SamDc73

7 ms·

31.
by SamDc73 8mo ago
I mean they are only running a small version of codex can they run the full one? Or the technology isn't there yet?
32.
by SamDc73 8mo ago
Yes! I tried to go as far as 32k on the context window but beyond that it won't be usable on my laptop (Ryzen AI 365, 32gb RAM and 6gb of VRAM)
33.
by SamDc73 8mo ago
This is model 12188, which claims to rival SOTA models while not even being in the same league. In terms of intelligence per compute, it’s probably the best model I can realistically run locally on my laptop for coding. It’s solid for scrip
34.
by SamDc73 9mo ago
I was waiting for someone to say "this is what happens when you vibe code"
35.
by SamDc73 9mo ago
https://samdc73.com/
36.
by SamDc73 9mo ago
https://imgur.com/a/EmtADiA This how the graph looks like, not clustered by tag or anything ... (I was expecting a view like Logseq or Obsidian)
37.
by SamDc73 9mo ago
https://codose.ai Still experimenting with different ways to make learning easier using LLMs. I put together Codose as a tool where you paste a link to an Exercism or LeetCode problem, and it spins up a code editor with an AI tu
38.
by SamDc73 9mo ago
Nice one. But the graph view seems broken? And for the data, it would be nice to have the original URL for each comment as a reference.
39.
by SamDc73 9mo ago
Nice project. One thing I’ve always disliked about RSS (and this could actually fix it) is duplicates. When a new LLM model drops for example there are like ~5 blogs about it in my RSS feed saying basically the same thing, and I really only
40.
by SamDc73 9mo ago
> If you want a half-decent model, at least look at NYC I don't have the data, honestly, but isn't NYC (and it's surrounding cities/suburbs) more dense than the Bay Area? In SF (the city) transpiration is quite decent
41.
by SamDc73 9mo ago
The electric trains are a bit faster, but still slower than something like BART. My main issues with Caltrain are the cost (public transit shouldn’t be more expensive than driving) and the schedule: depending on the station, missing a train
42.
by SamDc73 9mo ago
As long as houses are treated as an investment, there will be all kind of incentives to make them go up in value. China did "over" solve it by building loads of houses, so everyone's house value just crippled down like crazy.
43.
by SamDc73 9mo ago
I see a lot of people complaining about red meat. It’s not the healthiest food, but it’s a much weaker risk factor than diets high in processed foods (including processed meats), refined carbs, added sugar, and excess salt. For adults (25–6
44.
by SamDc73 9mo ago
> then just shat all over them with start menus made with React Native, control-alt-delete menus that are actually just webviews, and forcing Copilot down everyone's throats Thank god I've been using Linux long enough to not ex
45.
by SamDc73 9mo ago
> Dubai has a lower median wage than either London or Mississippi, but people don't think of it as economically broken. Dubai isn’t sold as a place to belong long-term. Most people move there knowing it’s temporary. The Bay Area is
46.
by SamDc73 9mo ago
This reads a bit similar to the build-your-own-x series https://github.com/codecrafters-io/build-your-own-x Feel like one of these things a lot of talk about but very tiny do ...
47.
by SamDc73 10mo ago
Yes, turbine blades can introduce radar clutter and affect certain military systems; but this has been know since the 1990s and has been engineered around for decades. China, the UK, Germany, and Denmark operate gigawatts of offshore wind i
48.
by SamDc73 10mo ago
I do like OpenCode, but I get small bugs here and there like flickering, freezing and sometimes just crash all together. But their configuration setup is the easiest and best out of all the other CLI tools
49.
by SamDc73 10mo ago
Even with something like a 5090, I’d still run Q4_K_S/Q4_K_M because they’re far more resource-efficient for inference. Also, the 3090 supports NVLink, which is actually more useful for inference speed than native BF16 support. Maybe i
50.
by SamDc73 10mo ago
I only run small models (70b at my hardware gets me around 10-20 TOPS) for just random things (personal assistant kind of thing) but not for coding tasks. For coding related tasks I consume 30-80M tokens per day and I want something as fast
51.
by SamDc73 10mo ago
This older HN thread shows R1 running on a ~$2k box using ~512 GB of system RAM, no GPU, at ~3.5-4.25 TPS: https://news.ycombinator.com/item?id=42897205 If you scale that setup and add a couple of used RTX 3090s with heavy
52.
by SamDc73 10mo ago
Fair point. The USA PATRIOT Act shouldn’t have existed in the first place. But one of its most controversial parts (Section 215) did expire in 2020 (it was barely going to make it, though). But you’re right overall: most of the Act’s powers
53.
by SamDc73 10mo ago
If privacy is your top priority, then sure spend a few grand on hardware and run everything locally. Personally, I run a few local models (around 30B params is the ceiling on my hardware at 8k context), and I still keep a $200 ChatGPT subsc
54.
by SamDc73 10mo ago
U.S. lawmaking has a built-in ratchet effect: passing new laws is politically easier than repealing old ones. An easy way to solve this is all laws should have an expiration date by default.
55.
by SamDc73 10mo ago
Last time I checked, it was a pain in the ass to self-host it
56.
by SamDc73 10mo ago
I'm yet to come across applications implementing A2A in real life.
57.
by SamDc73 10mo ago
GPT-5 was the first model that occasionally produced code that I could push without any changes Claude still tends to add "fluff" around the solution and over-engineer, not that the code doesn't work, it's just that it&#
58.
by SamDc73 10mo ago
I find the Stanford channel a bit hard to navigate, MIT has a dedicated channel for the courses and another one for the other stuff, I wish Stanford did the same. The same applies to Harvard too
59.
by SamDc73 10mo ago
Angela Yu course is good, told my brother about it and he had hard time figuring some stuff out cause some pieces were outdated (he has zero experience)
60.
by SamDc73 10mo ago
Honestly, I feel like it's easy to filter out good/bad courses on Coursera by stuff like University name (although gotta admit last time did a course there were 3 years ago)

More ›