furyofantares

That's cool.

I have agents going pretty much nonstop for hobby projects. This is great value for me because I get to see a lot of things I'm interested in come to life. It's very token-inefficient though and if usage limits were reduced much this wouldn't be worth doing.

For my actual job though, I don't let agents go for hours on end and I care about the code. I spend between $1000 and $2000 per month there. Of course we pay API pricing. I wonder how much you're getting for that $20 in API pricing, it could be 10x or even 50x. At least on the max plans it can be that high if you are hitting usage limits, I don't know about the 20.

We will use our big brains to draw bold conclusions about the most effective use of money, without the burden of conventional wisdom.

For our first bold conclusion we have determined that all of the world's money should be in the hands of people with big brains who intend to draw bold conclusions about the most effective use of money, without the burden of conventional wisdom.

This concludes the planning phase. There will be an additional planning phase once we've collected all of the money, we promise.

> About Me

> At my core, I am a creative. Whether it's games, videos, code, or music, I'm at my best when I'm creating something.

They've just handed you power tools.

There is a sadness in learning that I am less creative than I thought, that much of my self-identifying as creative came from having mastered some technical skills that used to be required to express my creativity.

Now people without those skills can express their creativity in the same domains. It's a great benefit for the world, I expect, but takes a minute to get used to the leveling of the playing field in areas I'd spent time mastering.

I won't try to convince anyone with specifics but it was definitely LLM generated. I think by someone who knows what topics they want to cover and did have something they wanted to post about. At a high level it is an infodump with almost no opinion expressed.

I learned a few things from it and have some interest in the topic. So this sort of a weird valley where I'm very annoyed by the writing but also getting something out of it. And it's more than I would have gotten out of promoting an LLM myself.

Which I know, because I have written a blog post about SDFs before, and I tried to author it with an LLM helping and the text just came out awful. I think I rewrote my entire post eventually - I am fairly confident at least that none of the LLM's words remained. I mention it because I've already read the SDF section of this blog largely, when an LLM generated mostly the same text 5 months ago.

Sorry, but it's unclear to me how that question is relevant even related to the conversation.

This is like when two years ago people said "oh yeah if these things are accelerating software development then where's all the new software??!!"

The answer then was "it still takes time, man".

That's the answer now, too!

It took time for them to accelerate the quantity of software, which is the thing these are most naturally good at, and it will take even more time for them to improve the quality of software, which has proven to be less of a superpower for LLMs than quantity is.

For my normal work I take ownership of the code, and end up with the exact code I want. I still have it go off and do a good amount of work a lot of the time, still queue up multiple tasks at the same time a lot of the time. Sometimes I throw it all away and re-prompt once it's time to commit to it, sometimes edit what it made, sometimes have it edit what it made etc.

For all of my side projects I'm full-on vibe. Well, almost: I do have opinions on what kinds of code it should write and set up my projects to get that. But I don't LOOK at the code.

I use a LOT more tokens on my side projects. I can have it working more or less constantly and it doesn't take up that much of my attention, but it is FAR less token efficient.

Oh - well the planner has a bunch of files it maintains for itself

logs/ keeps logs of various activities, including a daily unified log of my chats with it + it's messages to/from the other agnets

notes/ has notes on all the various projects, or any time i have it do research on something it'll keep notes here

reference/ similar to notes but more for content that has been pulled in externally

reminders/ stuff it's supposed to remind me about

scripts/ many different scripts it's written to do all the work it does

tasks/ open tasks

tasks/done/ completed tasks

So the PA itself isn't really a problem, it mostly just delegates to other sessions and if it has personal assistant-type work to do, it uses subagents. The per-project sessions are also meant to use subagents liberally.

But the actual answer is just one session per project and auto-compacting. Actually I run into usage limits on both codex and claude max subscriptions and so there is also a fair bit of it having to do codex/claude handoffs which is like a worse auto-compact.

I am quite sure I get worse results at a higher cost than I do for my actual day job, where I'm taking ownership of code, getting the actual code I want, and therefore doing a lot more manual context management.

But for side projects and other personal computer usage it's all vibing and auto-compacting.

The chat UI is duplicating the chatgpt/claude web chats, where you've got a list of chats on the left and each new thing you wanna do is expected to be a new chat, and maybe it has memories it writes to or can search chats to cross reference.

I've been using these things a bit differently lately, more of a personal assistant - probably more of the OpenClaw setup I'd guess, though I haven't used OpenClaw. I have my macbook running constantly and it has a session running for reach of my projects, plus one session called pa (personal assistant) which is set up to track tasks and to talk to all the other sessions.

The pa is the only session I directly interact with anymore. It's got it's own Signal account and a single Signal chat is the way I interact with all my sessions. Actually, there are a few sessions run on other machines too that it knows how to get to, but mostly it's just the per-project sessions on my macbook.

I think that's how I'm gonna want stuff like this to work - mostly in one single text thread that delegates to the others. The others don't have to be totally invisible, being able to drop into them individually would be fine as long as it doesn't interfere with the simplicity of the main thread.

Yeah, I think putting github.com in your repo is a mistake but it seems massively overstated. Seems trivial to fix all the code, or to just use go.mod, or first use go.mod and then fix your code.

I have to imagine google would consider this one to be essentially a bug - misreading user intention this badly. Akin to hallucinations.

I know many people want to ask questions to google directly and just read the AI result and very rarely check the links, and I know google wants to convert more people doing this before people switch to chatgpt etc.

