Hacker News: Philpax

New comment by Philpax in "One year of Roto, a compiled scripting language for Rust"

Philpax — Sun, 31 May 2026 12:46:21 +0000

Hot-reloading. You can edit your logic without rebuilding and restarting the host application; this cuts your iteration time from minutes to seconds, especially if the application is in a state that would need to be recreated.

New comment by Philpax in "Zig: Build System Reworked"

Philpax — Sat, 30 May 2026 15:21:45 +0000

What kind of scale are you thinking of?

New comment by Philpax in "Zig: Build System Reworked"

Philpax — Sat, 30 May 2026 15:16:20 +0000

I would argue that memory safety is part of devex: it's just one less thing you have to be constantly vigilant about.

New comment by Philpax in "Zig: Build System Reworked"

Philpax — Sat, 30 May 2026 14:53:24 +0000

I would be very surprised to see a large Rust codebase being harder to maintain than a large Zig codebase. The former makes it much easier to maintain invariants at scale.

New comment by Philpax in "Nitpicking the shell history scene in 'Tron: Legacy'"

Philpax — Thu, 28 May 2026 23:11:39 +0000

The first Matrix is 27 years old; Reloaded is 23 years old.

New comment by Philpax in "Claude Opus 4.8"

Philpax — Thu, 28 May 2026 17:51:50 +0000

It wouldn't be data distillation: instead, it would be teacher-student distillation. The teacher model has stronger representations that the student can mimic, which would give it more capability over training on the data itself.

New comment by Philpax in "Claude Opus 4.8"

Philpax — Thu, 28 May 2026 17:41:57 +0000

It's a refutation that we know how they work now. In the limit, though, yes, we are likely to be able to trace the process: it is possible, though, that understanding remains inaccessible because the trace is beyond comprehension.

If you can distil the model's reasoning for a decision into a billion yes/no questions, each covering largely-independent areas, can you really say you understand what its overall reasoning was?

New comment by Philpax in "Claude Opus 4.8"

Philpax — Thu, 28 May 2026 17:04:50 +0000

Could you honestly tell us what model you're on? I'm guessing Sonnet 4.6 or Opus 4.7.

New comment by Philpax in "Claude Opus 4.8"

Philpax — Thu, 28 May 2026 17:03:01 +0000

We do not understand and know how these models work. We know what their architectures are and how to create them, but we cannot explain their behaviours at a fundamental level. There is no definitive way for us to answer the question of "how did it produce response X for query Y?" - we're only grazing the surface with mechanistic interpretability.

New comment by Philpax in "Claude Opus 4.8"

Philpax — Thu, 28 May 2026 17:01:23 +0000

Okay, but that's not relevant to AI training?

New comment by Philpax in "Claude Opus 4.8"

Philpax — Thu, 28 May 2026 16:58:46 +0000

AI is grown, not built, and like with anything you grow, you'll never be able to predict exactly how it will turn out.

New comment by Philpax in "Anthropic Cofounder Chris Olah's Remarks on Pope Leo XIV's "Magnifica Humanitas""

Philpax — Mon, 25 May 2026 20:10:37 +0000

His argument is not that the existing global poor are going to be automated by AI, but that a great many people are going to join the global poor as their current livelihoods are automated.

Anthropic Cofounder Chris Olah's Remarks on Pope Leo XIV's "Magnifica Humanitas"

Philpax — Mon, 25 May 2026 19:12:29 +0000

Article URL: https://www.anthropic.com/news/chris-olah-pope-leo-encyclical

Comments URL: https://news.ycombinator.com/item?id=48270497

Points: 87

# Comments: 99

New comment by Philpax in "Leave Me Behind"

Philpax — Mon, 25 May 2026 17:36:26 +0000

You won't be left behind because you don't know how to use them - as you say, it doesn't take that much time to be at useful proficiency - but you will be left behind if you refuse to engage with them at all.

New comment by Philpax in "A Wayland Compositor in Minecraft"

Philpax — Fri, 22 May 2026 22:50:05 +0000

Yes, but part of the fun is doing it in Minecraft and using Minecraft's language for it (e.g. putting windows in your inventory, pulling them out of chests, etc)

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

Philpax — Thu, 21 May 2026 22:24:19 +0000

Huh? That's them intentionally demonstrating the slop style.

New comment by Philpax in "What's in a GGUF, besides the weights – and what's still missing?"

Philpax — Thu, 14 May 2026 23:20:44 +0000

Quantiser's choice, IMO. They're best-placed to decide what compromise to make for their particular model.

New comment by Philpax in "What's in a GGUF, besides the weights – and what's still missing?"

Philpax — Thu, 14 May 2026 22:00:35 +0000

I regret that the projection models ended up separate, and I too would have preferred for them to be in a single file. I'm not entirely sure why that ended up happening, but it very much runs counter to the single-file ethos I had in mind when I designed GGUF.

Hoping that someone will shepherd the cause of merging the two; I think I'm too out of the loop to do it this time around :-)

New comment by Philpax in "What's in a GGUF, besides the weights – and what's still missing?"

Philpax — Thu, 14 May 2026 21:58:17 +0000

Single-file deployments were an intentional design goal on my part. While most image models were/are single-file, LLM safetensors (at least at the time) were not, and I wanted to ensure that we enforced that at a structural level. I also didn't want to mandate a JSON reader for executors (e.g. llama.cpp), which the ST approach would have required. The bigger issue at the time, if I recall, was that ST couldn't support the new-and-upcoming quants that GGML had, and having our own file format offered us flexibility that ST couldn't.

New comment by Philpax in "What's in a GGUF, besides the weights – and what's still missing?"

Philpax — Thu, 14 May 2026 21:54:28 +0000

Yeah, I intentionally left space for the computation graph to be included in the GGUF spec in the hopes that this would be picked up by someone. I would have loved to have it in the first version, but I was prioritising getting the MVP spec out and implemented.

I'd still love to see this, but it would need a cheerleader very familiar with the current state of the GGML IR.