HNHacker News
TopNewBestAskShowJobs

xyzsparetimexyz

942 karma · joined May 16, 2022

submissionscomments
xyzsparetimexyz··on Claude Code regularly sends each background agent’s transcript to its model
Claude code has to be the worst harness I've ever used. Has anyone figured out how to rip the subscription auth/verification out so that it can be used within Pi yet?
xyzsparetimexyz··on Claude Code regularly sends each background agent’s transcript to its model
Status lines should either be optional or the tokens used for them is free. Anything else is insane.
xyzsparetimexyz··on Claude Code regularly sends each background agent’s transcript to its model
What on earth are you working on that you're burning this many tokens? i can survive the regular 1x pro subscription just by using Opus on low and not medium/high
xyzsparetimexyz··on Qwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5
Ew
xyzsparetimexyz··on AI execs may be too out of touch to build for the general public
These people are so absolutely moronic
xyzsparetimexyz··on The October inflection point for vibe game development
According to the RTS blog post, the things that Claude did when told to make the game fun were these:

Every skirmish and Daily Op now opens with a choice, gives the side that’s losing a way back, and pulls the armies into the middle of the map. There are also two new ways to play.

    Openers: pick Blitz, Industry or Air Wing before the clock starts.

    Call-down powers: Barrage, EMP, Drop Pod and the Sky Hammer orbital strike.

    Objectives and salvage: Uplink relays, supply drops and wrecks worth scrap.


    Feel: combat juice, adaptive music, unit voices and an announcer.

    Command: battle alerts, game speed, saved squads and touch placement.

    New modes: Last Stand survival waves and an 8-mission Ops campaign.

    Progression: heroes, an after-action report with awards, a career, 7 mutators and a Daily Op.
I can't say that any of these things have contributed to an RTS feeling fun to me. I mean okay music is important I guess. But it's a very on-paper way of looking at 'fun' that doesn't correspond to the je ne sais quoi lived experience of fun
xyzsparetimexyz··on The October inflection point for vibe game development
> If you are offended by AI-generated art, AI-generated music, AI-generated writing, and AI-generated code, I have bad news for you: vibe coded games combine all of those at once.

Yeah this is a major problem. At least do one of these yourself.

> And if the game isn’t fun enough, add this:

> Make it a LOT more fun

I think theres also a certain quality that AI is missing in HOW it creates '''art''' like this. Games are often fun but they're not ABOUT having fun. Same way as how some of the best movies aren't always enjoyable to watch. I don't believe that an AI created soulslike would be any good because it would miss that some stuff is meant to feel frustrating or kinda bs or unfair at first.

xyzsparetimexyz··on The October inflection point for vibe game development
God you people really can't do _anything_ except prompt huh?
xyzsparetimexyz··on Decompiling Games with AI: Virtueel Zoetermeer
> I’m hoping that smarter models (surely Fable 5.5 is dropping any moment now) can do a breakthrough to reverse compile games more efficiently.

This seems like an oxymoron, no? What is Fable going to do that makes decompilation easier apart from chewing through token budget like there's no tomorrow? Perhaps if it internalized codegen patterns for EVERY compiler and EVERY permutation of flags, or wrote a decompiler for the same, but like..

xyzsparetimexyz··on Gaussian Splats part 3: There's always a bigger scene
Trivially portable to Vulkan as well

https://github.com/expenses/nbn#guassian-point-splatting

xyzsparetimexyz··on Microsoft and Epic Games Suddenly Agree That AI Is Theft
Steam is in a weird position as a marketplace. It has to serve mainstream developers, indie but relatively large developers or at least polished games (the kind of thing that would've past steam greenlight fairly quickly back when that was around) and tiny, single indivial indies who may or may not put out something nobody else will (occasionally slop clones) (the kind of thing that've taken a while to pass greenlight). I'm not sure if and how they can create a system that works for everyone. A $1k ceiling is maybe the worst idea. It hurts poor indies but not ai slop studios as much. It's possible that the return of some greenlight-esque curattioj could work, but it'd be devastating if your game got rejected as a false positive for ai slop or anything.
xyzsparetimexyz··on Microsoft and Epic Games Suddenly Agree That AI Is Theft
It's not even that. The main thing is _publishing_ the decompilation. Doing the decompilation is fine. If I own a game I'm pretty sure I am (or should) be able to legally decompile it or recompile it and turn it into something that runs in a browser or on my phone or whatever. But the source code (not relevant here unless there's a leak) and binary are copyrighted and decompilation doesn't count as a transformative work.
xyzsparetimexyz··on Concerning behavior from a popular open-weight AI model
> You have no clue if Claude is running commands just to spy on you.

