anuramat

6 ms·

31.
by anuramat 3mo ago
whats the point of a commit message if it can be inferred by an agent from the diff?
32.
by anuramat 3mo ago
why?
33.
by anuramat 3mo ago
the article implies that the perfect balance is 1:1; I want to believe this is just some sort of a ragebait-based PR strategy
34.
by anuramat 4mo ago
> they are ... finding ... vulnerabilities ... that have not been worth the time and effort ... that's kinda the entire point
35.
by anuramat 4mo ago
> 80-90% of the results why would I ever want 80% of the results though?
36.
by anuramat 4mo ago
research, refactoring, architecture also just code-that-is-not-as-shit in general for clearly defined stuff smaller models are fine, but at that point I'd rather write the code myself in any case, if there's a chance a cheaper mod
37.
by anuramat 4mo ago
> this year or next maybe; I meant "in a single year eventually", not this year necessarily > what research transformers were published in 2017, and nobody anticipated agentic coding back then; but if I had to guess: 1. lean
38.
by anuramat 4mo ago
> 100000x in a year why not? fable 5 (2026) is an inf% improvement over having nothing (2022) in any case, stop pretending that it can be represented by a single scalar value, or at least use log scale units
39.
by anuramat 4mo ago
> think about what my next question might be or you could just say it yourself instead
40.
by anuramat 4mo ago
how is GPT more general than LLM? also it's not really "LLM vs RL", they're kinda orthogonal
41.
by anuramat 4mo ago
> 80-90% efficiency wdym by that > for daily tasks which are?
42.
by anuramat 4mo ago
> much more difficult as in time consuming or error prone? if it's just time consuming, it still might be faster than writing it yourself
43.
by anuramat 4mo ago
> how would the patient see it idk, I just know doctors who complained about the same thing for years
44.
by anuramat 4mo ago
he isn't wrong about the doctors though
45.
by anuramat 4mo ago
if you can't tell if slop is correct, how do you know your code is correct? starting with a mental model and then writing the code yourself surely makes it feel safer, but it doesn't mean it is besides, it doesn't even have
46.
by anuramat 4mo ago
well yeah, would be kinda funny if people started exposing oil and tobacco companies as ethically questionable
47.
by anuramat 4mo ago
> less likely how so?
48.
by anuramat 4mo ago
just use bwrap?
49.
by anuramat 4mo ago
it's not like anthropic will stop improving their models
50.
by anuramat 4mo ago
do they also avoid AI code review? AI can't write bad code if it can't write code it's not like it's impossible to use agents without going 100% vibes
51.
by anuramat 4mo ago
> so much better it's enough for them to be slightly better for this to make sense; I'm not sure most people would consider this to be a worse product either -- it's annoying for devs and makes hotswapping models more of a
52.
by anuramat 4mo ago
> NOT real justification I thought it was widely accepted that it's not; eg https://www.anthropic.com/research/natural-language-autoenco...
53.
by anuramat 4mo ago
I don't see any hubris in competition
54.
by anuramat 4mo ago
if you care about privacy, claude/gpt is strictly not an option and there's nothing to think about > sending your money akchyually if you do it right, you are sending negative money; fair enough otherwise
55.
by anuramat 4mo ago
unrestricted access to better models at compute prices = better synthetic data and faster research, so its not just about the product imho
56.
by anuramat 4mo ago
but "harder to reverse engineer" isn't manufacturing, that's protecting your moat
57.
by anuramat 4mo ago
there is zero downside to not switching though: just use claude while it's good and subsidized, switch if rugpulled
58.
by anuramat 4mo ago
how is summarized CoT a moat, and how is having the top 2 LLMs not a moat?
59.
by anuramat 4mo ago
wdym?
60.
by anuramat 4mo ago
no way, the contents of "reasoning_summary" are summarized? fyi openai does the same; not really surprising or particularly evil

More ›