a13o

Modern cynicism. Specifically the form you see on social media, where road-raging cult-like bubbles imagine post-truth to be the great explainer of their misery.

I used one well-known phrase, ‘Very Online’. Germans also coined the lovely ‘Hobbyless’ (Hobbylos). And then the catch phrase directed at these types, ‘touch grass.’

Your instinct is right that most people don’t hyper focus on this world view. But the training sets can’t get transcripts from the real world at scale, so they over-index on the more easily scrapable neuroses of the hobbyless.

Train your model on Very Online corpuses, ask Very Online questions, get Very Online answers

As long as frameworks and languages offer some novel eval to the agents, they’ll remain advantageous for certain projects. Rust is the most obvious case, giving the agent a borrow checker.

I think the frameworks with strong batteries-included conventions also offer unique guardrails for agents building more ambitious apps. Those conventions allow for highly opinionated linters and well-worn migration paths, which become an eval advantage.

Eventually we will develop new languages and frameworks that are tailored to agents maintaining projects on epic time horizons. Until then, we can squeeze free evals out of these existing convention-laden technologies.

This doesn’t have an interesting footprint on the intelligence/cost Pareto line compared to existing Opus 5.5 and GPT-6 models.

Ergonomics for hand composition are less relevant now. The most important trait is how does a long-lived project perform when written by a coding agent.

I’m curious if anyone has any takes on the different approaches to agentic styling.

I’ve tried Tailwind for this without further steering, and after a while the agent starts adding unnecessary utility classes everywhere. Then I taught it to micro-Ralph eval all its utility classes using screenshots and this works ok but burns a lot of tokens. Treats the symptom and not the problem IMO.

Maybe with the right set of constraints there’s a subset of Tailwind that composes better? I went back to vanilla CSS for now, on the hope that it will readily port to whatever the best thing ends up being later.

IMO we won’t see AGI until the models are sitting inside bodies with senses and gut biomes and opposable thumbs and the like, all influencing the context window in real time. Intelligence is a full-bodied experience

According to one poll, 75% of them will, yes

https://wtop.com/lifestyle/2025/08/what-kids-say-it-would-ta...

I had a lot of fun. I think the progression curve gets boring around Boss Level 11. Introducing a prestige system could extend the runway.

I thought both Factor and Divide were interesting twists to the battle system, but neither were given enough room to breathe. With different boss numbers in the early progression, Factor can be tuned to make battles winnable that otherwise aren’t. That saves Divide for later in the progression. As it is now, they both surface around the same time and crowd each other’s debut.

It’d also be good to highlight new developments by making them unlockable in the Clubs skill tree. E.g. make an unlock or unlocks for “2-star numbers auto attack”. And “Unlock Factor where factorials of boss numbers get huge stat increases.” It mostly gives you a way to spell out and educate new mechanics, but also it pads out the Clubs skill tree which is a bit thin comped to other games like this.

I hope you keep iterating on it, it’s a really good game!

From a non-dual perspective either nobody is a p-zombie or everybody is a p-zombie. I think the trajectory of LLMs will be that consciousness sensations are abundant and low value, and that will give all the non-dual territory over to the illusionists (who now have a tangible example to point to.) Emergence won't be disproven but it'll be about as interesting as a penny stock. Everyone else is going to go to ground as a dualist and argue for some entirely new unreachable aspect of human exceptionalism, but they'll have ceded qualia as they did for souls. If I had to guess, they'll focus on embodied cognition because building high-fidelity bodies still seems really hard.

It’s going to be a long road, but I think as LLMs and their offspring create more and more convincing arguments for silicon consciousness we will conclude that consciousness is about as real as humours, and we’ve all been p-zombies this whole time.

Maybe the literary creature shoe should have started on the other foot, and sent us in search of proof that we are or are not p-angels. That at least puts the burden of proof on the compatibilists where it belongs.

TV has the “puzzle box” genre and this is a much better fit for what Tunic is

Snakebird is my vote for best puzzle game of all time

This would have been great back when I used a search engine to visit web pages.

I think of it like a technology checkpoint. Make sure you got as far as everyone else when they gave up, so when the next innovation in that space comes along you can start back up on even footing.

You want to have your own pathway to production that dodges competitors’ patents, is somewhat defensible itself, maybe a brand, etc.

