Hacker News: androiddrew

New comment by androiddrew in "Taking on CUDA with ROCm: 'One Step After Another'"

androiddrew — Tue, 14 Apr 2026 10:41:57 +0000

I’ll try and get in touch with them. Thank you.

New comment by androiddrew in "GAIA – Open-source framework for building AI agents that run on local hardware"

androiddrew — Tue, 14 Apr 2026 00:49:05 +0000

You got it right I think. I’m sitting with two “AI Ready Radeon AI Pro 9700 workstation cards, which are RDNA4 not CDNA. My experience is that my cards are not a priority. Individual engineers at AMD may care, the company doesn’t. I have been trying since February to get ahold of anyone responsible for shipping tuned Tensile gfx1201 kernels in rocm-libs, which is used by Ollama.its been three weeks since I raised enough hell on the discord to get a response, but they still can’t find “who” is responsible for Tensile tuning, and “if” they are even going to do it for the gfx12* cards.

Don’t get me started with vLLM and AITER.

New comment by androiddrew in "GAIA – Open-source framework for building AI agents that run on local hardware"

androiddrew — Tue, 14 Apr 2026 00:00:51 +0000

Yup, meanwhile Jensen is on the Lexfriedman podcast stating the reason why CUDA is successful is because all thier devices run it. The on ramp is at the individual user.

I have and RDNA4 card and they certainly are prioritizing CDNA over a CDNA + RDNA strategy or a unification strategy.

New comment by androiddrew in "Taking on CUDA with ROCm: 'One Step After Another'"

androiddrew — Mon, 13 Apr 2026 12:10:47 +0000

I have been trying since February to get someone at AMD to shipped tuned Tensile kernels in the rcom-libs for the gfx1201. They are used by Ollama but no one on the Developer Discord knows who is responsible for that. It has been pretty frustrating and it shows that AMD has an organizational problem to overcome in addition to all the things technically that they want rocm to do.

New comment by androiddrew in "Will I ever own a zettaflop?"

androiddrew — Fri, 10 Apr 2026 00:06:22 +0000

Not with the price of silicon being what it is

New comment by androiddrew in "Things I Think I Think... Preferring Local OSS LLMs"

androiddrew — Thu, 02 Apr 2026 11:54:02 +0000

I love local first. I am finding that a 120B MoE is hitting the sweet spot for local hosted. Right now that takes a 2K strix halo, a 4k GB10 machine, or a 5k Mac Pro. 2 years from now I think hardware will take us back to the 2k ish range with good performance.

I love my dual GPU setup (2AMD Radeon r9700 64GB vram) but it costs 5x electricity than my GX10 (GB10 chip inside) and since layers are landing in system memory my TPS is half the GX10.

Now a dense model like Devstral2 24B slaps on the Dual GPU setup. I just haven’t gotten as much out of that as I have the 120 MoEs

New comment by androiddrew in "Ollama is now powered by MLX on Apple Silicon in preview"

androiddrew — Tue, 31 Mar 2026 11:14:01 +0000

Get turboquant 4 bit implemented and this would be game changer.

New comment by androiddrew in "The Little Book of C"

androiddrew — Thu, 26 Mar 2026 22:33:51 +0000

Wish they had this for zig

New comment by androiddrew in "Jury finds Meta liable in case over child sexual exploitation on its platforms"

androiddrew — Wed, 25 Mar 2026 10:55:49 +0000

Alternative headline: household spyware cash machine forced to pay $20 for being bad.

If you want to punish Meta then you have to punish the wonder boy who runs it. Not even share holders can fight off the guy spending 80B on the metaverse.

New comment by androiddrew in "Microsoft weighs legal action over $50B Amazon-OpenAI cloud deal"

androiddrew — Wed, 25 Mar 2026 00:10:07 +0000

The poster probably is hoping someone will post the archived version in the comments

New comment by androiddrew in "So where are all the AI apps?"

androiddrew — Wed, 25 Mar 2026 00:01:24 +0000

I heard a delightful term for building apps only for yourself “houseplant programming”.

New comment by androiddrew in "Tinybox – A powerful computer for deep learning"

androiddrew — Sun, 22 Mar 2026 13:26:26 +0000

Could you share what you are using for inference and how you are running it? I have a 64G VRAM/128G system RAM setup.

New comment by androiddrew in "Ozempic Is About to Go Generic for Billions of People"

androiddrew — Sat, 21 Mar 2026 13:58:56 +0000

Good, because it’s fucking ridiculous that pharma gets special patent loop holes to maintain a monopoly beyond what the basic protection grants.

New comment by androiddrew in "Ubuntu 26.04 Ends 46 Years of Silent sudo Passwords"

androiddrew — Sat, 21 Mar 2026 13:56:27 +0000

I don’t know why this keeps coming up. Has this been a big deal for everyone else? Like ok usability improvement, but the number of times I have read an article about this is silly.

New comment by androiddrew in "Thoughts on OpenAI acquiring Astral and uv/ruff/ty"

androiddrew — Fri, 20 Mar 2026 14:43:37 +0000

Yes, when the poetry people purposely added a feature to fail CI with a 1/10 chance because they wanted to depreciate a feature, I depreciated poetry.

New comment by androiddrew in "Leanstral: Open-source agent for trustworthy coding and formal proof engineering"

androiddrew — Mon, 16 Mar 2026 23:47:05 +0000

MOE but 120B range. Man I wish it was an 80B. I have 2 GPUs with 62Gib of usable VRAM. A 4bit 80B gives me some context window, but 120B puts me into system RAM

New comment by androiddrew in "How "Hardwired" AI Will Destroy Nvidia's Empire and Change the World"

androiddrew — Sat, 14 Mar 2026 22:24:18 +0000

Yeah, well might just come on your new laptop

New comment by androiddrew in "How "Hardwired" AI Will Destroy Nvidia's Empire and Change the World"

androiddrew — Sat, 14 Mar 2026 22:21:27 +0000

Give me a 120B dense model on one of these and yeah my API use will probably drop.

New comment by androiddrew in "New KV cache compaction technique cuts LLM memory 50x without accuracy loss"

androiddrew — Sun, 08 Mar 2026 01:03:14 +0000

I hope this is real.

New comment by androiddrew in "Why developers using AI are working longer hours"

androiddrew — Sun, 08 Mar 2026 00:37:54 +0000

I have never been in a flow state with an agent running. I use agents, but that isn’t flow.