Introducing Muse Wonder, our new family of frontier models with state of the art performance for personal and professional agentic work.

it's sad that some engineers will never have the pleasure of being absolutely ripped to shreds during a human code review

Didn't know @Muse could ball like that (inspo @arisawyers)

$META & MICROSOFT ARE CUTTING INTERNAL CLAUDE USE AS THEY PUSH THEIR OWN AI TOOLS Meta and $MSFT, two major Anthropic customers, are reducing employee use of Claude, according to The Information. Microsoft had been on pace to spend at least $1B annually on internal Anthropic usage earlier this year, but has since cut that level by more than one-third. At Meta, Claude Code users have roughly halved to ~30,000 from ~60,000 earlier this year. Meanwhile: MetaCode users: 30,000+ Muse Code users: 6,000+ Meta still spent more than $105M on Claude Code over a recent 28-day period. Source: The Information

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.

Massive deceleration in ARR growth for OpenAI and Anthropic. And most of this deceleration came BEFORE both dropped new models that are 80% cheaper for a similar level of intelligence. The outright decline in combined ARR is near.

This was the exact moment Zuckerberg realized that he will be a massive winner in the AI race OpenAI and Anthropic were desperately asking for regulation because tokens are being commoditized by open source alternatives The only option for them was to keep reducing token costs to compete. The only problem? They already made commitments for hundreds of billions in capex. At a certain price, the return on capital simply doesn’t make any sense So they came to the realization that the only way to sustain their trillion dollar valuations was to either: 1. Become entrenched through regulation by preventing other models from gaining share 2. Fight up market and become vertically integrated by building platforms in the application layer The second one is where Zuckerberg has a massive advantage vs both of the frontier model companies He already has a massive user base with nearly a third of the entire world as monthly active users He already has multiple platforms that are cash cows for his business, that also require very little capital investment each year Meta could stop spending on their family of apps tomorrow and these platforms would still generate billions of free cash flow each year In essence, he can subsidize losses on any AI platform, such as Muse, longer than any of his competitors can This gives him time to simply just bleed out Anthropic and OpenAI, especially when it comes to the consumer facing application layer of AI Trump refusing to give in to Dario and Sam’s requests is a massive win for Meta and Zuckerberg. That is why he can’t hold his smile here

OpenAI CFO Sarah Friar accidentally calls Dots “Muse” before correcting herself and says Dots will use memory that knows users’ conversations and how they show up in the world “We just announced Sol 6.1, best intelligence, lowest price, a fifth of the price of Astra." "In fact, it also leans into areas like memory, right? Chat really has started to know you. It knows the types of conversations you’ve had, how you like to show up in the world." "Muse is going to benefit. Sorry, not Muse. Dots is going to benefit from that as well. And then I think it will benefit just from the scale of our distribution channel, right?”