But google search does not (yet?) appear to me to be designed as a chatbot you are expected to converse with where it would be normal to just open asking about some guy named Dario in your personal life who isn't globally notable. There's a difference between pushing for a conversational query and a conversational personal assistant. Heck this is a few jumps past that even.

I wouldn't really ask when google got so weird, I'd ask when google's AI overview got so shitty, and the answer is ever since they introduced it.

Well the reason I couldn't tell if you were aware is because they aren't synonyms and your comment is perfectly reasonable, but I picked up an (incorrect, it seems) vibe that you thought they were unrelated. Which, they kind of are! It's a fairly interesting story how it came about as a property of string theory but appears to be leaving it behind.

It just kept going and going and I was quite confused as to when it was going to give me any feedback, but I guess so was just getting them all right and there's no clear feedback in that case?

I can't tell if you're aware that the holographic principle came from string theory

Trying to cheat, you happen upon a place you can write notes, and you know you're part of a swarm of clones of yourself. So you reason most likely there will be others who end up in the same place, and you leave some notes, and indeed other clones of you do end up in the same place.

Do something else that day?

If someone sends me LLM output that they read and thought was fine, then that's OK with me.

They might be wrong and it's not fine, and I have a few people I know who have a much higher tolerance than I do for LLM outputs, and I'm aware of that when they send them to me and I often reject it.

But it's a totally different category than someone who only skimmed what they sent me.

If someone sends me something they read and digested and think it's good, but I don't think it is, then I'm not upset with them. If someone sends me something without that first step, then I am.

(Of course they could tell me they didn't read it, and that's fine too, I can choose to read it or not.)

Nice! It even supports COLR and SVG fonts, both of which I have options for when making a variant of my Keldritch font (https://kellydornhaus.com/keldritch/)

Is it strange that I

1) Have no interest in Harry Potter (wasn't able to get through the first book, and maybe saw two of the movies)

2) Have disdain for most of what I see from both rationalists and EAs

3) Thoroughly enjoyed HPMOR

I thought it was kind of like The Martian, but long

Also try being 7 years old again, and watching the first moon landing on live TV.

Maybe now we can stop posting the nonsense take that the frontier labs have hit a wall and are trying to distract from that for IPO reasons.

Also maybe we can stop saying "we can't slow down because China will never slow down" - I don't really think slowing down is right, BUT if slowing down is correct then maybe we should be talking about China slowing down instead of just saying "won't happen" without any evidence that Chinese labs don't have similar concerns.

It doesn't need to be bulletproof to be very valuable.

However much effort is required to fake it - it's proof that the image is either legit or that much effort went in. There's TONS of cases where it's plausible for someone to have put in the effort to fake a photo with AI (nearly zero effort required) but not remotely plausible that they set up some elaborate high quality photo of a fake.

It's also much more damning if you get caught faking it. Think of the examples where police have been caught posting altered images on social media. The lame excuse that some intern didn't realize it would do more than just upscale the image won't fly if some elaborate setup was required.

I got sick of stuff like Solitaire and Hearts and other standard playing card games requiring subscriptions or coming with ads etc in 2026 - that stuff came with your computer in my youth and got enshittified. Mainly I hate seeing my family playing all this pesterware.

So I'm making one app that has all the games I can think of without any of that crap. https://suddenlycards.com

It has online play which is P2P with a very simple service for connecting players to each other. It also has bot play, with a mix of heuristic-based bots for some games, and bots with small neural nets (of various approaches) for others, depending on what has worked best for each game.

I'm also working on my own game framework for vibing games (which the above is authored in). I have about a dozen prototypes going, and I've fully ported Spectromancer (a game I own the rights to) to it. https://nanogame.app/spectromancer

Looks like they updated the link to that page now too. Thanks.

Same price as image-2, and I'm guessing it's not more tokens (given the speedup).

allllll the way at the bottom

> Pricing and availability

> Images 2.5 is rolling out today to ChatGPT, ChatGPT Work, and Codex users across all tiers on desktop, mobile, and web.

> GPT‑Image‑2.5 Sunburst and GPT‑Image‑2.5 Flare are available in the API. See pricing details here.

and the link to the pricing details is a page where i am either too dumb to find the pricing details, or they don't exist

Linux is a marvel of the world and a large amount of its contributions have been made by people being paid to do so for commercial purposes - and I am not sure that would be true without the GPL. Is there a convincing argument there?

And outside open source there has been a staggering amount of creative work at all quality levels done for commercial purposes under the protection of copyright. Is there a convincing argument that the last century of software and music and books and etc would have been just as staggering without copyright?

I would love to hear such an argument. In my youth I thought copyright was blatantly stupid and should be abolished. I still sorta think that but I don't have a convincing argument in the face of the absolutely massive amount of good work that has been produced under it.

Sure, you can break symmetry (in this case making the decision makers not identical because they have different random number generators available), but the remaining symmetry means identical mixes must be chosen, and so a mixed strategy would only be chosen if it maximize his value for both people cooperatively.

Maybe it's a bit subtle that they said clones and I said identical decision makers; I'm letting you fill in the gap for how much clones may diverge and how much that matters.

Failing to cooperate with literal clones of yourself in a prisoner's dilemma would be a spectacular failure. There's only two things that can happen with identical decision makers: they both cooperate or they both defect. So identical decision makers who know they're identical can cross off the asymmetrical entries in the payoff matrix and the decision to cooperate becomes trivial.