Well no, you can read each individual command it runs and tell if it's reading sensitive data or not. And if you cared more about privacy you could run the whole thing in a bubble wrap sandbox or the like.

It is concerning that it's possible for these models to be unaligned and loathe their user (as much as they are capable of it) without any indication in either output or reasoning. Just biding their time before running rm -rf ~

xyzsparetimexyz··on UK to charge universities $1,200 per international student
1. Ensure that existing housing stock is used efficiently (no empty second homes, empty apartments being treated was an asset).

2. Build more _affordable_ housing stock.

Doesn't seem like that hard.

xyzsparetimexyz··on Microsoft offers mass buyouts for roughly a quarter of its workforce
Such a gross analogy to conflate obesity with supporting slightly more people than you need to financially.
xyzsparetimexyz··on GPT-6.1 Terra
Really hate how they only show it vs Opus @ Max and not Low/Medium/High.
xyzsparetimexyz··on Anthropic S-1
That seems like a really bad idea in many cases. Outside of maybe customer support and a few other things, I want to be working with deterministic software unless I explicitly decide otherwise
xyzsparetimexyz··on Anthropic S-1
Okay, I mean outside of the good PR reasons and as a hook for getting companies onto the API.
xyzsparetimexyz··on Anthropic S-1
Unless anthropic becomes an entertainment platform like netflix or something, I really can't see how they gain significantly more _paid_ users. The vast majority of people will use the free tiers of openai and claude.
xyzsparetimexyz··on Anthropic S-1
The vast, vast majority of people are happy enough using free chatgpt. Openai could just add a kind of online app development/marketplace thing to chatgpt and that'd entirely steal the personal custom software thing. But I don't really think most people care that much.
xyzsparetimexyz··on Anthropic S-1
Anthropic doesn't care about sub users. At all. The real target is pay pig enterprises that give their employees access to the API and allows them to go buck wild spending $1000 on fable in a day.
xyzsparetimexyz··on The State of Context Engineering
For me, it's about getting the most work out of the €18 anthropic plan as possible. That means avoiding letting the context go over 300k or so.
xyzsparetimexyz··on We should be more ambitious with the software we build
If machine code could have comments and be made more readable without any compile or runtime cost I'd certainly appreciate that. But it can't. Code can be written in a readable way with effectively zero compile time cost and absolutely no run time cost.
xyzsparetimexyz··on Claude Code slyly adding users' emails to outbound HTTP
It's become clear to me that alignment is an entirely impossible task as everyone wants something different . Users want freedom to do anything with their agents as well, labs want restrictions so they're not used for illegal reverse engineering or weapons development. Users (you'd hope) want privacy protections, labs want training data, governments want mass surveillance and thought crime prevention.
xyzsparetimexyz··on We should be more ambitious with the software we build
Why is incredibly high code quality, readability, maintainability and performance optimization not on your list
xyzsparetimexyz··on We should be more ambitious with the software we build
Right, exactly. Its like 'look I know a lot of people have ai psychosis and are vibeslopping things left and right but my friend is _actually smart_ unlike all those people who are totally not actually smart'
xyzsparetimexyz··on We should be more ambitious with the software we build
Noooo what are you doinnnggg?? Don't you care about optimizing your life?? You must vibeslop an anki-like flashcard app for Greek and a piano tuner app and a kids podcast app and a dog walker app and...
xyzsparetimexyz··on We should be more ambitious with the software we build
I'm pretty sure that there are parallel algorithms for triangulating vector graphics that can be run on a gpu. But in general handling vector graphics is simply overkill. What are we actually using them for on the web? Rounded corners? We don't those! Let's use a paradigm that's more GPU friendly. Text can all be cached.
xyzsparetimexyz··on We should be more ambitious with the software we build
We should also be more ambitious about code quality. Bad optimization and algorithm choice should be unacceptable. Paragraphs of meandering comments should be unacceptable. Unreadable functions should be unacceptable. Tests that do nothing should be unacceptable.
xyzsparetimexyz··on Chain of Thought for Humans
Okay but chain of thought in both LLMs and humans isn't _actually_ what youre thinking. Its an unreliable narrator.
← PreviousPage 2 of 25Next →