Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
atemerev
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
16 ms
·
301.
▲
by
atemerev
2y ago
Well, of course I check the output and correct it as needed. It is still much faster than writing it myself. And less boring. As for the documentation part — I infer that you hadn't used state of the art models, had you? They do not wr
302.
▲
by
atemerev
2y ago
"Given this code, extract all entities and create the database schema from these", "write documentation for these methods", "write test examples", "write README.md explaining how to use scripts in this dir
303.
▲
by
atemerev
2y ago
Well, if you have ADHD, you are always fifteen. Novelty-seeking is built in, part of physiology, cannot be ignored. I am 42, and LLMs still provide a lot of excitement to me.
304.
▲
by
atemerev
2y ago
It is 5x if you are already a senior SE knowing your programming language really well, constantly suggesting good architecture yourself ("seed files" is a brilliant idea), and not accepting any slop / asking to rewrite things
305.
▲
by
atemerev
2y ago
This is excellent, and matches my experience. Those lamenting the loss of manual programming: we are free to hone our skills on personal projects, but for corporate/consulting work, you cannot ignore 5x speed advantage. It's over.
306.
▲
by
atemerev
2y ago
No, I meant "good at persuading", it is not 100% efficiency of course.
307.
▲
by
atemerev
2y ago
Aider is free and open source, Cursor is free. You can connect Aider to various free models on OpenRouter, some of them state of the art.
308.
▲
by
atemerev
2y ago
What is this, some OpenAI employee fan fiction? Did Sam himself write this? OpenAI models are not even SOTA, except that new-ish style transfer / illustration thing that made all us living in Ghibli world for a few days. R1 is _better_
309.
▲
by
atemerev
2y ago
I prefer Aider, but that's a matter of taste.
310.
▲
by
atemerev
2y ago
By practice. It requires some skill too, to constantly remove slop and keep the project from deteriorating. Usually by suggesting good architecture, asking for tests, and rewriting / cleaning up some bad code endlessly. Still much fast
311.
▲
by
atemerev
2y ago
HIP is now somewhat viable (and ROCm is now all HIP). But — too late. First versions of ROCm were terrible. Too much boilerplate. 1200 lines of template-heavy C++ for a simple FFT. Can't just start hacking around. Since then, the CUDA
312.
▲
by
atemerev
2y ago
Including Hacker News, presumably.
313.
▲
by
atemerev
2y ago
Physics is famously non-deterministic. Quantum physics is built on irreducible randomness, it is incompatible with determinism (except for evolutions of probability quasidistributions). And we don't even need to try to find "quant
314.
▲
by
atemerev
2y ago
My (similarly religious) opinion is that because we _can_ simulate biological systems, computers can and will have free will too. Human brain is just a physical object that has some particular structure. We have free will (yes, I believe in
315.
▲
by
atemerev
2y ago
Hardly any actual physical simulations simulate every possible moment in time. We just calculate consequences of current events and put them in the queue to happen sometime in the future. And there's evidence that the real world works
316.
▲
by
atemerev
2y ago
For many people in post-Soviet countries, it was the other way around: raised as atheists, many of them found religion when it became allowed. I wonder if the same thing will happen with China.
317.
▲
by
atemerev
2y ago
Rule of law is universal. If they are illegal immigrants, it should be established by law and proper legal procedures. Otherwise, it is you who can be an illegal immigrant tomorrow, just because some dictator says it. You can prove otherwis
318.
▲
by
atemerev
2y ago
"most companies can survive a management self-destruct" — if it is an oil extraction company, maybe. Otherwise, knowledge and processes can be easily destroyed for good.
319.
▲
by
atemerev
2y ago
Do these look like MS-13 gang members? https://www.theguardian.com/us-news/2025/mar/18/germany-inve... https://www.dw.com/en/german-nationals-us-immigration-detain... https:/&
320.
▲
by
atemerev
2y ago
I know assembling things is half the fun, but if you just want a universal ESP32 device with screen and buttons, I recommend M5stack. Cheap, programmable, uniquitous.
321.
▲
by
atemerev
2y ago
Well, I do. I still like Outlook, and it works good enough as a web app. Same for OneNote. Azure as the development platform is crazy, of course. But I also have a friend who I respect much, and who is probably a better engineer than me, wh
322.
▲
by
atemerev
2y ago
It is usually available on Macs. Of course, there is still some specialized software for Windows, e.g. many CADs. But all these things are now marginal. If you are a software engineer and use Windows for your work, you are now a minority.
323.
▲
by
atemerev
2y ago
There are much more ads than viable targets.
324.
▲
by
atemerev
2y ago
The golden goose is now Azure. Windows is (or will be soon) a free complementary OS to make more people use Azure services.
325.
▲
by
atemerev
2y ago
Windows is a launcher for games, there is no other use for this system (correction: it still runs some specialized professional software that is too expensive to port). And now, even most of the games work on Linux too.
326.
▲
by
atemerev
2y ago
This is the best of possible worlds. You may not like it, but this is what peak performance looks like.
327.
▲
by
atemerev
2y ago
Mostly ad targeting (you can infer a lot of things from the global graph of contacts). Meta is an attention routing company.
328.
▲
by
atemerev
2y ago
That too, of course. WhatsApp itself was a work of art at that point, its success should be studied and hopefully emulated.
329.
▲
by
atemerev
2y ago
WhatsApp purchase was for that sweet sweet data of everyone's contact lists (this was their original innovation for onboarding — just give us access to your phone book and we'll tell you who else is on WhatsApp). Their earnings we
330.
▲
by
atemerev
2y ago
It is a fucking long time, but the ribosome is fucking complex (and ribosome biogenesis is _particularly_ fucking complex). My Fermi estimation skills give me the probability of ribosome evolving on Earth as "fucking whatever, more res
More ›