Hacker News: guywhocodes

New comment by guywhocodes in "Reddit User Uncovers Who Is Behind Meta's $2B Lobbying for Age Verification Tech"

guywhocodes — Tue, 17 Mar 2026 12:10:49 +0000

This is probably protected free speech

New comment by guywhocodes in "Is Mozilla trying hard to kill itself?"

guywhocodes — Wed, 17 Dec 2025 12:31:11 +0000

It made me stop using FF on android.

New comment by guywhocodes in "Server DRAM prices surge 50% as AI-induced memory shortage hits hyperscalers"

guywhocodes — Tue, 04 Nov 2025 16:27:41 +0000

I was looking att filling my EPYC servers empty slots, what I paid $90/stick 2-3 years ago is now $430

New comment by guywhocodes in "We reduced a container image from 800GB to 2GB"

guywhocodes — Sun, 02 Nov 2025 12:32:12 +0000

The TLDR: > We tackled critical container image bloat on our Sealos platform, fixing a severe disk space exhaustion issue by shrinking an 800GB, 272-layer image to just 2.05GB.

They say they made a 800GB container image, so your issue is about singular vs plural?

Regardless, I don't really get why anyone would self report like this. Is next article going to be about how they don't encrypt passwords and when they accidentally dropped prod DB they could restore account from logs because it had the passwords in clear text?

New comment by guywhocodes in "LinkedIn will soon train AI models with data from European users"

guywhocodes — Mon, 22 Sep 2025 10:32:15 +0000

I wonder what this will do for enterprise sales posting

New comment by guywhocodes in "The HackberryPi CM5 handheld computer"

guywhocodes — Thu, 11 Sep 2025 06:36:56 +0000

When I've used i2s it has required setting spi clocks that made my spi devices not function. While it does have all of these IO buses, using more than one at a time is a bit of tetris. I wouldn't be surprised if there is some hardware constraint making i2s impossible.

New comment by guywhocodes in "Lossless LLM 3x Throughput Increase by LMCache"

guywhocodes — Sat, 28 Jun 2025 17:41:30 +0000

This is exactly what llm-d is

New comment by guywhocodes in "MCP in LM Studio"

guywhocodes — Thu, 26 Jun 2025 06:02:12 +0000

MCP Host is terminology from the spec. It's the software that makes llm calls, build prompts, interprets tool call requests and performs them etc.

New comment by guywhocodes in "Resurrecting flip phone typing as a Linux driver"

guywhocodes — Tue, 24 Jun 2025 05:41:15 +0000

A few years ago I started building this. The idea was to send the word and when cycling send the same number of backspaces before the next.

I guess I got busy with other things

New comment by guywhocodes in "GPT 4.5 level for 1% of the price"

guywhocodes — Sun, 16 Mar 2025 13:06:36 +0000

Europe 2.0

New comment by guywhocodes in "The AMD Radeon Instinct MI300A's Giant Memory Subsystem"

guywhocodes — Sat, 18 Jan 2025 18:54:07 +0000

More like the value of drivers that doesn't require one in-house team per customer to "fix" driver crashes in the customers' particular workloads.

New comment by guywhocodes in "Cognitive load is what matters"

guywhocodes — Thu, 26 Dec 2024 08:42:22 +0000

A tale as old as the SPA

New comment by guywhocodes in "Cognitive load is what matters"

guywhocodes — Wed, 25 Dec 2024 19:41:05 +0000

Is this bait?

New comment by guywhocodes in "Bringing K/V context quantisation to Ollama"

guywhocodes — Thu, 05 Dec 2024 08:41:59 +0000

This is par for Ollama, look at the log_probs issues/prs and you get an idea of how well Ollama is run.

Ollama is IMO a model downloader for llama.cpp so you can do roleplay with ease.

New comment by guywhocodes in "Large Enough"

guywhocodes — Wed, 24 Jul 2024 18:59:40 +0000

This is pretty much equivalent to the statement "multicharacter tokens are a dead end for understanding text". Which I agree with.

New comment by guywhocodes in "Hacked Nvidia 4090 GPU driver to enable P2P"

guywhocodes — Fri, 12 Apr 2024 20:21:20 +0000

We are talking about a literal monologue while poking at a driver for a few hours, this wasn't a huge project.

New comment by guywhocodes in "Spent brewer's yeast as a biosorbent for metal recovery from polymetallic waste"

guywhocodes — Tue, 19 Mar 2024 11:25:30 +0000

You can probably just burn it and recover the metals from the ashes chemically.

New comment by guywhocodes in "Bookish Diversions: Do Audiobooks Count?"

guywhocodes — Sun, 03 Mar 2024 22:11:28 +0000

I've found this to be particularly true with William Gibson, tried to listen to some but a split second lapses in attention forcing me to go back often and I gave up. And just pick up a text.

New comment by guywhocodes in "Ask HN: What have you built with LLMs?"

guywhocodes — Mon, 05 Feb 2024 20:36:21 +0000

I built a tool to create "average llm" probability of code for checking how aligned code is with what an LLM would output. Working on adding context from a project to check how the style of a section aligns with the style, content and domain of a project.

Idea is to use it to identify code that sticks out, because that usually what's interesting or bad.

New comment by guywhocodes in "Ceph: A Journey to 1 TiB/s"

guywhocodes — Sat, 20 Jan 2024 09:10:57 +0000

Did something very similar almost 10 years ago, EBS costs were 10x+ the cost for same perfomance CEPH cluster on the node disks. Eventually we switched to our own racks and cut it almost in ten again. We developed the inhouse expertise for how to do it and we were free.