TheAceOfHearts

There is a weird sense of entitlement that AI data centers should be allowed to get built anywhere, but I don't think the positive case has been made that it's actually that much of a net positive in many locations. The US is massive and if AI data centers are such a positive boon then it should be easy to find a place where they're actually welcome. Unlike traditional data centers where latency might be relevant, most AI data centers aren't latency sensitive so the default plan should be to build them far from major population centers. Just buy up a massive chunk of empty land in the middle of nowhere and drop all the data centers there.

There is some misinformation around resource usage, but most data center constructions aren't being fully candid and transparent with their expected resource usage either. I think the key issue is that many major constructions haven't built any trust with local communities nor have they listened to people's concerns, but they feel entitled to force through their projects anyway.

If you think that these AI data centers are great and incredibly beneficial, go advocate for your local government to bring construction to wherever it is that you live. Let people sort themselves out based on their interests.

It all started off as obvious bullshit: in 2024 Elon Musk was claiming that he could cut $2 trillion in waste, fraud, and abuse. He then walked that back to $1 trillion in cuts, but he didn't even get close.

At least for me, the jump in productivity has resulted in building stripped down one-off software for my highly specific use-cases.

You can use an LLM to create anything but you still need to know what it is that you're building, and you need to think through how everything should work or the LLM will just fill it with sausage. You can tell that the models are still quite jagged and limited by the mixed quality from a lot of the software that these presumed trillion dollar companies are putting out. The future is sausage.

A lot of people severely underestimate the value and impact of the arts because they don't produce immediately visible results. But artistic works are often a massive source of inspiration and they help people through dark times.

I think the more likely explanation is that he's just not as familiar with frontend web development and react, so he underestimated the difficulty. He was asking for help on Twitter for how to make pills inside of a text field. His failure was caused by a knowledge check rather than a skill check.

If I remember correctly, he wanted to fix the web search UI to match the features offered by Discord. He even crowdsourced some free work from his followers by asking them for help in implementing visual pills in text fields or something like that.

The basic feature set he wanted to implement wouldn't have been very difficult for someone who is experienced with react, but I'm imagining there's lots of minor quirks with the "last mile" of details related to internationalization / localization and accessibility.

Microsoft actually offered free VMs with different OS and IE versions to make testing easier.

Yeah I could see it being a problem if you're doing work on closed source or repos with sensitive credentials. Since my usage has all been on open source projects I'd be happy to share everything I'm doing if it can help train better models.

Do you have a substantive reason why you dislike this? What is the problem if it's opt-in? Nobody is forcing you to share your usage if you don't want to.

I'd prefer it if all the model builders could train on my usage rather than being limited to a single company. That'll hopefully help make all the models better in the long-term.

Someone should setup a plugin or something for Claude Code that makes it easy to log all inputs and outputs for people who are willing and interested in sharing their usage. I don't want Anthropic to be the only company that can train on my usage, I want to share my usage so it can be used for training all new models.

Once you have a system for collecting all logs, you just need a place where they can be submitted. Ideally it would be a freely licensed dataset that is publicly available for everyone.

Has anyone built this yet?

I took up acrylics painting a few years back and I've been surprised by how much is lost in photos and videos. The two colors with which I've noticed this the most are ultramarine blue and prussian blue. I don't think it's just the color though, part of it comes down to how light is reflected off the painting and where you're standing, as well as the texture and the brush strokes. I have a few paintings hanging in my room and occasionally I'll look at them for a while and it'll reveal a new perspective to me that I had previously missed, despite being the one who made it.

This post is making me feel a bit inspired to go outside and immerse myself in the forest to take in the greens. Thanks for sharing.

I've been thinking a lot about soul cultivation as a concept, and the general structure of the soul, and doing a bit of writing on the topic. I feel like this topic is surprisingly under-discussed and under-explored relative to how impactful it is. By soul I mean "the part of you that is an observer", in case this isn't clear. I think a lot of discourse gets caught up with metaphysical speculation instead of focusing on what is there and what is knowable.

Most recently I was also probing people about how they conceptualize of the soul, making my own drawings, and asking others for drawings. If you have a few minutes I would also be interested in seeing how you would draw a soul, given pen and paper or equivalent materials. It often feels like for a lot of people the concept of the soul gets comingled with very confusing definitions.

There's a general problem where certain concepts become so overloaded that just disambiguating and clarifying what is meant becomes a challenge. I will note that if your first thought or question is whether the soul is even real, you might be confused about the definition or we might be referring to different concepts.

