Torch Bearer

@skdh @AshtonForbes The litmus test is in who can actually do more than talk using the wisdom and understanding they claim to have. Similarly for medicine, if you need actual heart surgery you have virtually zero chance of surviving if you choose a self taught internet grifter over an MD surgeon.

@AshtonForbes which part of "where did the plates come from" is it that you do not understand?

It's embarrassing how PhD physicists don't understand quantum mechanics or the Casimir effect. They get the most basic things wrong, repeatedly. The Casimir effect does prove we can extract zero point energy, otherwise the plates wouldn't come together on their own. That movement is a force that can be used for work. That's undeniable and specifically what Robert Forward showed in his 1984 paper. What everyone is actually upset about is PERPETUAL extraction, which in the case of plates requires them to be pulled apart. A valid argument for why that system won't work. Naturally, people have been working on designs to get around that, including Sonny White's idea. What will shock you is it can be accomplished with plasma and you start to wonder what's really going on with fusion. I've given up trying to reach these lost academics. Physics can't progress until they die because they're too much of fragile cowards to debate, since they know they'll lose and look foolish.

@PeterBerezinBCA There are definitely a lot of sociopaths coming out of the woodwork.

AI companies really need to come up with a better pitch to the public than “You’re all gonna lose your jobs and end up paying way more for electricity”.

@brivael It's pretty obvious that within our lifetimes we could have a full sensory experience of a virtual world indistinguishable from the real world. Imagine Neuralink, Compute, AI, etc in 20 years given current progress. And it's accelerating by all measures.

Elon Musk pense qu'on est dans une simulation. Moi, mon film préféré c'est Matrix. Et voici ma théorie sur la simulation : C'est improvable, ok. Mais en tant que philosophie de vie, c'est redoutable. Le principe : pour que la simulation reste active, il faut qu'il y ait du fun dedans. Sinon les designers s'ennuient et ils débranchent. Or, qu'est-ce qui tue le fun ? Les systèmes rigides. La bureaucratie. Quand tu enlèves les libertés individuelles, tu tends mécaniquement vers un système boring. Des formulaires, des comités, des normes, des sous-comités sur les normes. Plus personne ne crée, plus personne ne prend de risque, plus personne ne joue. Et là, les mecs derrière l'écran regardent leur dashboard et se disent : "bon, faut faire quelque chose." J'ai vu passer une théorie qui m'a fait mourir de rire : le Covid aurait été envoyé par les designers exprès. Pas pour nous nuire — pour pousser la bureaucratie le plus loin possible. La forcer à se révéler dans toute son absurdité. Confinements, QR codes, autorisations de sortie, comités d'experts qui se contredisent en boucle. Un stress test à l'échelle planétaire. Le but : faire péter le système par excès, pour permettre le reset. Et c'est exactement ce qu'on est en train de vivre. Trump, Musk, Milei — ce sont les incarnations du patch. DOGE qui démantèle les agences fédérales. Milei qui tronçonne l'État argentin en direct. La tech qui reprend le narratif. Le retour brutal des libertés individuelles comme valeur centrale. On assiste à un renouveau de civilisation. Et il est massivement basé sur la liberté de l'individu de créer, de buildre, de prendre des risques. Conclusion opérationnelle : Traitez la vie comme un jeu vidéo. Accumulez un maximum de skills. Buildez des trucs. Faites des choses qui vous donnent du fun, ou qui donnent du fun à l'humanité — et accessoirement, aux types qui nous regardent depuis l'autre côté de l'écran. Soyez intéressants à regarder. C'est littéralement votre seule mission.

Thats certainly true to some extent, except the quality of life today is lower than it was in 2007 for most western countries. US quality of life probably peaked from 1985 to 2008, and birth rates were actually up and sustainable during that time. Another period regarded as the highest quality of life in the US was the postwar 1950s-1960s, where we also had a huge birthrate surge. I cant speak to China and bangladesh, though my list certainly impacts their population to some extent. China did a lot of other things that inevitably created their current situation. I forgot to mention a massive issue, fertility itself, which has plummeted globally due to a variety of factors like microplastics. Testosterone has also plummeted. One interesting dynamic at least in the US, is that birthrate is very different based on political affiliation. There is definitely a huge psychological component.

@DragonStacker @paulg All of your mentioned reasons might be applicable to US, but not to Bangladesh or China where fertility rates also plummeted. When the quality of life rises, people stop having many kids. A woman in Gaza gives twice much (3.5) birth than in US (1.6).

If Steve Jobs were still alive, he would have the moral authority to face and maybe even to solve this problem. But I doubt anyone in the phone business now does.

@PositivFuturist It's a form of thinking in the limit, and it helps you see the truth of most arguments.

The best way to defeat liberals in argument is to argue from the left of them. $15 minimum wage, why not $100? Asylum? why not open borders? They have no principles they can call on to limit their own arguments.

@petergyang Subagent delegation with codex or claude can put a lot of code down really quick.

How do people even do AI hackathons these days you're just sitting around waiting for the agents half of the time?

@elonmusk @XFreeze Be that as it may, it's going to break a lot of API integrations through OpenRouter. I know several in use SaaS apps I built are impacted. Resources should be scaled back as a function of traffic, but not deprovisioned. They are not ancient; they were 2025.

