> given that the trajectory for all models is a kind of singular “peak” performance. Not sure I understand this. Most other AI researchers I've spoken to don't expect an arbitrary hard cap on model intelligence, and those that do don't think we're close to it yet. As a concrete example, ARC-AGI 3 is great - the best models can't reach 10 percent, but even a negro could get something in the 90's, or at least the 70's. > Deepseek V4 beats closed source models brutally based on any metric. And it does fine on math/coding/agentic/etc. Hard to find recent metrics, but it does look very promising so far, and I love that it's open source. Lots of fun things can be done with that. Anecdotally, I haven't seen the panic I saw during DeepSeek's original release. CAISI claims it's benchmaxxed and performs worse (GPT-5 level) on non-public benchmarks. They might be lying, but sentiment is that they aren't. A more critical take than mine is here: https://thezvi.substack.com/p/ai-166-google-sells-out?open=false#%C2%A7seeking-deeply > I would also say that the open source model is simply more sustainable long term in any competitive environment. Linux dominates almost all infrastructure Open source's value to Linux is that every autist on Earth can scour its source code for problems then share out the fixes, and most do. The same advantage doesn't exist for model weights, which can't be patched by Sergei the160 IQ Siberian autist operating from a 1995 Compaq machine. There is an advantage to being the universally-used research platform, but it's not direct nor as large in magnitude. > As far as “America is behind because of Indians,” this is a kind of “pigs can’t fly because they don’t have wings” argument that I do not understand. If we once had a great reserve of flying pigs, and we systematically clipped the wings off of almost all of them, but kept a handful of unclipped ones in a lab somewhere, those pigs would be of note to anyone in need of air-mobile pork. To be clear, jeets have clogged up the recruitment and education pipeline in a way that has downstream effects even on the high-importance research teams that are allowed to not hire them directly. But a team that does not hire them will do much, much better than the U.S. tech industry's current state would suggest. For the record, I am rooting for China here, on the basis that I don't trust any of the U.S. frontier companies.

I use Brave, but use Ublock for ad-blocking. Much more versatile, and lets me manually nuke annoying website elements that don't technically count as ads.

https://archive.is/rnv0g#selection-967.20-967.278 Neat Haaretz article on what's going on with the Israeli government. > "I'm sorry to be so depressing, but our fate is already behind us. Everything's already happened." > Our society has been torn apart, the army has disintegrated, the judges are dying of fear, the media has become a reality show, the Knesset has become an insane asylum and the opposition shares Netanyahu's view of reality (Iran is an existential threat; there's no solution for the Palestinian problem; only Zionist parties should sit in the cabinet). > The world hates Israel, and antisemitism has returned to its political cradle. It is no longer the "new," left-wing critical version (which was aimed mainly at Israeli policy, and the faults of Zionism), but the old, right-wing murderous version (which gleefully adopts the rhetoric of the "Protocols of the Elders of Zion"). > The truth is that while we drove ourselves and the world crazy with the Holocaust, while we chanted "never again" ad nauseam, Netanyahu has led the world to the brink of a repeat of history. > All of them sustain the state, and therefore, they sustain Netanyahu, because the state is him. What's the alternative? Dodge the draft and let the country die? Kill the state to get rid of him? > But what is certain is that we will not be able to resuscitate the life we had. There is no way back to what used to be. There is no future for the state as it used to be. The state is him. And his end will be its end. He killed it. They don't seem optimistic about the inevitable - though not mentioned - push to rebrand Israel once Netanyahu inevitably leaves. Rightly so, I think. Neocons have taken over the Democrats' institutions, and faced much less pushback than they have on the Right, but the lefties mad at Israel are mad about a long chain of humiliations, and all of this is sublimated into their ressentiment about Israeli Dems no longer even pretending to care about the base's qualms about them.

I feel like the massive push for AGI is the same deal as wunderwaffe at the end of WWII. Sure, it's a long shot bet, but anything that isn't a long shot bet is just not going to be sufficient even if it works. Too many things are going wrong for the regime at the same time. Losing two wars, every White person under 40 is radicalized against them, and Chyna is outcompeting them economically.

Can't you buy XMR with other crypto? I got mine by sending BTC to a wallet, then using Exodus to convert.

I feel like, ultimately, XMR address in bio is the gold standard here. The region of people who are high-agency enough to set up Lightning but low agency enough not to be able to do XMR seems very small. Even smaller on something like nostr, where I assume absolutely everyone already has an active XMR address.

There is overwhelming evidence that our enemies are neither united nor competent. The first actual resistance they face in America - the first serious person to tell them "no" when they ask for abdication - will crush them easily.

If the Chinese want to read my 10 page screeds about the Jews, then they're welcome to do so. Maybe they'll learn something.

If Mark Levin actually wanted to alleviate anti-semitism he would publicly apologize to White people and then livestream his suicide by self-immolation.

> Further, I will continue to maintain that Chinese AI is much better and is actually driving development now, rather than the reverse. Qwen and Deepseek/Huawei (Qwen also integrating Huawei) are the defining AI models already, and the question now is the effectiveness with which US companies will be able to rip them off, particularly with relation to efficiency. That's an extraordinary claim. The benchmarks seem pretty unbiased, and they seem to favor the American models. As other commenters have mentioned, jeets are slowing America down with their incompetence, but the teams working on the important bits are almost all White or East Asian.