Hacker News: pshirshov

New comment by pshirshov in "Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark"

pshirshov — Fri, 22 May 2026 15:24:54 +0000

That's curious, I've been trying to do some parametric modeling with Claude - and its performance was abysmal.

New comment by pshirshov in "Was my $48K GPU server worth it?"

pshirshov — Fri, 22 May 2026 10:47:56 +0000

Questions:

1) Was the energy bill factored in? 2) Have you extracted any comparable value out of this?

New comment by pshirshov in "VSM Loop: Viable-System Orchestration for Multi-Agent R&D"

pshirshov — Wed, 20 May 2026 16:57:27 +0000

I guess noone would ever pay any attention but this thing works wonders (even its sub-skills are very valuable). It turns Claude/Codex into persistently working machines with much higher output quality - and without need for a custom harness (which actually could improve token usage, quality and discipline but comes with maintenance costs).

With these skills Claude can work completely autonomously for days and weeks only interrupting the flow when some input is genuinely required from the user.

A sandbox and yolo mode are must.

The theory behind, unfortunately, doesn't have the recognition it deserves.

VSM Loop: Viable-System Orchestration for Multi-Agent R&D

pshirshov — Wed, 20 May 2026 16:52:53 +0000

Article URL: https://github.com/7mind/nix-config/blob/main/pkg/llm-prompts/skills/vsm-loop/content.md

Comments URL: https://news.ycombinator.com/item?id=48210610

Points: 1

# Comments: 1

New comment by pshirshov in "Apple Silicon costs more than OpenRouter"

pshirshov — Sun, 17 May 2026 19:33:03 +0000

... but comes with privacy guarantees *

* but Apple will collect all your keystrokes anyway

New comment by pshirshov in "Omnisearch – A lightweight metasearch engine written in C"

pshirshov — Fri, 15 May 2026 17:57:37 +0000

I can't figure out what's that from the readme. Am I supposed to read the code and infer the ways to use it?

New comment by pshirshov in "Building ML framework with Rust and Category Theory"

pshirshov — Fri, 15 May 2026 10:23:07 +0000

> Rust

> No HKTs

> Category Theory

New comment by pshirshov in "RISC-V Router"

pshirshov — Thu, 14 May 2026 22:13:06 +0000

But the fundraising goal is.

New comment by pshirshov in "RISC-V Router"

pshirshov — Thu, 14 May 2026 21:27:10 +0000

> Router

> Ethernet: 1 WAN Gb, 1 LAN Gb

> $250000

Awesome.

New comment by pshirshov in "Reimagining the mouse pointer for the AI era"

pshirshov — Wed, 13 May 2026 11:45:14 +0000

Could we play bullshit bingo?

Bullshit!

You don't need any new metaphors to support such (questionable) flows - at all.

Swipes instead of selection rectangles are annoying - you don't see the traces of the swipes on these demo gifs! So, you've effectively "selected" something - but you have to keep in mind WHAT you selected.

Total ridiculous bullshit.

New comment by pshirshov in "If AI writes your code, why use Python?"

pshirshov — Tue, 12 May 2026 14:08:14 +0000

They tend to ignore such instructions on first circular issue - even with Opus you have to kick it really hard, insist on generalization and intervene manually. In my opinion this is not a productive/workable approach for large projects.

Typical failure mode: "I fix pyright error A, it causes pyright error B, pyright is broken, I will exclude both A and B through pyright config and will add ignore annotations for both A and B and will write a couple of idiotic comments about that".

New comment by pshirshov in "If AI writes your code, why use Python?"

pshirshov — Tue, 12 May 2026 13:14:34 +0000

I've tested many flows involving linters. Results are far from ideal - agents tend to work around linters, mass-add ignore annotations, etc, especially in situations when fixing one warning/error triggers another (and that happens regularly).

New comment by pshirshov in "If AI writes your code, why use Python?"

pshirshov — Tue, 12 May 2026 12:33:35 +0000

Models are trained on human data and they make exactly the same mistakes - and a lot of mistakes humans don't usually make.

Also they are extremely bad at high-level design.

New comment by pshirshov in "If AI writes your code, why use Python?"

pshirshov — Tue, 12 May 2026 12:17:31 +0000

Well, Python is not the only language with a mature ecosystem.

Also, in many cases it's cheaper to rewrite a small lib instead of fighting crappy code - but that applies regardless of the target language.

New comment by pshirshov in "If AI writes your code, why use Python?"

pshirshov — Tue, 12 May 2026 10:57:42 +0000

No reason, unless the project is simple. The more you can offload onto your compiler/typer - the shorter is the feedback loop, the better agents work.

Lack of strictly enforced static typing make agents fail much sooner with Python. In my opinion, Rust and Scala are the best targets for agentic flows - and, coincidentally, they have the most advanced typers among mainstream languages.

But any statically typed language behaves better than any dynamically/duck typed language. When I say "better" I mean delivery time and the amount of shipped defects.

Another thing which helps (but not generally applicable) - ask your agent to verify critical protocols with formal proof in TLA+/lean/coq. Agents are bad at formal proofs - but generally are much better than most of the humans.

New comment by pshirshov in "Venom and hot peppers offer a key to killing resistant bacteria"

pshirshov — Mon, 11 May 2026 19:14:21 +0000

Same thing - I do not know. It got pretty decent media coverage back then - and nothing since.

New comment by pshirshov in "Venom and hot peppers offer a key to killing resistant bacteria"

pshirshov — Mon, 11 May 2026 13:28:46 +0000

"Researchers have developed", yeah. When I read such things, I always recall https://en.wikipedia.org/wiki/Epimerox - this thing promised wonders - very broad spectrum, very low toxicity - and, most importantly, it was targeting a conservative essential protein - so nearly zero resistance. And there were no updates for more than a decade.

Something developed in a lab is something we, most likely, will never see - and will never know why the thing didn't reach the 2nd stage (or the 1st).

New comment by pshirshov in "My first in-prod corrupted hard drive problem"

pshirshov — Fri, 08 May 2026 23:33:10 +0000

ZFS and ECC do protect against silent page corruption that gets faithfully replicated to every mirror.

New comment by pshirshov in "My first in-prod corrupted hard drive problem"

pshirshov — Fri, 08 May 2026 20:28:23 +0000

So, you were not using a striped mirror ZFS for a prod database? What could go wrong, yep.

New comment by pshirshov in "AI slop is killing online communities"

pshirshov — Fri, 08 May 2026 08:17:44 +0000

This one doesn't work for a long time.