kimjune01

4 ms·

31.
by kimjune01 3mo ago
it's magic!
32.
by kimjune01 3mo ago
i like the tiny vision but it's yet to break out from its research project phase
33.
by kimjune01 3mo ago
vibe coding becomes much more tolerable when you disambiguate your agents with goals, constraints and heuristics up front.
34.
by kimjune01 3mo ago
this kind of benchmark is particularly susceptible to a narrow grader, where the input request is vague and therefore a significant range of SQL output would be valid, yet graded incorrect.
35.
by kimjune01 3mo ago
i tried to filter out a long list of these but still claude struggles to write like a human
36.
by kimjune01 3mo ago
understatement is a kind of humor, when he said 'redditor'
37.
by kimjune01 3mo ago
it would be nice if these benchmark reports actually specified which tasks they passed and which ones they didn't.
38.
by kimjune01 3mo ago
Chat data is the closest we ever got to purchase intent than demographics or keywords, attribution given a market position is already solved
39.
by kimjune01 3mo ago
the investors want their money to make more money
40.
by kimjune01 3mo ago
I do this
41.
by kimjune01 3mo ago
Attribution would be as easy as knowing which chatbot they came from? Unless chatbot publishers (openai in this case) allows you to target based on the attribution, that extra data will not be actionable, anyway.
42.
by kimjune01 3mo ago
the long tail of chatbot users will never pay for subscriptions, that's been proven since WhatsApp tried to charge a dollar or two per year per user
43.
by kimjune01 3mo ago
In a chatbot-first world, it's gonna have to either be generic impression ads that race to the bottom or contextual recommendations. Of course, the fear is that the advertiser hoards personal data or the chat answers are skewed to make
44.
by kimjune01 3mo ago
and LLMs are really good at filling in the blanks. It's like a really big import on generic application knowledge, and if you draw the circles well enough, it will draw the rest of the owl.
45.
by kimjune01 3mo ago
eval shops who depend on these things for a living will have to version their datasets, to prevent contamination
46.
by kimjune01 3mo ago
i meant to make the distinction between 'prompt' and 'spec' in that prompts are something that a human types in by hand. Interesting programs are a combination of canon and novel input
47.
by kimjune01 3mo ago
whether the hypothesis originates at the users or in your guess for the users is irrelevant in that they're both hypotheses with no certain outcome until trial.
48.
by kimjune01 3mo ago
I agree with iteration, but that's an orthogonal argument to how effective it is at turning prose to code. You are asking, LLMs are capable of translating a 835-page spec, but where does the spec come from? Follow the stream up from th
49.
by kimjune01 3mo ago
prompts are not specs because most often, prompts are underspecified. You can prompt fizzbuzz well enough for it to be functional, but not ffmpeg. So the distinction is less about Claude's capability, more about the quality of the inpu
50.
by kimjune01 3mo ago
a 17 year old can take a nude selfie and be charged as an adult in possession of CSAM
51.
by kimjune01 3mo ago
i thought this was what openclaw was for?
52.
by kimjune01 3mo ago
i found the same thing with Rust, verification steps are really helpful for speed and correctnesss
53.
by kimjune01 4mo ago
lean compiles or it doesnt
54.
by kimjune01 4mo ago
this is known as the oracle problem
55.
by kimjune01 4mo ago
one of those words that lost meaning, like samaritan or literally
56.
by kimjune01 4mo ago
In the days when Sussman was a novice, Minsky once came to him as he sat hacking at the PDP-6. "What are you doing?", asked Minsky. "I am training a randomly wired neural net to play Tic-tac-toe", Sussman replied. "
57.
by kimjune01 4mo ago
oh this again, i found this on stumbleupon circa 2007
58.
Auditing DeepSWE(june.kim)
3 points by kimjune01 5mo ago|0 comments
59.
Show HN: 97% on SWE-bench Verified with subscription-token agents(github.com)
2 points by kimjune01 5mo ago|0 comments
60.
by kimjune01 5mo ago
using AI to red-team your thoughts and assumptions is the fastest way to get smart since the dawn of time

More ›