Hacker News: xpct

New comment by xpct in "Why thinking out loud with someone beats thinking alone"

xpct — Wed, 17 Jun 2026 21:43:14 +0000

2017 is a bit early to refer to them as LLMs. I'm not sure when exactly we started to refer to LMs as 'large', but I don't think it was before GPT2 (2019). That said, from the NLP work I've done, it was much more interesting working on small specialized models.

New comment by xpct in "A robot is sprinting towards you. Do you want it running on Claude or Grok?"

xpct — Wed, 17 Jun 2026 21:35:06 +0000

Those short sentences are also of the X hype account cadence, though they've fully embraced LLM text by now

New comment by xpct in "Mechanical Watch (2022)"

xpct — Wed, 17 Jun 2026 03:33:18 +0000

This blog remains one of the most inspirational interactive websites for me. It really nails both clarity and remains interactable, fun. When (if ever) LLMs are able to generate interactables like these, is when I will hang up my coat.

New comment by xpct in "Nobody clicks your share buttons"

xpct — Tue, 16 Jun 2026 07:33:24 +0000

And here I thought I was an oddball for manually cutting off end parts of links

New comment by xpct in "No, everyone is not using AI for everything"

xpct — Sun, 14 Jun 2026 16:44:30 +0000

Neither, just an ad

New comment by xpct in "Not everyone is using AI for everything"

xpct — Sun, 14 Jun 2026 16:35:08 +0000

> Not having used agents and not being able to comment on what to do and what not to do with them is immediate out since early this year.

From all the tech that we have, agents are really not that hard to learn on the job. They're also not a magical silver bullet.

New comment by xpct in "Not everyone is using AI for everything"

xpct — Sun, 14 Jun 2026 16:31:48 +0000

I understand the pressure to get employed from your perspective, but differences in opinion should be voiced out and typically aren't the thing leading to rejection from the company. It's common that engineering leads seek out people with different backgrounds and views to work on the same team. If anything, answering truthfully will make you stand out from others who've responded in a generic, heavily hedged way.

New comment by xpct in "Ask HN: What internal tool did you build that became a product?"

xpct — Sun, 14 Jun 2026 04:40:30 +0000

I loved the 'Build Software Like It Needs To Last' one. I feel like it's something that could be made into a poster and hanged where I could see it first thing in the morning.

New comment by xpct in "Shepherd's Dog: A Game by the Most Dangerous AI Model"

xpct — Sat, 13 Jun 2026 10:14:41 +0000

It seems to me that most media genres discover the most interesting parts relatively early, then most subsequent work is deeply derivative. I feel that way about video games, digital music, movies.

I'd wager it's because ideas are simpler to explore orthogonally, giving an overview of what's possible.

New comment by xpct in "Statement on US government directive to suspend access to Fable 5 and Mythos 5"

xpct — Sat, 13 Jun 2026 02:20:18 +0000

As someone who's also worried about delegating too much thinking to LLMs, I wonder if letting your own citizens use the good models is detrimental.

New comment by xpct in "Statement on US government directive to suspend access to Fable 5 and Mythos 5"

xpct — Sat, 13 Jun 2026 01:54:58 +0000

Jokes on you, we're releasing the new, 'more efficient', 'less intelligent', Capybara 5 model. It's been 'reprogrammed' to only score 49.8% on the 'PyTorch basics' benchmark!

New comment by xpct in "Statement on US government directive to suspend access to Fable 5 and Mythos 5"

xpct — Sat, 13 Jun 2026 01:47:51 +0000

I've actually not thought about deployments in remote jurisdictions that much. I also don't think the models are dangerous enough to warrant it, but do you reckon the big labs have plans thought out for deleting remote model copies, such that they couldn't be scrubbed off cold NVMEs?

New comment by xpct in "US Government directive to suspend access to Fable 5 and Mythos 5"

xpct — Sat, 13 Jun 2026 01:43:43 +0000

Likely more surveillance when it comes to electricity expenditure.

New comment by xpct in "US Government directive to suspend access to Fable 5 and Mythos 5"

xpct — Sat, 13 Jun 2026 01:38:34 +0000

That would depend on what gets leaked, as I'm not so sure that the weights by themselves would be enough to replicate the architecture. I imagine some part of the secret sauce will remain in the architecture, and the tensor dimensions may not be enough to decode it.

I'm sure if proprietary models continue to be a big thing, the methodology of their storage and loading on hardware will be obfuscated quite a bit.

New comment by xpct in "Malware developers added nuclear and biological weapons text to to their spyware"

xpct — Fri, 12 Jun 2026 23:02:25 +0000

Oh cool, haven't heard of these before. Unfortunately strings like that can just be sed'd out.

New comment by xpct in "Malware developers added nuclear and biological weapons text to to their spyware"

xpct — Fri, 12 Jun 2026 23:00:21 +0000

Actually, even Opus 4.8 completely switched off on me and suggested Haiku when I asked about today's Arch Linux AUR malware.

New comment by xpct in "Mmorpg World of ClaudeCraft, vibe coded with Fable 5"

xpct — Fri, 12 Jun 2026 21:55:33 +0000

This was made in 2 days and 91% of the Max 20x plan, as the author stated on the Reddit thread, so roughly ~$200. Supposedly, existing free assets were used and weren't generated.

I'd say demos like these stand to profit the most from LLMs, if the goal is to make as much as possible in a few days: a barrage of quests are easy to generate, so are gear choices, and some skills for the initial 9 classes to pick from. A human would generally spend a lot of time here, thinking about whether the class/skill choices fit their world, what type of progression is fun and isn't. It's also where player testing would be important for a game to set good pacing and balance the difficulty.

Of course, the game itself is barely playable, it randomly stutters when I walk too far away from camp, the character controls are unintuitive, etc. A lot of this stuff could be chipped away by spending more time on the project and testing it yourself, getting a feel for what you want the game to be. That by itself should require a game to take more than a few days, if we expect others to play it and enjoy it. Something simple like movement controls could take many game iterations to iron out, and those aren't hard technical tasks.

Still, I can't entirely wrap my head around the fact that I live in a world where a machine can create this with minimal intervention by humans, and do a somewhat OK job at it, to the point where I'm willing to spend 10 minutes playing it.

New comment by xpct in ""Don't You Just Upload It to ChatGPT?""

xpct — Fri, 12 Jun 2026 21:29:23 +0000

I propose we move away from the framing of "Model years" - they're standard human research years. Yes, likely more people are working on it, and also working harder, but ever since we acquired a certain amount of compute in the world, many people were able to independently find the same patterns and train models.

New comment by xpct in "How Terry Tao became an evangelist for AI in math"

xpct — Fri, 12 Jun 2026 13:11:01 +0000

I'm sure a system that could do this is economically optimal.

Why are you looking forward to this, though?

New comment by xpct in "If you are asking for human attention, demonstrate human effort"

xpct — Fri, 12 Jun 2026 13:03:38 +0000

I hope he's thought out his next vocation, since he's so eager to automate his current one.