Hacker News: why_only_15

New comment by why_only_15 in "The SEC Should Block Anthropic's IPO to Protect the Public"

why_only_15 — Tue, 03 Nov 2026 04:54:24 +0000

Over and over again they talk about how Anthropic's margins are getting squeezed. They're so squeezed in fact that Anthropic has merely 85% gross margins! They're earning merely $6 in revenue for every $1 they spend!

New comment by why_only_15 in "The AI Race Just Got Awkward"

why_only_15 — Wed, 30 Sep 2026 16:44:00 +0000

Why do you think the Chinese labs figured this out before the western labs? No reason to believe that whatsoever.

New comment by why_only_15 in "Saving another 100TB of RAM"

why_only_15 — Mon, 21 Sep 2026 09:50:49 +0000

yeah there's some KV cache offload

New comment by why_only_15 in "Saving another 100TB of RAM"

why_only_15 — Mon, 21 Sep 2026 09:50:29 +0000

This is just wrong unless you're using some confusing definition. Notice that companies trying to do lots of inference aren't looking for CPUs, they're looking for GPUs/TPUs.

New comment by why_only_15 in "Saving another 100TB of RAM"

why_only_15 — Fri, 18 Sep 2026 22:19:40 +0000

CPU DRAM can't really be used for inference efficiently -- inference mostly wants memory bandwidth, not memory capacity, and GPU DRAM has >10x more bandwidth. The fabs can switch between them but you can't switch after the fact.

Show HN: Helping agents break out of their sandboxes in a monitorable way

why_only_15 — Wed, 16 Sep 2026 23:22:24 +0000

Article URL: https://sparrowsystems.co/

Comments URL: https://news.ycombinator.com/item?id=49734386

Points: 1

# Comments: 0

New comment by why_only_15 in "Introducing System One Models and Jev"

why_only_15 — Tue, 15 Sep 2026 20:10:51 +0000

They have various benchmarks, e.g. how much time it takes them to do wikipedia page -> page games. Jev seems to take the same or fewer hops but in ~10x less time and for ~10x less money.

It's totally reasonable to compare against LLMs doing chain of thought if it gets comparable performance.

Show HN: What people watched on Netflix since 2023

why_only_15 — Mon, 03 Aug 2026 13:13:38 +0000

Netflix makes reports since 2023 showing how much engagement everything in their catalog gets. This joins together all those into one report.

Most surprising to me: 1. What a mega-hit KPop Demon Hunters was. The #2-5 ranked movies got ~500M hours viewed each, KPop Demon Hunters got 1081M. 2. How big old content is. Suits from 2011 is the #2 show on Netflix. Grey's Anatomy starting in 2005 is #3. Gilmore Girls from 2000 #4. 3. The "scale effect" where releasing a new season of a show engenders massive viewing of previous seasons. Stranger Things season 5 got 880M hours of viewing in H2 2025 when it released. But it caused 1200M additional hours of viewing of seasons 1-4. 4. Massive popularity of movies I had never heard of. Leave the World Behind, a movie I had never heard of, was the #2 movie of the last three years. It got ~100 minutes of watchtime per Netflix subscriber. Of the top 10 movies I had only heard of 3.

This was vibe-coded, so could be incorrect, especially in details. I think the overall results of the report are correct in terms of # of hours watched since 2023, though. Reproduction script: girl.surgery/netflix_engagement/reproduce.py.

Comments URL: https://news.ycombinator.com/item?id=49155362

Points: 1

# Comments: 0

New comment by why_only_15 in "Postmortem for Kernel Soundness Bug #14576"

why_only_15 — Sun, 02 Aug 2026 02:28:32 +0000

To what degree can you say something like "if the kernel doesn't have metaprogramming [or some other set of features] it's fine". When were the last bugs with a reduced feature set?

