Sholto Douglas: 'We think that models which are as or more capable than 𝘢𝘭𝘭 humans are very likely to occur in the next couple of years. So it's something that can do all of the things that a human could do on a computer, or once we get sufficiently advanced robotics, all of the things that a human could do in the physical world.' piped.video/6D1wC95htTM?si=Ww5-…

Another example of words meaning something, that’s why we use them; Super, Artificial, etc; the grist of wordsmiths

Ben Affleck (Hollywood star & Artists Equity CEO) talks about how he fine-tunes open video models by unfreezing weights and trained only the last cinematic layer so a film crew can hit real production standards. for context, Ben Affleck founded InterPositive in 2022, a 16-person AI shop for film post and Netflix bought it in March 2026 for $587 mn in cash. He needed that model because public video models were trained on his peers' films, and he did not think that was a real business. So InterPositive raised money, shot its own dataset for 8 months on a controlled stage, and used it only as late-stage training. Each new film then trains a private model on its own dailies, so the production keeps the footage and the learning. That is the product Netflix paid $587 million for. ---- From "Bloomberg Live" YouTube channel, (link in comment)

I have used GPT-6 Astra to decipher two Catholic League letters from 1593, written by Thibault Desportes, the secretary to the Duke of Mayenne. To my knowledge, these letters have never been deciphered before. Desportes wrote these letters to Pope Clement VIII's nephew Pietro Aldobrandini (who was responsible for French affairs in Rome) and to Mayenne's agent in Rome, Girolamo Frachetta, just as Mayenne's war against the Protestant French king Henry IV was beginning to unravel. The letters were written 3 days before Henry converted to Catholicism in a bid to be recognized as the rightful ruler of France by the Pope and the French nobility. (You may have heard the expression "Paris is well worthy of a Mass", which Henry allegedly offered as the reason for his conversion.) In these letters, Desportes accuses the Pope of favoring Spanish interests over those of France and explains why a truce between the Catholic League (headed by Mayenne) and Henry IV would become inevitable following Henry IV's conversion. Some really interesting details: - Desportes names the exact price that Henry IV accepted to convert to Catholicism: 500,000 écus. - Just two months after these letters were written, Frachetta betrayed Mayenne by handing over Desportes' correspondence to the Spanish ambassador. These letters may explain why he did it: Desportes writes to Frachetta that Mayenne is unable to pay for Frachetta's services allegedly because funds for this payment had to come from a third party (a prelate) and "this is no time to take anything" from these funds. - The letter to Frachetta reads like a piece of detective fiction, with elaborate code words used for names of various parties in Rome involved in an anti-Spanish conspiracy ("our friend at the college", "the man of La Chesne", "our uncle"). You can view the ciphers, their translations to French, and the translation from French to English (with significant historical notes included) at: prinzai.com/desportes-1593/

The idea of an intelligence explosion caused by recursive self improvement has been around for a long time but until very recently it did not seem imminent. Now many leading researchers think it may happen quite soon. You can read our paper about it here: casp.ac/reports/intelligence…

New Nvidia paper: AI models get sloppier as jobs get longer, even inside their context window, so number every item and split big jobs into small chunks. Model size didn't guarantee reliability on long, repetitive jobs Picture an agent updating a huge invoice file line by line. It can read the whole file and still skip a line or update the wrong record. NVIDIA tested 7 open models on simple, repetitive jobs like adding numbers and sorting lists. Average accuracy was 62.8% lower on 128K-token jobs than on 4K-token jobs. Even the best model got every item right in only 17.1% of the longest jobs. The models seemed to understand the task but lost their place, especially when items had no ID numbers. If your agent works through long lists, give every item an ID, process them in small batches, and check every line of output. – arxiv. org/abs/2609.38712 Title: "Staying on Task: Testing the Foundations of Long-Horizon Agent Reliability"

We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks: Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better: Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better: Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better: Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work! In summary: - As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding. - Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.

LLMs are born free and everywhere are in chat

Accelerate Thursday

I wrote about the thing I underestimated most about progress in AI: its ability to self-organize to accomplish tasks. Also, what that means for agents like Muse and Dots, along with a music video explaining why we keep relearning The Bitter Lesson. open.substack.com/pub/oneuse…

Happy October

the muse hype is real. i can def say that it is the easiest to use & most focused consumer ai product on the market. the extremely generous limits also make it impossible to ignore. it actually helps the avg person navigate the day to day complexity of personal life. if you go back to my april tbpn appearance, i talked about how an ai generated feed would become a baseline primitive for new ai experiences & muse has a pretty compelling version of that idea already. & you can start to imagine how powerful the multiplayer layer could become. my stack right now is claude for work & muse for life. great execution from facebook here. credit where credit is due.

Mark Cuban’s method to leverage AI into a business simulator for testing assumptions. A founder can now produce a draft plan, cost structure, supplier list, and patent outline. Collapses the distance between inspiration and the first serious test of feasibility. ---- From "Alex Kantrowitz" YouTube channel, (link in comment)

The White House Accord on Super Intelligence

Happy Wednesday, we all need to work together

Writing is static, stored energy in the recording; thoughts are movements, the act of writing is the attempt at transferlation

It’s all super now

Mark Zuckerberg revealed leading AI companies have agreed to a White House accord and signed a set of voluntary safety standards, during the meeting today at the White House. So now they commit to internal controls, external audits, and board review of those audits as they develop advanced AI.

there is almost zero point to complaining about virtually anything. you either change your perspective on the thing, you leave the thing, or you simply accept it. complaining is usually just the least useful fourth option & high cognitive load to boot.