What do you think are the most important problems that the US is currently facing?

It helps to look at some figures to bring it into perspective instead of making off hand comments about how it's not a lot. Remember these are human lives!

After some searching, I found estimates ranging between 600 thousand to 1.6 million people living with Hepatitis B in the US.

If we can help 20% of those people, that means significant life improvements to somewhere between 120 thousand and 320 thousand people.

If we take the upper end, that's half the population of Wyoming.

Does anyone have advice for getting it down? I feel like no matter what I mix it with, it grosses me out and it becomes extremely challenging to get it down. I've tried pouring the dry powder in my mouth and just chugging it down quick as well as mixing it with various liquids, but they sandy texture makes me gag. I've managed to sustain taking creatine for a while but it never gets any better.

I was recently talking to someone about that! I wasn't sure if it was my imagination, but I felt like Opus 4.6 was way more diligent about looking things up online and making sure that its response was accurate. While Opus 4.7 seems content to just throw out an answer as quickly as possible with little care for accuracy; I started to always remind it to do an online search and to double check its work, to the point where I had to add a custom memory.

That reminds me, do skilled players actually use the tilt keys? I remember being confused for years as to the purpose of tilt keys because I hadn't used a real pinball machine, and I can't remember it nudging the ball enough to merit the risk.

It's good, I have this song I generated last year with Suno which stuck with me and I just tried having Flow generate a variant and it was acceptable. Sometimes the lyrics get modified for no reason. It would be better if you could control emphasis by specifying tags or something along those lines, but it seems fun to play around. If there was an intermediate step where a symbolic or partial processing of your input was shown for tweaking, it would be immensely powerful.

One of the key issues that I encountered after a few song generations is that it feels very rushed, like it's constrained to this 3 minute limit per song so it forces every section of the song to conform to a very specific structure. I tried increasing the limit to 4 minutes but it still gave me 3 minute songs.

Honestly, I feel like this product is showing up a bit late to the party and it's not really feeling particularly innovative. There's nothing egregiously bad about it, but it doesn't seem to add anything new or special that I could notice.

I don't have a microphone hooked up so I can't try the voice interface, but it would be really fun if you could sing to it in order to iteratively compose a song. It could clean up your voice a bit and add music. Or being able to hum out a beat which it converts into a track which you slowly build up. Is anyone able to try if those capabilities are possible with the existing product?

Overall, I'm not sure if a chat interface is the best way to produce a song. It feels very restrictive to have full songs as the primary iteration mechanism. In a text file and in code you can inspect or modify different sections or components very easily. I think a more human-focused tool would provide an on-ramp towards full music production, where you can focus on the parts that you care about and enjoy, while the AI tool fills the other parts with sausage. Right now you can chat with the tool but it appears to be quite limited in the kind of changes that it can make.

It would be quite outstanding if MIT Plasma Science & Fusion Center released their core courses on OpenCourseWare. Considering the potential impact of this technology and how much it's needed, humanity doesn't seem to be trying that hard to make it work.

Gemini is great, but when I last tried it it wasn't as good as Claude Code with agentic workflows. When I tried Antigravity is was very unreliable, like the tooling had yet to catch up to the model such that it wasn't able to fully leverage the model's intelligence and capabilities. I think it comes down to how you're using the models, so I'll ask: how are you interfacing with the LLM?

Well, there's layers. When I started using nixOS I read through the guide and wiki but I also used LLM assistance to help create a stable starting point. Then over time I've incrementally added new things to my configuration through a mix of LLM assistance and reading online material.

I think the initial migration towards nixOS is the hardest point, since it requires learning a bunch of new things all at once in order to get the system into a usable state that matches your expectations and preferences. The key benefit of using an LLM is that it makes it really easy to get your system into a useful initial state, and then you can safely learn and experiment incrementally with a mix of tools.

When I started off I didn't understand everything, but at this point I feel I have a very good understanding of everything in my configuration file.

Replaceable cables have always been a high priority for me. I've been using SHURE SE215 for almost 2 decades at this point.

Pokemon cards have gone full circle, GameStop now has an online service where you can gamble on cards digitally just like lootboxes. You buy a roll at different price points to win a PSA graded card from a set of probabilities, and then you can sell it back for 90% market value to GameStop or have them ship it to you.

The proliferation of gambling over so many domains has radicalized me against it in a way that I didn't think would've been possible a few years ago.

