Lester Leong

scalable optimist.
building things i want to use. everything else is a side effect.

[Lester Leong @LesterL]

This is Cozyplanet.

Drift across the globe.
Listen to bossa nova in São Paulo.
Watch the snow fall in Reykjavik.
Read hyperlocal news from *anywhere*.

All through a UI that feels like 1996.

[Lester Leong @LesterL]

@AndrewCurran_ Lemme guess. Sparse attention?

[Lester Leong @LesterL]

@GroundhogStrat We can probably agree to some extent here, although I will say that staying relevant is just as important as being well-known in the first place, and also that outside of EA/tech circles, Eliezer still isn't widely known.

[Lester Leong @LesterL]

@foomagemindset @GroundhogStrat It was based on this tweet: https://t.co/X3Xpq6oQRA

[Eliezer Yudkowsky @allTheYud]

@xlr8harder For $10,000 I will pose in a picture where I and 4 other guys are standing behind Aella, who is sitting on a sofa, with the 5 of us wearing fedoras.

[Lester Leong @LesterL]

I mean, it probably also depends on whether the underlying message genuinely resonates, as well as the skill of the communicator. Trump has made many gaffes throughout the years that would have sunk a Howard Dean; the fact it hasn't sunk Trump seems to indicate that the difference lies with him. Also, though, for every Howard Dean, there are probably hundreds more politicians you haven't heard of, who failed simply because they never got their message out. I would rather be a Howard Dean with a 50/50 chance, than a nobody with a 0% chance.

[Lester Leong @LesterL]

But you're talking about it now, right? That's the playbook. Spread fast and deep within your in-group, garner enough attention outside of it so more people talk about your in-group. Even to this day people talk about Trump's opening "rapists and criminals" campaign remark about Mexico. Many didn't like it, but enough did (or *eventually* did, after digging deeper), to form one of the most powerful voting blocs in American history. And we still don't have a "big beautiful wall".

[Lester Leong @LesterL]

@GroundhogStrat Then you just weren't part of the in-group that would have understood it as such. To me, it was a direct nod to the $10,000 debate on AI safety that Eliezer took on several weeks ago despite the circumstances and the other person's demeanor.

[Lester Leong @LesterL]

As you know, this is based off of a tweet where Eliezer proposes doing this exact thing for $10,000. I understood that to mean that he preferred *not* to do it, but decided to quantify how much he didn't want to do it, instead of just communicating a vibe. Which of course is a core pillar of the rationalist movement that applies to decision making of all sorts, especially those related to AI safety, around which the movement was founded.

[Lester Leong @LesterL]

Note that you will need to self-host this somewhere (ie, GCP Cloud Run), and you will need API keys for the various features like AI-generated prompts.

Otherwise, feel free to fork and extend!

[Lester Leong @LesterL]

Testing an AR walking travel HUD I vibecoded for the Meta Ray Ban Display glasses in old Tokyo.

Inspired by open-world video games, you can:
- see POI markers around you in AR
- check off objectives
- bring up important travel notes like phone numbers, directions, etc
- convert to and from any currency
- get AI-generated briefs on dining, shopping, history and things to do at city, local, and hyperlocal levels

Fully open source and available at: https://t.co/yq6pUGzySd

[Lester Leong @LesterL]

Why did I make this? Because it's fun, and feels kind of like playing Fallout or GTA in real life.

More importantly, it keeps you present and aware. Constantly checking your screen for directions or looking stuff up makes you read as a clueless tourist, and that's exactly what flags you as a target for pickpockets and worse.

[Lester Leong @LesterL]

I think that report is overindexing on exchange ratios. It's a terrible proxy for overall casualty ratios, since Ukraine has lost more territory than Russia has gained (however small), and you cannot recover bodies from territory you don't own. With that said, Ukraine's rate of land loss has slowed down significantly in recent months, and in the same time period, Russia has lost more soldiers than it could replace.

[Lester Leong @LesterL]

@RokoMijic First project that RSIs itself into being better than all humans at everything. May or may not have argmax-like optimization tendencies, but having them will make it much worse

[Lester Leong @LesterL]

@liron @ESYudkowsky @47fucb4r8c69323 The best arguments on Doom Debates that I've seen thus far are from @RokoMijic and @QuintinPope5 - more like this please

[Lester Leong @LesterL]

@RokoMijic The entire mass of Mercury would be sufficient to build a practical Dyson swarm. That leaves Mars untouched, whose gravity makes it suitable as a habitable nucleus

[Lester Leong @LesterL]

@samuel_spitz Disagree, the steelman argument for Jensen is actually pretty sound. He's either too used to media-speak to properly lay out the correct argument, or is afraid to say something that could result in negative movement for NVDA

[Lester Leong @LesterL]

Also what really matters for the purposes of American national security is not China's total compute, but compute assymmetry between the two countries. What does it matter if China's compute grows by 10%, if American compute grows by 40% within the same time period (because America maintains its tech stack leadership), and China is restricted to n-1 generation chips? It means we get to keep the ecosystem centered in the US, which means better talent, better capital flow, better depth/breadth of companies, and ultimately a higher probability that we continue to make breakthroughs like Mythos before China does. And so we plug the holes well before China gets the same capabilities.

[Lester Leong @LesterL]

Jensen is essentially making a systemic argument, while Dwarkesh is making a more atomistic one. The correct framing for him would have been to consider it in relative or game theoretic terms. For example, does it make a difference if China's compute increases if American compute also increases, but by a much bigger margin? Jensen is saying that by ceding the market to China, we risk shifting the entire ecosystem to China, which means that China will speed ahead in terms of talent, funding, supply chain, and depth/breadth of companies. Which then increases the probability that they come out with the next Mythos before we do.

[Lester Leong @LesterL]

Dwarkesh did a great job. I get what Jensen was trying to say, and I agree with the gist of it, but he could have answered more directly. For example, he could have said, the more American technology diffuses around the world, the more centered the ecosystem will be around America, and so America will get more talent, more funding, more companies will build here first, and that results in other innovations across the other layers happening in America first (like Mythos) as opposed to China.