `scalafmt` really slow down builds "considerably"? I find that difficult to believe, relative to compile time.

That's a fantastic design idea, and it seems to require all the other lines to have the same indentation "prefix".

Haven't used much C#, but I love Scala's `.stripPrefix` and `StringContext`.

If these are the problems (or, your problems), then it seems that it doesn't matter if AI wrote it or not -- just that the writing isn't "good".

I believe there is at least one other thing I got from the post: that he shouldn't have to abandon Scala, perhaps because doing so is to give in to a sort of injustice (in his mind)?

You don't know someone until you know someone.

I find it ironic that you talk about distrust but then use ACAB, which I assume means "all cops are bad" (cursory googling).