I've been saying this for a while, but if I had to use Grok for anything programming-related I'd feel very sad and unproductive. I was playing around with a local TTS model codebase but having some issues getting it to work, so I tried explaining the problem to all the major models to see how they performed. Grok performed the worst by a significant margin, and the worst part was that it easily became stuck trying minor changes that didn't solve the key problem.

If we are to take any claims of Recursive Self Improvement seriously at all, then having a competent coding model seems like a key asset where you need to guarantee that you're remaining competitive. Why wouldn't you make coding models a top priority if you expect it to ultimately help your internal teams become more productive and effective?

There's also not an unlimited supply of researchers and engineers for them to keep burning through people at the rate at which they've been working. Although I guess for people with short timelines it makes sense to sprint hard, while people with longer timelines are more likely to treat this as a marathon. Maybe the years of burning bridges and developing such a toxic reputation are finally catching up to Elon. I think part of the harm that Elon has done is framing all the work in xAI as engineering while being highly dismissive of research, but a lot of research requires running experiments or thinking about problems and exploring them for long periods of time. If you're just grinding out work nonstop you don't really have time to let your mind wander and explore new ideas.

Honestly, I'm surprised they've done such a terrible job with programming. I remember around summer last year it was quite apparent how far behind they were with coding tools, but Elon was posting about taking that domain a bit more seriously. Why didn't any of those efforts materialize into real outputs? Something must be truly dysfunctional inside of xAI for them not to be shipping anything at all, especially considering Elon's propensity to ship undercooked products while continuing to iterate on them, as he has done in many previous cases.

I've noticed that Elon has also gone very hard on social media posting a ton of criticisms against the other big AI company CEOs like Daario Amodei. This suggests to me that he must feel very threatened, otherwise he wouldn't be resorting to such childish behavior. He must feel incredibly frustrated that no amount of money is able to make him more competitive within the AI space.

Surely any sane person vibe coding a note taking app just has it save all the notes as markdown files to disk? At that point making a backup is trivial and they're unlikely to get corrupted.

I vibe coded a tiny MUD-style world sim where LLMs control each character. It's basically a little toy sandbox where LLMs can play around. There's no real goal to this, I just thought that it would be fun, like a more advanced tamagochi.

One of the issues I encountered initially was that the LLMs were repeating a small set of actions and never trying some of the more experimental actions. With a bit of prompt tweaking I was able to get them to branch out a bit, but it still feels like there's a lot of room for improvement on that front. I still haven't figured out how to instill a creative spark for exploration through my prompting skills.

It has been quite exciting to see how quickly a few simple rules can lead to emergent storytelling. One of the actions I added was the ability for the agents to pray to the creator of their world (i.e. me) along with the ability for me to respond in a separate cycle. The first prayer I received was from an agent that decided to wade into a river and kneel, just to offer a moment in stillness. Imagining it is still making me smile.

Unfortunately, I don't have access to enough compute to run a bigger experiment, but I think it would be really interesting to create lots of seed worlds / codebases which exist in a loop. With the twist being that after each cycle the agents can all suggest changes to their world. This would've previously been quite difficult, but I think it could be viable with current agentic programming capabilities. I wonder what a world with different LLM distributions would look like after a few iterations. What kind of worlds would Gemini, Claude, Grok, or ChatGPT create? And what if they're all put in the same world, which ones become the dominant force?

Hypothesis: limiting usage / tokens could have a positive effect on project quality, since it forces the developer to think more carefully about the problems they're working on. When you're forced to stop and slow down, you try to be more deliberate with token usage. But if you have unlimited tokens you can just keep generating infinite lines of code without thinking as hard about the problem.

I've seen people on social media bragging about how they're able to produce a mountain of code as if this was praiseworthy.

Well, I'm currently reading through "Designing Virtual Worlds" by Richard Bartle. He's known for being one of the creators of MUD (multi-user dungeon). I'm not far along enough to make a judgement on the quality of the contents, but I keep seeing the book title pop up everywhere so it seems important.

Since we're on the subject of waste removal, most mammals also pee for roughly the same amount of time [0], around 20 seconds. If you look up a video of an elephant urinating, it's quite the spectacle and the flow is voluminous.

[0] https://www.nationalgeographic.com/animals/article/mammals-p...

That's really cute, it reminds me that Will Wright (creator of The Sims) has referenced this book "The Ants", by Bert Holldobler in multiple occasions as a key inspiration for his games (and in particular SimAnt) and systems thinking. Did you come across that during your research phase or had you not heard about it? I haven't read it yet, but maybe someday I'll get around to it.