c0rruptbytes

5 ms·

31.
by c0rruptbytes 3mo ago
my experience right now with fable via bedrock is they collect data, 5.6 has the option for ZDR at least
32.
by c0rruptbytes 3mo ago
few months behind? 5.6 came out last week and kimi is benching around the same right now
33.
by c0rruptbytes 3mo ago
everyone loving these linux sandboxes, even Apple wrote one https://github.com/apple/containerization/tree/main/examples...
34.
by c0rruptbytes 3mo ago
i think to fall in love with Pi, bundled skills are a bit antithetical - you realistically only need a couple of skills that you maybe design yourself
35.
by c0rruptbytes 3mo ago
inference is profitable, these companies are in the red because they're paying a premium to get the compute now versus later (because compute is the only moat when open models are catching up) we're literally looking at insane mar
36.
by c0rruptbytes 3mo ago
OpenAI prioritized compute much much earlier on, so they’re probably just more able to provide the model while Anthropic seems to be busting out the seams Sonnet 5 today was incredibly slow for example
37.
by c0rruptbytes 3mo ago
Deepseek isn't philanthropy, it's a hedgefund trying to short the western AI market by saying "hey we can do 90% of they can (arguably better at a density metric) for a 1/10th of the cost" it's my theory at lea
38.
by c0rruptbytes 3mo ago
if they’re paying for the tokens, what’s the problem
39.
by c0rruptbytes 4mo ago
i'm in, i think prices are gonna suck anyway, i own a playstation and that shit sucks, i want to do more couch co-op with my partner and the steam library opens up so much indie games can i build a mini pc myself? probably but meh
40.
by c0rruptbytes 4mo ago
TV is heavily subsidized from data collection and ads, not sure it's a perfect comparison
41.
by c0rruptbytes 4mo ago
https://blog.kilo.ai/p/did-claude-opus-48-distill-alibabas it happens to all models…when the internet is increasingly generated, things happen
42.
by c0rruptbytes 4mo ago
MCP tool search fixes the major issue imo, MCP clears skills/clis in every other way
43.
by c0rruptbytes 4mo ago
I would try a 6-bit MoE and maybe with unsloth's studio, they claim to have auto tool fixing which is where i see a lot of issues with MoEs I'm on a 48gb M5 Pro right now and it's been okay, a lot of my rough experiences have
44.
by c0rruptbytes 4mo ago
large contexts degrade the performance - attention doesn't work will for large windows like that and cloud models are kind of hacking it local models do involve some context engineering to get it okay, but it's not that rough
45.
by c0rruptbytes 4mo ago
q4 isn't rubbish, but it's a compromise for a good value, q6 is essentially a no-compromise quantization and it's what i recommend for MoEs in my experience for agentic workflows
46.
by c0rruptbytes 4mo ago
I'm talking about the common use case that I think hacker news people have: you get a macbook for work, you run the macbook they're not going to start giving GPUs to employees to run local models
47.
by c0rruptbytes 4mo ago
I don't know about good, I use a lot of local models and they're still pretty painful to run locally You have dense models (qwen 27b, gemma 31b) who are pretty smart, but pretty slow You have MoE models (gemma 26b, qwen 35b, north
48.
by c0rruptbytes 4mo ago
Minimax M3 too, and huawei claims to be releasing non-nvidia dependent training software too. openPangu 2.0 could be a shake-up if it holds up as a good model China may not care about open source, but they know they will personally fund AI
49.
by c0rruptbytes 4mo ago
i’m running m4 pro 48gb right now omlx + gemma 12b 6 bit + pi it’s feasible for sure MoEs for speed (qwen 35b, cohere 30b, gemma 26b) Dense for more methodical work (qwen 27b [reigning champ], gemma 31b, gemma 12b) MoE i recommend 5bit+ Den
50.
by c0rruptbytes 4mo ago
it does not result in great results left unattended, it’ll start creating slop or hardcoding solutions but overtime if you adjust your verification rubric, it’s not too bad, gets pretty good, if you do make it do TDD, it gets kinda crazy an
51.
by c0rruptbytes 4mo ago
There's `honcho` for memory, i'm starting to play with it now, but I feel like I've seen a lot of projects pop up for it
52.
by c0rruptbytes 4mo ago
I like Zed... but AI dev workflows get complicated fast you start with claude code or codex and it's cute, but then you realize - hmm configuration is cheap, the AI can do it! then you start looking into MCPs and skills, fuck it, oh-my
53.
by c0rruptbytes 4mo ago
seems more like a culture problem, i have my calendar very public, all my junior devs know ill get on a zoom with no hesitation and they actually seem to enjoy the screen sharing, every zoom is recorded with AI summary/transcript so th
54.
by c0rruptbytes 4mo ago
we hired a few juniors at our fully remote company - no issue this is ft trying to help their real estate portfolio
55.
by c0rruptbytes 4mo ago
fair, i think i was referring more to 1.58 bit architecture in general since the original paper (Figure 3) shows that we eliminate FP16 multiplication and addition just for INT8 addition. I need to dive deeper into bonsai overall if it diff
56.
by c0rruptbytes 4mo ago
AI doesn’t have a model or agent availability problem to be fair, it does have a positive outreach problem and pewdiepie can do extremely well there just my 2c
57.
by c0rruptbytes 4mo ago
ideally if ternary models work, the math is extremely easy for computers (addition/subtraction vs 16 bit multiplication)
58.
by c0rruptbytes 4mo ago
> If Rust is the new C++, Zig is the new C. thank you, this helps!
59.
by c0rruptbytes 4mo ago
I run Deepseek at home...only people getting money is my electricity company
60.
by c0rruptbytes 4mo ago
i'm not cool and hip like hacker news devs, but I've been seeing Zig a lot, is this the new cool thing on the street? no more Rust?

More ›