Listed in the article are the National Society for the Prevention of Cruelty to Children and the Internet Watch Foundation, which monitors and removes child sexual abuse material from the internet.

The recent Meta lawsuits also mention opposition from the National Center for Missing and Exploited Children and Meta's own executives: Monika Bickert (head of content policy) and Antigone Davis (global head of safety). Both executives mention the danger end-to-end encryption poses to children when attached to a social media graph.

https://www.reuters.com/legal/government/meta-executive-warn...

If you’re worried about fitting the window, make a RAG holding an AST transformation of your codebase

I’ve given these sorts of haystack search struggles completely over to the LLM. Whether it’s finding a bug in code or searching documentation for the answer, I view it as a near-obsolete skill. The past couple of decades it was important to know how to chase down documentation and zero in on the one line of config you were missing. Now it’s not.

I don’t view this as a hollowing out of my skill tree, I view it as freeing myself to focus on modern skills I need to develop. Such as learning how to steer LLM context windows towards maintainable solutions in large codebases.

I’m sure I’ll be thankful now and then that I know how to manually sift through stack traces for answers. But I expect those moments to be rarer and rarer. I basically never look at machine code, but I bet that used to be an important skill for programmers many decades ago.

I liked the results of vectorizer.ai and recraft.ai

Input image is important too. When working with the generalist LLM on the raster art, give it context that you are making a logo, direct it to use strokes and fills and minimal color palette, readable at small sizes, etc.

If you search for ‘vectorization AI’ there are a handful of specialized tools and apis that can do it. It worked well for a handful of logos I wanted to convert. Nano banana generated the raster logos, and these other tools vectorized them

Atomic commits compose easier. In case you want to pull a few out to ship as their own topic. Or separate out the noisy changes so rebases are quicker. Separate out the machine-generated commit so you can drop it and regenerate it on top of whatever.

My commit messages are pretty basic “verbed foo” notes to myself, and I’m going to squash merge them to mainline anyway. The atomic commits, sometimes aided by git add -p, are to keep me nimble in an active codebase.

I wouldn’t say A Theory of Fun is “the book.” It’s more a coffee table read. “The book” is Jesse Schell’s The Art of Game Design

Most people don’t stay. They burn out and find work/life balance in other fields.

Once they realize how depressed their wages were in the games industry there’s no hope of getting them back.

I don’t think volume of engagement is the main issue with social media. Rather, it’s the scope of access. Social media exposes us to too many people and we forget their humanity. Instead of information spreading across the globe along a lattice of trusted relationships, it teleports through bias-confirming wormholes.

Funny enough, the time zone restriction acts as a crude proxy for locality and slightly scratches the itch; more than the time window does.

It’s a start. I’m bummed at how narrowly scoped this is. When the RFC period was open I wrote in to highlight how apartments charge surprise pet rent fees that don’t appear until the application process.

> How are you supposed to know if you have milk that will quickly go bad etc.?

The milk will say "Best if Used By <date>".

This information can be found in the fifth sentence in the article.

The wording in the fraud cancellation emails gave me a good laugh.

"My payment provider said you used a stolen credit card. Why did you do that? I've revoked your access."

What happens now? We unbundle search results? Sounds cool!

Self-identifying as an introvert or an extrovert is a limiting belief. Free yourself from this pseudo-science garbage.

I replaced breakfast with a bottle of Huel six years ago, and I don't see a reason to stop. I still cook for lunch and dinner and I skip the Huel on occasion to brunch with friends. No need to be a zealot about these things.

I do keep some of the meal bags in the pantry for emergencies. I've also replaced lunch with them at times to count calories and lose weight. I think these are good tools to have in your kitchen. They have characteristics that no other food option has, and that makes them valuable.

I played a party game where you had to describe surviving a deadly scenario ("your car went off a bridge") and a LLM would decide if your answer would work or not. A few rounds in we found the best strategies where answers like:

I escape happily. I do not perish.

There's a small blocklist of obvious words like 'survive' and 'die'; but once you get blocked on those, it's a tell that this strategy will work with the right unblocked synonyms.

Basically if you ever find yourself adversarial with a LLM, figure out The Game and directly subvert it. There's no amount of propositions that can prepare it for human ingenuity at the meta level.