@XFreeze These are ancient

MASSIVE UPDATE FOR xAI DEVELOPERS xAI is officially retiring several older models today Deprecated models include: • Grok-3 • Early Grok-4 fast variants • grok-code-fast-1 You can now migrate all text, reasoning, and code workloads to the new Grok-4.3 Grok-4.3 is an absolute beast: ✅ 1-million token context window ✅ Adjustable reasoning effort (None / Low / Medium / High) ✅ Tops industry leaderboards for agentic tool calling and instruction following ✅ Cheaper API pricing: $1.25 / 1M input tokens and $2.50 / 1M output tokens Best intelligence-per-cost model available today If you're using image generation, simply swap grok-imagine-image-pro for the new grok-imagine-image-quality Update your endpoints and start building with the latest state-of-the-art model

@FortressLugh I was in college at the time. Everyone loved it as far as I remember. I saw it a few times in the theater.

43 year old guy here. It definitely wasn't. I saw it in theaters. Everyone loved it, and Pitt especially. He was widely praised at the time for his performance in the role. He had a brooding intensity that really captured the essence of Achilles, and his physical performance in combat was excellent. Very convincing. I'm hardly alone in this view. Everything about Troy was better than this new Odyssey disaster: from casting to clothing.

@alphafox Just saw that Netflix was billing us ~$30 per month. Turns out nobody in the house has used Netflix for months. I've had a subscription for like 15 years. Felt surreal to cancel it, but I'm sure af not going back to DVD stacks.

People are ditching their streaming services and going back to DVDs after recent price hikes from streaming providers - good idea?

@jun_song Opus is still the best for Frontend. It simply has better taste. Kimi 2.6 and GLM 5.1 are fierce competitors though.

Best AI models now : Backend : GPT-5.5 Frontend : Kimi-K2.6 Agentic : MiMo-V2.5-Pro Best Price : Deepseek-V4-Flash Worst Price : Opus-4.7 Fast Research : Grok 4.3 Image : GPT-Image-2 Video : Seedance 2.0 (#1 for 3 months) Yes, Claude sucks.

@BassonBrain Verizon and AT&T are poorly run companies that lack any integrity or agility and have been for a long time. It's time for those parasites to be put out to pasture.

Verizon, AT&T and T-Mobile said on Thursday they agreed in principle to form a new JV with an aim to address long-time coverage gaps, especially in rural areas, by using satellite-based technologies. This comes as the industry increasingly worries about what Elon Musk’s @Starlink Mobile might do to shake up the terrestrial mobile space. Musk has said he’s not going to put the U.S. terrestrial carriers out of business, but at the same time he’s expanding Starlink and buying up more spectrum...

@thdxr Wonder if that is why it has been so slow this last week. Even low effort feels extremely slow for really simple tasks.

something is going wrong with gpt 5.5 caching doesn't look like much on this chart but this it's now using 2.5x as many input tokens as a week ago and dropping

@TheCinesthetic Definitely one of the best TV shows ever made. I am really enjoying re-watching it. On par with Expanse all things considered.

The lack of awards love for Edward James Olmos in Battlestar Galactica still feels wrong. Adama could command a room with a speech, then completely break your heart just sitting silently with the weight of leadership on his face.

@arena @AnthropicAI @ErnieforDevs The vast majority US business automation and random AI tooling is being built on Chinese models. Simply due to cost, and sometimes the ability to self host.

US vs China update. Stanford's AI Index put the US–China gap at 2.7%. Here's what two years of real-world use from the Text Arena shows. Gap three years ago: +278. Today: +29. @AnthropicAI's Claude Opus 4.6 Thinking vs. Baidu's @ErnieforDevs Ernie 5.1 at the top. The US has never lost #1, but the race keeps closing.

@thdxr That would/should work particularly well with the Rust rewrite.

so should bun even have its own parsers and bundler anymore or should it just build with the void0 crates

Separate processes via two user launch agents on MacOS, listening on separate ports. Both set to parallel=1. Here are more details: 27B - Model: Qwen3.6-27B-UD-Q6_K_XL.gguf - Vision projector: mmproj-F16.gguf 35B - Model: Qwen3.6-35B-A3B-UD-Q6_K_XL.gguf - Vision projector: mmproj-F16.gguf Both run with full 262k context, Metal GPU offload, flash attention, Q8 KV cache, continuous batching, and MTP speculative decoding via: --ctx-size 262144 --gpu-layers 999 --flash-attn on --cache-type-k q8_0 --cache-type-v q8_0 --cache-ram 32768 --spec-type mtp --spec-draft-n-max 3 --cont-batching

@DragonStacker 80 tok/s on a 35B at Q6 is wild. are you running both models in the same llama.cpp instance or separate processes? curious how you're handling the context switching

local LLM people: what are you actually running right now? everyone talks like they have a DGX Spark under the desk, but I’m curious what the real setups look like DGX Spark 128GB unified memory? RTX 6000? RTX 5090 32GB? RTX 3090 24GB? MacBook Pro? Mac Studio M3 Ultra, if you somehow found one? or are you running something completely different that people are sleeping on?

@SawyerMerritt @Starlink Looks like a cheap acquisition target for a company like amazon trying to break into LEO internet. Customers would have to be migrated but there is still value there.

Hughesnet has lost 57% of its broadband subscribers (880,000 subs) since @Starlink went live in December 2020. Hughesnet says that “we currently do not have the necessary cash on hand” to pay off debts, leading the company to warn that “substantial doubt exists about our ability to continue as a going concern.” In Q1, Hughes reported a net loss of $7.6 million. Hughesnet reported that its subscribers were at 681,000, down from 1.56 million in Dec 2020, of which about 1.19 million were US customers, including retail and enterprise users. Hughesnet has lost about 100,000 subscribers each year since late 2020. The trend continued in Q1 2026, with the subscriber count decreasing by about 20% year over year from 853,000. pcmag.com/news/starlink-…

@stevibe @UnslothAI Those 27B and 35B mtp drops are 🔥. Been running both since yesterday.

Whoa, @UnslothAI is on a roll!