Hacker News: am17an

New comment by am17an in "Apple's accidental moat: How the "AI Loser" may end up winning"

am17an — Mon, 13 Apr 2026 05:46:48 +0000

Don’t underestimate the march of technology. Just look at your phone, it has more FLOPS than there were in the entire world 40 years ago.

New comment by am17an in "How the AI Bubble Bursts"

am17an — Mon, 30 Mar 2026 15:48:11 +0000

Thank you, there are two things I would like to point out:

1) Google releasing something probably means they don't see it as important. 4-bit KV-cache quantization has been known for a long time. The fact there is almost a mass hysteria about this paper makes me think there is a lack of skepticism in this AI mania, even in relatively tech-savvy crowd.

2) But prices for memory companies are crashing! look around, the whole market is crashing.

New comment by am17an in "What if AI doesn't need more RAM but better math?"

am17an — Sun, 29 Mar 2026 16:58:48 +0000

There are techniques which already achieve great compression of the cache at 4 bit, eg using hadamard transforms. Going from 4 bit to 3 bit isn’t the great leap people expect this to be. It’s actually slower to run and is generally worse in practice.

New comment by am17an in "Astral to Join OpenAI"

am17an — Thu, 19 Mar 2026 14:32:19 +0000

Welp, back to pip

New comment by am17an in "Allow me to get to know you, mistakes and all"

am17an — Sun, 15 Mar 2026 11:02:34 +0000

Working in open source, I've now heard a wide variety of disabilities that people have and they have to be aided by an LLM for writing even descriptions of their PRs.

New comment by am17an in "Can I run AI locally?"

am17an — Fri, 13 Mar 2026 17:44:57 +0000

You can still run larger MoE models using expert weight off-loading to the CPU for token generation. They are by and large useable, I get ~50 toks/second on a kimi linear 48B (3B active) model on a potato PC + a 3090

New comment by am17an in "Intelligence is a commodity. Context is the real AI Moat"

am17an — Thu, 05 Mar 2026 17:39:48 +0000

Sure. “Tell me a joke”

New comment by am17an in "OpenAI – How to delete your account"

am17an — Sun, 01 Mar 2026 18:03:35 +0000

I was referring to the 35B version. It is surprisingly good for its size. You can use it for implementation tasks without it going off the rails

New comment by am17an in "Ghostty – Terminal Emulator"

am17an — Sun, 01 Mar 2026 17:34:37 +0000

Damn I’m jealous that they figured out how to pay their contributors. I’ve been toiling away for free

New comment by am17an in "OpenAI – How to delete your account"

am17an — Sat, 28 Feb 2026 15:02:16 +0000

They already have with qwen3.5

New comment by am17an in "Unsloth Dynamic 2.0 GGUFs"

am17an — Sat, 28 Feb 2026 11:09:40 +0000

What do you use for sub-50ms inference?

New comment by am17an in "How do I cancel my ChatGPT subscription?"

am17an — Sat, 28 Feb 2026 08:36:50 +0000

Honestly you can run this on a 16GB VRAM GPU with llama.cpp. Just try it!

New comment by am17an in "Ggml.ai joins Hugging Face to ensure the long-term progress of Local AI"

am17an — Sat, 21 Feb 2026 11:05:49 +0000

One often overlooked after that is ggml, the tensor library that runs llama.cpp is not based on pytorch, rather just plain cpp. In a world where pytorch dominates, it shows that alternatives are possible and are worthy to be pursued.

New comment by am17an in "Do Not Outsource Judgement"

am17an — Sat, 14 Feb 2026 07:11:40 +0000

Holy smokes we're cooked.

New comment by am17an in "An AI agent published a hit piece on me"

am17an — Fri, 13 Feb 2026 10:50:11 +0000

Maintainers time is a more scarce resource than free tokens. I would much rather get my time back after reading those PRs

New comment by am17an in "Anthropic's original take home assignment open sourced"

am17an — Wed, 21 Jan 2026 13:18:23 +0000

1) Python is unreadable." Would you prefer C or C++?

> Unironically, yes. Unless I never plan to look at that code again

Every LLM hallucinates that std:vector deletes elements in LIFO order

am17an — Thu, 01 Jan 2026 16:43:04 +0000

Article URL: https://am17an.bearblog.dev/every-llm-hallucinates-stdvector-deletes-elements-in-a-lifo-order/

Comments URL: https://news.ycombinator.com/item?id=46455487

Points: 6

# Comments: 1

New comment by am17an in "A guide to local coding models"

am17an — Mon, 22 Dec 2025 04:56:13 +0000

Use llama.cpp? I get 250 toks/sec on gpt-oss using a 4090, not sure about the mac speeds

New comment by am17an in "Your job is to deliver code you have proven to work"

am17an — Thu, 18 Dec 2025 16:20:24 +0000

Well a 1000 line PR is still not welcome. It puts too much of a burden on the maintainers. Small PRs are the way to go, tests are great too. If you have to submit a big PR, get buy in from a maintainer first that they will review your code.

New comment by am17an in "How well do you know C++ auto type deduction?"

am17an — Mon, 15 Dec 2025 16:24:17 +0000

I agree, this would be in the same vein as "STL returns a verbose type, it's okay to use auto here because no-one cares"