Hacker News: stabbles

New comment by stabbles in "Our eighth generation TPUs: two chips for the agentic era"

stabbles — Wed, 22 Apr 2026 23:20:49 +0000

Other than TPUs they're also planning for 960,000 Rubin GPUs [1] which can do 33 teraflops fp64 each, so over 30 classical exaflops, and with emulation it could be more than 100 exaflops.

[1] https://blogs.nvidia.com/blog/google-cloud-agentic-physical-...

New comment by stabbles in "AI may be making us think and write more alike"

stabbles — Tue, 07 Apr 2026 14:36:55 +0000

Another one I noticed is "or maybe I hallucinated that" instead of "or maybe I dreamed that". Researchers will be horrified to learn that even talk about LLMs affects people's vocabulary.

New comment by stabbles in "AI may be making us think and write more alike"

stabbles — Tue, 07 Apr 2026 12:48:01 +0000

Oh no, LLMs threaten our individuality ⸻ what will we do?!

New comment by stabbles in "I am definitely missing the pre-AI writing era"

stabbles — Mon, 30 Mar 2026 09:49:36 +0000

Are grammatical errors and typos fashionable now? Reading this post it seems the anti-thesis in the LLM era is not to edit at all, but rather write down a stream of consciousness to make it "personal".

New comment by stabbles in "The RISE RISC-V Runners: free, native RISC-V CI on GitHub"

stabbles — Sun, 29 Mar 2026 18:09:32 +0000

My experience with RISC-V so far is that the chips are not much faster than QEMU emulation. In other words, it's very slow.

New comment by stabbles in "Epoch confirms GPT5.4 Pro solved a frontier math open problem"

stabbles — Tue, 24 Mar 2026 09:31:25 +0000

You're glancing over the fact that mathematics uses only one token per variable `x = ...`, whereas software engineering best practices demand an excessive number of tokens per variable for clarity.

New comment by stabbles in "Liberated Systemd"

stabbles — Sat, 21 Mar 2026 10:30:05 +0000

The other day someone commented on this site that in the age of agentic coding "maintaining a fork is really not that serious of and endeavor anymore." and that's probably the case. I'm sure continuously rebasing "revert birthday field" can be fully automated.

Then the only thing remaining is convincing a critical mass that development now happens over at `Jeffrey-Sardina/systemd` on GitHub.

New comment by stabbles in "Astral to Join OpenAI"

stabbles — Fri, 20 Mar 2026 07:22:16 +0000

That's overstating things. The biggest piece of infra is PyPI, to which uv is only an interface. They do distribute Python binaries, but that's not very impressive.

New comment by stabbles in "World Happiness Report 2026"

stabbles — Thu, 19 Mar 2026 17:19:39 +0000

https://files.worldhappiness.report/WHR26.pdf

New comment by stabbles in "World Happiness Report 2026"

stabbles — Thu, 19 Mar 2026 17:19:21 +0000

You would think that Finland's unemployment rate (10%+) would influence its ranking, but that's not the case at all.

New comment by stabbles in "Python 3.15's JIT is now back on track"

stabbles — Wed, 18 Mar 2026 09:11:28 +0000

You really need dedicated types for `int64` and something like `final`. Consider:

    class Foo:
      __slots__ = ("a", "b")
      a: int
      b: float

there are multiple issues with Python that prevent optimizations:

* a user can define subtype `class my_int(int)`, so you cannot optimize the layout of `class Foo`

* the builtin `int` and `float` are big-int like numbers, so operations on them are branchy and allocating.

and the fact that Foo is mutable and that `id(foo.a)` has to produce something complicates things further.

New comment by stabbles in "Python 3.15's JIT is now back on track"

stabbles — Wed, 18 Mar 2026 08:00:30 +0000

That was how the Mojo language started. And then soon after the hype they said that being a superset of Python was no longer the goal. Probably because being a superset of Python is not a guarantee for performance either.

New comment by stabbles in "Grace Hopper's Revenge"

stabbles — Tue, 17 Mar 2026 10:13:49 +0000

The TL;DR: code should be easy to audit, not easy to write for humans.

The rest is AI-fluff:

> This isn't about optimizing for humans. It's about infrastructure

> But the bottleneck was never creation. It was always verification.

> For software, the load-bearing interface isn't actually code. Code is implementation.

> It's not just the Elixir language design that's remarkable, it's the entire ecosystem.

> The 'hard' languages were never hard. They were just waiting for a mind that didn't need movies.

New comment by stabbles in "Stop Sloppypasta"

stabbles — Sun, 15 Mar 2026 22:28:27 +0000

I wouldn't call "ChatGPT says" an equivalent of LMGTFY. The former is people in awe with the oracle, the latter is people tired of having to look something up for others.

New comment by stabbles in "Redox OS has adopted a Certificate of Origin policy and a strict no-LLM policy"

stabbles — Tue, 10 Mar 2026 10:06:41 +0000

For well-intended open source contributions using GenAI, my current rules of thumb are:

* Prefer an issue over a PR (after iterating on the issue, either you or the maintainer can use it as a prompt)

* Only open a PR if the review effort is less than the implementation effort.

Whether the latter is feasible depends on the project, but in one of the projects I'm involved in it's fairly obvious: it's a package manager where the work is typically verifying dependencies and constraints; links to upstream commits etc are a great shortcut for reviewers.

New comment by stabbles in "The death of social media is the renaissance of RSS (2025)"

stabbles — Mon, 09 Mar 2026 10:18:49 +0000

Sure, naming is important, but the RSS icon was well known. It was part of the Firefox address bar.

New comment by stabbles in "A decade of Docker containers"

stabbles — Sat, 07 Mar 2026 21:52:31 +0000

Does Nix do one layer per dependency? Does it run into >=128 layers issues?

In Spack [1] we do one layer per package; it's appealing, but I never checked if besides the layer limit it's actually bad for performance when doing filesystem operations.

[1] https://spack.readthedocs.io/en/latest/containers.html

New comment by stabbles in "Package Managers à la Carte: a formal model of dependency resolution"

stabbles — Sat, 28 Feb 2026 12:01:40 +0000

It's not just that, it's also a filesystem layout issue. If you install everything in `/usr` or `/lib/pythonX.Y/site-packages` you cannot have two versions / variants of the same package installed concurrently.

For that you need one prefix per installation, which is what Nix, Guix, and Spack do.

New comment by stabbles in "Julia: Performance Tips"

stabbles — Fri, 27 Feb 2026 11:48:23 +0000

There was a nice talk recently at FOSDEM about Reactant and Oceananigans.jl: https://fosdem.org/2026/schedule/event/TYF8ZD-accelerating_s...

New comment by stabbles in "Julia: Performance Tips"

stabbles — Fri, 27 Feb 2026 09:35:06 +0000

Back in the day I had trouble convincing my C++ friends to give Julia a try, because Julia's garbage collector was a showstopper. But if you follow these performance tips related to pre-allocation, in-place mutation, avoiding temporary allocations (and maybe avoiding cycling references), you don't struggle with GC performance issues.

Looking back, I think the tooling Julia had from the start, combined with the REPL, made it actually really nice to fix these performance issues. Much better than compiling and linking a binary and running it through some heap profiler tool. You could simply do

    julia> @time f()
      x.y seconds (N allocations: M bytes)

and iteratively improve part of the code base instead of profiling the entire application.

(To be fair: back then escape analysis was not implemented in the compiler, and it was hard to avoid silly allocations)