daxfohl

32 ms·

421.
by daxfohl 1y ago
Nit, in Haskell it's a MonadPlus. Which IIRC is a monad that supports filtering.
422.
by daxfohl 1y ago
True. I didn't get to the fact that it was sales due to pay wall.
423.
by daxfohl 1y ago
I don't recall reading any official statements, ever, that said layoffs were due to AI replacing engineers. Do you have examples? I think that line gets thrown in by reporters that don't know what they're talking about, or H
424.
by daxfohl 1y ago
They're still hiring though. If layoffs are a prereq for hiring fresh blood, maybe it's worth the cost? Especially if hiring is weighted toward new grads and junior engineers, which are in the midst of a historically bad employmen
425.
by daxfohl 1y ago
Shifting focus is not narrowing focus. They should be able to shift employees accordingly too.
426.
by daxfohl 1y ago
Well, after some more thought, I realized that doesn't account for things that are truly innovative. I imagine humans will have a lock on that for a while. But it does raise the question, how much of our work is true innovation? When I
427.
by daxfohl 1y ago
Though in a way, this feels similar to things like garbage collection, disk defragmentation, or even query planning. Yes, you could build libraries that do these sorts of things for you, but in all likelihood the LLM providers will embed cu
428.
by daxfohl 1y ago
Or maybe it's that each of these things is pretty simple in itself. Clipping context is one line of code, summarizing could be a couple lines to have an LLM summarize it for you, etc. So not substantial enough for a formal library. Whe
429.
by daxfohl 1y ago
I'm surprised there isn't already an ecosystem of libraries that just do this. When building agents you either have to roll your own or copy an algorithm out of some article. I'd expect this to be a lot more plug and play, an
430.
by daxfohl 1y ago
Seems like there'd be an opportunity for open source tooling here. Context visualized, summarizers, explorers, A/B testers, etc. Also LLM pre-caching of context summaries since IIUC any context change requires full N^2 recalculati
431.
by daxfohl 1y ago
Cool, I'll give it a shot. So far I've just been blindly feeding into the AI and crossing my fingers. I'll try displaying the spectrogram graphically, and I imagine that'll help figure out what the next step needs to be.
432.
by daxfohl 1y ago
I was hoping to make a piano practice assistant for my kids, that would take sheet music in MusicXML format, listen to the microphone stream, and check for things they frequently miss like rests, dynamics, consistent tempos. Surprisingly th
433.
by daxfohl 1y ago
Finally learning TLA+ by plugging a very simplified multithreaded Java simulation of an old project's distributed, (hopefully) eventually-consistent algorithm into LLMs and asking for translations. I'd previously tried to learn TL
434.
by daxfohl 1y ago
Yeah, PM, data science, compliance, accounting...all largely automatable. You just need a few directors to call the shots on big risks. But even that goes away at some point because in a few months it'll have implemented everything you
435.
by daxfohl 1y ago
Yeah, I give it about two years until we get to "Hey AI, what should we do today?" "Hi, I've noticed an increase in users struggling with transactions across individual accounts that they own. It appears some aspect of m
436.
by daxfohl 1y ago
I have to imagine google considered this and decided against it. I assume it's that all the high-perf matmul stuff needs to be ASIC'd out to get max performance, quality heat dissipation, etc. And for anything reconfigurable, a CP
437.
by daxfohl 1y ago
Yes. I tried the prolog style requirements last night too and it wasn't great either. Some things are easier to say in "rules" and some things are easier to say in "steps". Trying to get all the way to "rules&q
438.
by daxfohl 1y ago
If you're a big software company, not much. If you're a small non-tech business, it could be an easy way to automate some things without hiring a software engineer.
439.
by daxfohl 1y ago
Though now that I say that, I wonder if maybe the approach should be, instead of specifying the onboarding procedure in terms of steps, maybe it would be better to specify a goal, the constraints, and some optimization function, and allow t
440.
by daxfohl 1y ago
I won't say "never", but I guess I left out the caveat that I tried this for a hack week demo once. Working in compliance and onboarding, there's a whole lot of logic around what you have to collect and check, what you h
441.
by daxfohl 1y ago
My hunch would still be no; human agents are able to cooperate without needing to do everything in a global shared record, so I'd expect AI agents would as well. If you (or any other AI agent) feel the need to check that the AI agent d
442.
by daxfohl 1y ago
This is somewhat my take too. The way most vibe coding happens right now does create a lot of duplication because it's cheap and easy for LLMs to do. But eventually as the things we do with coding assistants become more complex, they&#
443.
by daxfohl 1y ago
A big difference is that compilers are deterministic, and coders generally don't review and patch the generated assembly. There's little reason to expect that LLMs will ever function like that. It's always going to be a back-
444.
by daxfohl 1y ago
That's more of a hobbyist thing I'd say. Corporations developing these things will of course want to use some centralized system that they trust. It's more efficient, they have more control over it, it's easier for avera
445.
by daxfohl 1y ago
Nothing works automagically. You still have to build in all the operational characteristics that you would for any traditional system. It's deceptively easy to look at some AI agent demos and think "oh, I can replace my team'
446.
by daxfohl 1y ago
Makes sense, and each model has a max context length, so they could charge per token assuming full context by model if they wanted to assume worst case.
447.
by daxfohl 1y ago
Partially related, is charging by token sustainable for LLM shops? If the compute requirements go up quadratically, doesn't that mean cost should as well?
448.
by daxfohl 1y ago
Can you expand any more on the nagging feeling?
449.
by daxfohl 1y ago
Downdetector is end-user-reported, and the numbers look reasonable, so I'd say it's doing fine. I think more likely the cloudflare outage is causing failures that end users are reporting as outages in the service cloudflare is sup
450.
by daxfohl 1y ago
They should make the AI lead the postmortem.

More ›