New comment by why_only_15 in "Google fixed more Chrome bugs in June than over the past two years, thanks to AI"

why_only_15 — Fri, 31 Jul 2026 10:45:16 +0000

The whole point is that these are security bugs that didn't necessarily bother people so far but could be exploited

New comment by why_only_15 in "How AST-grep Rewrote Tree-sitter in Rust and Made It 30% Faster"

why_only_15 — Mon, 27 Jul 2026 03:24:29 +0000

I really don't like this. Don't post your AI-generated articles here. https://www.pangram.com/history/fce67fb4-4f6a-4eb4-a323-5048...

New comment by why_only_15 in "Running Gemma 4 26B at 5 tokens/sec on a 13-year-old Xeon with no GPU"

why_only_15 — Thu, 16 Jul 2026 02:27:45 +0000

The correct comparison is not sonnet, but qwen3.5-27b on a cloud. Alibaba's pricing [0] is $0.20/m input $1.56/m output, so $0.72 for the prompt processing or $0.22 for the generation. Yours is still cheaper but the margins are less.

My guess is that this math gets less good with MoE (because you will be limited by VRAM, but clouds won't).

[0]: https://openrouter.ai/qwen/qwen3.5-27b

New comment by why_only_15 in "Running Gemma 4 26B at 5 tokens/sec on a 13-year-old Xeon with no GPU"

why_only_15 — Thu, 16 Jul 2026 02:20:15 +0000

Why is the post AI-generated? If you're going to make something for us to read you pay us the courtesy of actually writing it first.

https://www.pangram.com/history/61fbc90e-180d-4d91-a85d-16ff...

New comment by why_only_15 in "Running Gemma 4 26B at 5 tokens/sec on a 13-year-old Xeon with no GPU"

why_only_15 — Thu, 16 Jul 2026 02:15:59 +0000

The post itself is totally AI-generated. It has tons of tells, in addition to Pangram saying so.

https://www.pangram.com/history/61fbc90e-180d-4d91-a85d-16ff...

New comment by why_only_15 in "When AI Builds Itself: Our progress toward recursive self-improvement"

why_only_15 — Fri, 05 Jun 2026 02:27:50 +0000

why do you think so? they provide some evidence of this in the article, but there have been several improvements in e.g. nanogpt-speedrun or openai parameter golf made by AIs

New comment by why_only_15 in "Where are the economies of scale in homebuilding?"

why_only_15 — Fri, 29 May 2026 03:57:21 +0000

Does China subsidize housing construction? Is your claim something like "China encourages/forces consumers to save at high rates which lowers interest rates and makes there be more construction"?

New comment by why_only_15 in "Project Glasswing: An Initial Update"

why_only_15 — Fri, 22 May 2026 20:39:28 +0000

what's the origin of your $20k/vuln estimate?

New comment by why_only_15 in "Talkie: a 13B vintage language model from 1930"

why_only_15 — Tue, 28 Apr 2026 06:08:30 +0000

The formal Ottoman name was Kostantiniyye=Constantinople until the empire's fall in 1922. The official shift happened in 1930, with the Turkish Postal Services Law changing the name to Istanbul.

New comment by why_only_15 in "DRAM has a design flaw from 1966. I bypassed it [video]"

why_only_15 — Fri, 10 Apr 2026 07:40:42 +0000

This is a quite old technique. The idea, as I understood it, was that lots of data at Google was stored in triplicate for reliability purposes. Instead of fetching one, you fetched all three and then took the one that arrived first. Then you sent UDP packets cancelling the other two. For something like search where you're issuing hundreds of requests that have to resolve in a few hundred milliseconds, this substantially cut down on tail latency.

New comment by why_only_15 in "Google releases Gemma 4 open models"

why_only_15 — Fri, 03 Apr 2026 02:07:05 +0000

I am in this world, but am not familiar with this specifically.

My guess is that they found a bug with their implementation of the model using the weights Google released. These bugs are often difficult to track down because the only indication is that the model is worse with your implementation than with someone else's.