Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
dyl000
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
5 ms
·
31.
▲
by
dyl000
1y ago
Anubis isn’t letting me through ;(
32.
▲
by
dyl000
1y ago
regular people, not really interested in tech chasing 'easy tech salaries' coming up against the fact its a hard industry.
33.
▲
by
dyl000
1y ago
answer: no
34.
▲
by
dyl000
1y ago
openAI models have an issue where they are pretty good at everything but not incredible at anything. They're too well rounded. for coding you use anthropic or google models, I haven't found anyone who swears by openAI models for c
35.
▲
by
dyl000
1y ago
it was only a matter of time, they have too much good data to not train their own models, not to mention that claude API calls were probably killing their profitability. open source alternative https://huggingface.co/SWE-ben
36.
▲
by
dyl000
1y ago
Remove the gil first…
37.
▲
by
dyl000
1y ago
I was a real big fan of vb.net! Built so much with it.
38.
▲
by
dyl000
1y ago
o3 is useless because of its hallucinations
39.
▲
by
dyl000
1y ago
I don't understand... it's a vscode fork where they're earning $20/mo for customers that are regularly costing much more. They have no moat. I've used all 500 of my fast requests this month with cursor, I've 10
40.
▲
by
dyl000
2y ago
Looks like a tissot PRX
41.
▲
by
dyl000
2y ago
llama has order of magnitude mode compute requirement than deepseek.
42.
▲
by
dyl000
2y ago
this is hilarious, you define misinformation as things that go against mainstream consensus. Yes, please -- Believe everything you're told. You're not testing peoples ability to separate wheat from chaff, you're just diffing
43.
▲
by
dyl000
2y ago
I wonder how long before we start to acknowledge that AI labs are heavily gaming benchmarks and they are mostly useless as a way of judging model performance. The latest one to be caught was Meta, but they've all been doing it for a wh
44.
▲
by
dyl000
2y ago
I’d assume effect would increase how much memory JS uses, you’re creating a bunch of memory junk. In theory it should make you able to contain errors with exception handing.
45.
▲
by
dyl000
2y ago
Thats one of my primary issue with the Node community. When I wrote Go for a few months I was blown away by how many projects I came across that hadn't had a commit to their repo in 3-4 years but people were still recommending them, th
46.
▲
by
dyl000
2y ago
I like ideas from effect, I use my custom result type all over my application as well as my custom tryCatch & promise.allSettled wrappers. Issue comes down to you need to wrap all of your code in effect/results. Something like rust
47.
▲
by
dyl000
2y ago
https://github.com/vercel/next.js/pull/75877 memory leaks in the runtime and frameworks all the time. This setTimeout leak crashed our prod machine multiple times until we rolled back.
48.
▲
Everything I Was Lied to About Node.js Came True with Elixir
(d-gate.io)
16 points
by
dyl000
2y ago
|
4 comments
49.
▲
by
dyl000
2y ago
I do get access to a version of reddit that isn't astroturfed by bots?
50.
▲
by
dyl000
2y ago
Apparently performance on larger context kinda sucks. Still impressive we have such large context on open source model.
51.
▲
by
dyl000
2y ago
I was bearish on mongodb after I used it for the first time.
52.
▲
by
dyl000
2y ago
Mine was wrong 1 Problem Understanding The task is to count the number of 'r's in the word 'strawberry'. This requires identifying each occurrence of the letter 'r' in the given word and keeping a count. 2 Data