keeda

38 ms·

571.
by keeda 8mo ago
> When surveyed, 30% to 50% of developers told us that they were choosing not to submit some tasks because they did not want to do them without AI. This implies we are systematically missing tasks which have high expected uplift from AI
572.
by keeda 8mo ago
Actually, no those are separate things, and you can see for yourself with this simple experiment. Just adding "Check your assumptions" to the prompt -- note, without specifying what assumptions! -- leads them to reason to the righ
573.
by keeda 8mo ago
> They'll eventually train it to answer this specific question, and then it will perform worse in some other aspect. Not necessarily. Simply asking models to "check your assumptions" -- note, without specifying what assu
574.
by keeda 8mo ago
I'm imposing but could you try these runs again with this small change: Simply append “Make sure to check your assumptions.” to the question. Note, it does not mention what assumption specifically. In my experiments, after the models g
575.
by keeda 8mo ago
Ehh, it's called offshoring for programmers too. Maybe "outsourcing" in some places. The Indian IT giants themselves call it the onsite / offshore model, with a few "onsite" people -- liasons typically on H1B e
576.
by keeda 8mo ago
All these points are brought up all the time but the upshot is, based on reporting from Microsoft and StatCounter, Windows marketshare actually grew . Point is, we techies might chafe at and complain about all these anti-consumer shenaniga
577.
by keeda 8mo ago
Periodic reminder (note, originaly posted in 2013): https://arstechnica.com/gadgets/2018/07/googles-iron-grip-on... At the risk of posting memes to HN: https://imgflip.com/i/akp488
578.
by keeda 8mo ago
I wouldn't hold my breath: https://arstechnica.com/gadgets/2026/01/windows-11-has-hit-1... https://gs.statcounter.com/os-market-share/
579.
by keeda 8mo ago
Yes, we as a society urgently have to figure out how to learn and educate with AI. There are even studies showing that students who use AI to do their work do not learn the necessary skills. And I'm also hearing grumblings about entry
580.
by keeda 8mo ago
Let's also not forget the multiple other studies that found significant boosts to productivity using rigorous methods like RCTs. However, because these threads always go the same way whenever I post this, I'll link to a previous t
581.
by keeda 8mo ago
The real takeaway here -- also corroborated by the DORA 2025 report https://dora.dev/research/2025/ -- is that more than anything, AI amplifies your current development culture. Organizations with strong quality c
582.
by keeda 8mo ago
Writing a good ticket is not a common skill. IMO it seems deceptively easy but usually requires years of experience to understand what to include and express it in the most concise yet unambiguous terms possible for the intended audience.
583.
by keeda 8mo ago
Haven't looked into this much myself, but throwing it out there: https://statmodeling.stat.columbia.edu/2025/04/19/for-15-yea...
584.
by keeda 8mo ago
The whole thing is wild. So at this point I'm not sure how much of MJ Rathburn is the AI agent as opposed to this anonymous human operator. Did the AI really just go off the rails with negligible prompting from the human as TFA claims,
585.
by keeda 8mo ago
I got the "you should walk" answer 4 out of 5 times with free ChatGPT, until I told it to, basically, "think carefully": https://news.ycombinator.com/item?id=47040530
586.
by keeda 8mo ago
The broad point about assumptions is correct, but the solution is even simpler than us having to think of all these things; you can essentially just remind the model to "think carefully" -- without specifying anything more -- and
587.
by keeda 8mo ago
Easily fixed by appending “Make sure to check your assumptions” to the question: https://imgur.com/a/WQBxXND Note, what assumption isn't even specified. So when the Apple “red herrings trashes LLM accuracy” stud
588.
by keeda 8mo ago
And yet they sat on transformers until OpenAI kicked off the AI boom by actually productizing that research in ChatGPT. Though it's possible they were just being cautious, my uncharitable view is that they knew this would disrupt their
589.
by keeda 8mo ago
There are some interesting dynamics going on at Ars. I get the sense that the first author on the pulled article, Benj Edwards, is trying to walk a very fine line between unbiased reporting, personal biases, and pandering to the biases of t
590.
by keeda 8mo ago
At first glance, dropping "safety" when you're trying to benefit "all of humanity" seems like an insignificant distinction... but I could see it snowballing into something critical in an "I, Robot" sense (
591.
by keeda 8mo ago
This makes sense to me because I've been having very accurate results with models from even 2+ years ago... but I had to "hold them right." Even when reasoning models and coding agents were just a gleam in Altman's and A
592.
by keeda 8mo ago
In the initial analysis of a single worker, yes, but when scaling up per-human savings to use by the wider population, the aggregate resource savings compensate for training resource usage within a few days, weeks at most.
593.
by keeda 8mo ago
Hah, interesting, I intended my exclamation marks come across as surprise or emphasis rather than glee! If it helps better interpret my posts, these days my frame of mind is "We're living in very interesting times" in the sen
594.
by keeda 8mo ago
This includes things like drinking, sanitation, etc. Derived the number from here: https://www.epa.gov/watersense/statistics-and-facts Mostly lines up with this reference too, which focuses only on water usage at work:
595.
by keeda 8mo ago
True, but I think in these scenarios they rely on prompt caching, which is much cheaper: https://ngrok.com/blog/prompt-caching/ I have no expertise here, but a couple years ago I had a prototype using locally depl
596.
by keeda 8mo ago
The thing is, there are many interplaying dynamics here that are impossible to unravel. This is why I called it "napkin math", because figuring out the full ramifications of this change is a pretty large economic problem that nobo
597.
by keeda 8mo ago
> These things are so hideously inefficient. Quite the opposite, really. I did some napkin math for energy and water consumption, and compared to humans these things are very resource efficient. If LLMs improve productivity by even 5%
598.
by keeda 8mo ago
I've mentioned this before ( https://news.ycombinator.com/item?id=45987316 ), but I follow a labor economist on LinkedIn who puts things into a balanced perspective. The thing he reminds us often is you can't take t
599.
by keeda 9mo ago
I think there is another dynamic that this article hints at but does not call out explicitly: work becoming much more intense because AI is now doing most of the cognitively less intense -- maybe even relaxing -- "donkey work." Th
600.
by keeda 9mo ago
I have a slightly different theory for AI fatigue. Programming essentially comprises three aspects: problem-solving, coding up the solution, and seeing the solution work. We used to have a balance between them that AI agents are disrupting.

More ›