daxfohl

23 ms·

451.
by daxfohl 1y ago
Oddly i'm not having any issues with anything I regularly use.
452.
Everything Is Down
75 points by daxfohl 1y ago|46 comments
453.
by daxfohl 1y ago
Maybe these features already exist, but I'd like a way to: 1) Type check without necessarily building the whole thing. 2) Run a unit test, only building the dependencies of that test. Do these exist or are they remotely feasible?
454.
by daxfohl 1y ago
Good point. One could say a doctor is the same job as a mechanic, but that doesn't capture the whole story.
455.
by daxfohl 1y ago
> although we’ll make plenty of mistakes and some things will go really wrong, we will learn and adapt quickly If the "mistake" is that of concentrating too much power in too few hands, there's no recovery. Those with the
456.
by daxfohl 1y ago
> although we’ll make plenty of mistakes and some things will go really wrong, we will learn and adapt quickly Famous last words.
457.
by daxfohl 1y ago
Though put that way, it seems hard to rationalize high salaries for software roles where this tax deduction would apply. Granted, supply-and-demand, but still.
458.
by daxfohl 1y ago
What about oncall? What about fixing bugs, or KLO, or security patches, or devops, or tweaking feature flags, or dealing with customers? If you're 100% allocated to a greenfield project that's behind closed doors until 2027, sure.
459.
by daxfohl 1y ago
FOMO. They don't want to risk being the one company left behind because their engineers haven't learned to use AI as efficiently as others.
460.
by daxfohl 1y ago
The other side of me thinks that maybe the eventual landing point of all this is a merger of engineering and PM. A sizeable chunk of engineering work isn't really anything new. CRUD, jobs, events, caching, synchronization, optimizing f
461.
by daxfohl 1y ago
It's like trying to learn math by reading vs by doing. If all you're doing is reading, it robs you of the depth of understanding you'd gain by solving things yourself. Going down wrong paths, backtracking, finally having that
462.
by daxfohl 1y ago
I went through the same phase a couple decades ago and also came out with Middlemarch on top. Oddly I'd never heard of it prior, and only grabbed it because it was in some classics list at B&N. It didn't seem like anything I e
463.
by daxfohl 1y ago
That's true for now, and of course will be possible in the future too, but I think there's a potential future where AIs just own the whole stack: client, server, some inscrutable protocol between them, some ML model on both ends
464.
by daxfohl 1y ago
In a couple years when everyone is just manipulating AIs to make huge blobs of who-knows-what-but-it-works code, we'll hark back to how simple hand-coded React was.
465.
by daxfohl 1y ago
I'd read the first as adjusting to market demand, not playing with people's lives. If if were construed as playing with lives, that could apply to basically any investment.
466.
by daxfohl 1y ago
Agree, but two questions: First, is AI really a better scapegoat? "Reducing headcount due to end of ZIRP" maybe doesn't sound great, but "replacing employees with AI" sounds a whole lot worse from a PR perspective (
467.
by daxfohl 1y ago
It's already started, with AI resume spamming on one side and the associated AI assisted screening on the other. "Deep research" generated funding requests on one side and associated AI generated funding request summarization
468.
by daxfohl 1y ago
Yeah, I immediately thought of "The Wrong Abstraction" by the same author. Putting the branch inside the for loop is an abstraction, saying "the for loop is the rule, and the branch is the behavior". But very often, some
469.
by daxfohl 1y ago
IME this is frequently a local optimum though. "Local" meaning, until some requirement changes or edge case is discovered, where some branching needs to happen outside of the loop. Then, if you've got branching both inside an
470.
by daxfohl 1y ago
I like this a lot. At first, putting ifs inside the fors makes things more concise. But it seems like there's always an edge case or requirement change that eventually requires an if outside the for too. Now you've got ifs on both
471.
by daxfohl 1y ago
Oh, so I think what it does is take a nearly infinite-dimensional nonlinear space, and transform it into "the N dimensional linear space that best preserves approximations of linear combinations of elements". That way, any two (or
472.
by daxfohl 1y ago
I wonder if this could be used to find redundant code
473.
by daxfohl 1y ago
Wait, but if gender was composed of say two dimensions, then there'd be no way to distinguish between "the gender is different" and "the components represented by each of those dimensions are individually different"
474.
by daxfohl 1y ago
Or it could make abstractions even more important and useful. If it's able to identify common but nuanced patterns across the entirety of github, it could build libraries and frameworks that satisfy them. Then it would fine tune itself
475.
by daxfohl 1y ago
Mostly agree with these points, but it does say that it automates some of the fixes in complex cases that the non-AI tool couldn't handle. Besides, gotta start somewhere. It's probably a PoC, for a platform that will eventually ha
476.
by daxfohl 1y ago
> Proving things about low-level algorithms [...] more difficult than proving things about high-level domain logic I'm not sure I buy that. Like, prove that you never double-charge a sale, or a deactivated user can never make a purc
477.
by daxfohl 1y ago
Yeah, the issue is that proofs are harder than people think, even for trivial things (try a few easy leetcode problems in Lean with a proof of correctness), and less useful than people think, especially when you get past low level algorithm
478.
by daxfohl 1y ago
And it is continually improving. Rust borrow checker is an example of this. But as far as jumping into the deep end of dependent types, that's a whole other ball of wax. Like, imagine (or try!) writing leetcode solutions in Lean or Idr
479.
by daxfohl 1y ago
I also wonder how hard it would be to create your own agent that remembers your preferences and other stuff that you can make sure stays in the LLM context. ...Maybe a good first LLM assisted project.
480.
by daxfohl 1y ago
So is the real engineering work in the agents rather than in the LLM itself then? Or do they have to be paired together correctly? How do you go about choosing an LLM/agent pair efficiently?

More ›