daxfohl

11 ms·

151.
by daxfohl 9mo ago
A million times, this. Sometimes they luck into the intent, but much more frequently they end up in a ball of mud that just happens to pass the tests. "8 unit tests? Great, I'll code up 8 branches so all your tests pass!" Of
152.
by daxfohl 9mo ago
So we've graduated from unmaintainable slop code to unusable slop products. Sorry, this just doesn't feel like progress toward any meaningful future. But I'm sure it will unburden lots of investors of their money.
153.
by daxfohl 9mo ago
Often when doing that, you look up the company on blind and it's got a 3 star rating with low pay, bad management, and no vision, though decent WLB. My theory is that a lot of companies are one trick ponies. By the time you've hea
154.
by daxfohl 9mo ago
They already do. What they suck at is common sense. Unfortunately good software requires both.
155.
by daxfohl 9mo ago
Not just the abstraction ladder though. Also the situational awareness ladder, the functionality ladder, and most importantly the trust ladder. I can kind of trust the thing to make code changes because the task is fairly well-defined, and
156.
by daxfohl 9mo ago
Actually they're the only ones who do: copy and paste into chatgpt with "distill this please".
157.
by daxfohl 9mo ago
I don't even think that "singularity-level coding agents" get us there. A big part of engineering is working with PMs, working with management, working across teams, working with users, to help distill their disparate wants a
158.
by daxfohl 9mo ago
And while it appears that lots of problems can be contorted into translation, "if all you have is a hammer, everything looks like a nail". Maybe we do hit a brick wall unless we can come up with a model that more closely aligns wi
159.
by daxfohl 9mo ago
Interesting, IIUC the transformer architecture / attention mechanism were initially designed for use in the language translation domain. Maybe after peeling back a few layers, that's still all they're really doing.
160.
by daxfohl 9mo ago
They're currently still at the "create a million branches until the tests pass" phase. Once they are more capable of genuine design, refactoring, and maintenance, their code will arguably be more readable than code written by
161.
by daxfohl 9mo ago
Damn, I'm sure even I'd have caved at some point. Did you get to VHDL? Your project etched in pure silicon? Yes! Must! Oh no, my job is safe no longer!
162.
by daxfohl 9mo ago
Hasn't it just been subsumed by AdS/CFT now? IIUC that's a layer of abstraction but still primarily string theory under the hood. That's still an active area isn't it? Or is that dying too?
163.
by daxfohl 9mo ago
Even in the utopia scenario, that experiment has been taken to its natural conclusion on rats back in the 70s and the results were...interesting, to say the least. (google "Universe 25"). I feel like in many ways, a devolution to
164.
by daxfohl 9mo ago
Hard to say. In business we'll still have to make hard decisions about unique situations, coordinate and align across teams and customers, deal with real world constraints and complex problems that aren't suitable to feed to an LL
165.
by daxfohl 9mo ago
They've gotten pretty good IME so long as you guide it to think out of the box, give it the right level of background info, have it provide alternatives instead of recommendations, and do your best not to bias it in any particular dire
166.
by daxfohl 9mo ago
In some cases that could happen; in particular there may be a lot of UI and cross-app-integration style stuff that starts to get offloaded to users, so users can have AI code up their own UI for using some services together in the way that
167.
by daxfohl 9mo ago
From a PM perspective, the main differentiator between an engineering team and AI is "common sense". As these tools get used more and more, enough training data will be available that AI's "common sense" in terms of
168.
by daxfohl 9mo ago
Users don't want changes that rapidly. There's not enough people on the product team to design 20x more features. 20x more features means 400x more cross-team coordination. There's only positive marginal ROI for maybe 1.5-2x
169.
by daxfohl 9mo ago
I mean, it's somewhat moot, as even the formal hypothesis ("what is this proof proving") can be more complex than the code that implements it in nontrivial cases. So verifying that the proof is saying the thing that you actua
170.
by daxfohl 9mo ago
Agreed on both counts. Though for the first one it's still easier to implement things when bugs create compile or local unit/integration test errors rather than distributed service mismatches that can only be caught with extensive
171.
by daxfohl 9mo ago
I don't think "looks difficult" has been driving wages. FAANG etc leadership knows what's difficult and what's not. It's just marginal ROI. If you have a trillion-dollar market and some feature could increase t
172.
by daxfohl 9mo ago
Oh I've thought this for years. As an L7, basically my primary role is to serve as someone to bounce ideas off of, and to make recommendations based on experience. A chatbot, with its virtually infinite supply of experience, could oste
173.
by daxfohl 9mo ago
Though I think it's a very steep sigmoid that we're still far on the bottom half of. For math it just did its first "almost independent" Erdos problem. In a couple months it'll probably do another, then maybe one ea
174.
by daxfohl 9mo ago
Could be. Most projects I've worked on tend to span multiple services though, so I think AI would struggle more trying to understand and coordinate across all those services versus having all the logic in a single deployable instance.
175.
by daxfohl 9mo ago
Same, and similarly something like a "create a holistic design with all existing functionality you see in tests and docs plus new feature X, from scratch", then "compare that to the existing implementation and identify opport
176.
by daxfohl 9mo ago
The central equation is still marginal ROI. Will adding an engineer provide positive higher marginal ROI than some other investment? If so, hire; if not, fire. With the scale and size of the software industry about to explode even higher, o
177.
by daxfohl 9mo ago
Agreed, but: There's been a notable jump over the course of the last few months, to where I'd say it's inevitable. For a while I was holding out for them to hit a ceiling where we'd look back and laugh at the idea they&#
178.
by daxfohl 9mo ago
The last 20 years has seen a brain drain from other engineering and science into software. My guess is we'll see that finally start to reverse. Which, I think is great! How much value can yet another CRUD app possibly add to the world
179.
by daxfohl 9mo ago
This resonates with my experience in codex 5.2, at least directionally. I'm pretty persnickety about code itself, so I'm not to the point where I'll just let it rip. But in the last month or two things have gone from "I&
180.
by daxfohl 9mo ago
Soon it'll all be AI, so that'll fix it.

More ›