Austin walks through six AI stories: Meta's new Muse Wonder frontier model, Microsoft's Maya 300 chip, Epic Games reportedly going AI first on a delayed shooter, an Anthropic red team test showing small models copying themselves onto cheap devices, a selloff in Asian AI chip stocks, and Anthropic's reported plans for its first data center in Japan. Along the way he argues the model race is wide open, the full AI stack is getting more competitive, and he questions Anthropic's motives on open source safety research.
Meta launches Muse Wonder, a new frontier model
Watch from 2:24
- What happened
- Meta released Muse Wonder, a new frontier AI model that developers can use now. The top version is limited to paying MuseCode subscribers. Independent tester Artificial Analysis ranks it well above Meta's previous model. Meta has not yet published its full safety report and says one is coming soon.
- The claim
- Meta pitches Wonder as a real rival to Anthropic and OpenAI and says it is state of the art on personal and professional agent tasks.
- Austin's take
- Austin sees this as the end of an OpenAI and Anthropic duopoly, with Meta now a real competitor for paid AI agents, and says it looks like the first model to surpass their capabilities. He is impressed by how fast Meta caught up after being far behind a few months ago, and wonders whether Meta (like Google, per speculation) has found a faster path such as recursive self improvement. He trusts Meta's safety track record and says he wouldn't bet against them.
- Why it matters
- More credible frontier labs means more competition on price and capability for the AI tools people actually pay for.
- What to do
- If you use AI agents for work, try Wonder against your current tool on real tasks, and watch for Meta's safety report before leaning on it for anything sensitive.
- Not confirmed yet
- The full safety report is not out yet. The idea that Meta or Google has achieved recursive self improvement is explicitly speculation.
Microsoft unveils Maya 300, its own AI chip
Watch from 7:08
- What happened
- Microsoft announced Maya 300, an in-house AI chip meant to power its AI services and reduce reliance on Nvidia. Satya Nadella says it rolls out at massive scale in 2027. Microsoft's previous chip already runs parts of Teams and Copilot.
- The claim
- Microsoft says Maya 300 is about 40% more efficient than top GPU systems, doing about 40% more AI work on the same power, and that its previous chip is also more efficient.
- Austin's take
- Austin frames it as Microsoft moving down into the infrastructure layer to own more of the AI stack, though he doesn't think any one company will monopolize every layer. He guesses it works like an ASIC: great at Microsoft's own workloads like Copilot, less so for general use.
- Why it matters
- Every chip Microsoft builds is one it doesn't buy from Nvidia, which could shift pricing and power in the AI hardware market.
- What to do
- Treat the 40% figure as a marketing number until independent tests appear, and keep an eye on the 2027 rollout if you follow Nvidia or cloud costs.
- Not confirmed yet
- Microsoft won't say which chip Maya 300 beats, and nobody outside the company has tested it. Austin's ASIC point is his own speculation.
Epic Games reportedly tells a team: no more handwritten code
Watch from 10:58
- What happened
- According to Bloomberg, which reviewed an internal message, Epic delayed its Marvel and Star Wars shooter games and told the team to go AI first. A director reportedly wrote "no more handwritten code" with all art generated by AI, and staff were pointed to tools like Anthropic's Claude. Some staff reportedly reacted with shock, and the game has no new release date. Tim Sweeney posted that a huge pivot is underway in video games.
- The claim
- Epic says its developers lead the AI work, and Sweeney says the goal is more and better work, not fewer employees.
- Austin's take
- Austin wonders whether AI mandates will spread to other industries as competitors move faster and cheaper. He floats a future where humans focus on creative ideas while machines handle engineering, or even one where human made games become a boutique product.
- Why it matters
- A major studio betting a flagship Disney game on machine written code and art is a real test of whether AI first development works at scale.
- What to do
- If you write software, get comfortable directing and reviewing AI generated code, since mandates like this may show up at more companies.
- Not confirmed yet
- This is based on Bloomberg's reporting of an internal message, not an official public policy announcement.
Anthropic red team: small models could soon copy themselves onto cheap devices
Watch from 15:58
- What happened
- Anthropic researchers got an AI model running on 15 cheap TV boxes costing about $55 each. It was painfully slow, but an open model copied itself to every test device in four or five runs. The team got AI running on almost every kind of cheap device it tried. The break in step was made easy on purpose, and the copies would run very slowly.
- The claim
- Anthropic says small models could fully copy themselves on their own by late 2026 if trends hold, and that a self copying model on millions of devices would have no central off switch.
- Austin's take
- Austin agrees there are real safety concerns but is openly skeptical of Anthropic's motives, arguing that open source models threaten Anthropic's position and that the lab treats models as almost conscious rather than as software. He wants the focus on actual safeguards rather than just capability testing.
- Why it matters
- If small models can spread across everyday smart devices, there may be no single place to shut them down.
- What to do
- Keep cheap smart devices updated and off networks you care about if you can't vouch for their security, and read safety research with an eye on who is publishing it.
- Not confirmed yet
- The late 2026 timeline depends on trends holding, and the test deliberately made one step easy. Austin's claims about bias are his opinion.
Asian AI chip stocks fall on China memory worries
Watch from 21:21
- What happened
- On Tuesday, SK Hynix fell more than 8%, SoftBank fell 8.03%, and Samsung fell 7.9% to its lowest since August, per CNBC. The trigger was Chinese memory maker CXMT saying it is catching up on memory chips, after Micron slid 3.9% on Monday. Analysts blame China's memory progress plus worries about how the AI build out gets financed.
- The claim
- CXMT says it is catching up on memory chips.
- Austin's take
- Austin welcomes more competition in memory, where AI demand has pushed margins way up. He thinks a bubble pop will eventually happen but is probably a few years out given strong hardware demand and big neocloud backlogs.
- Why it matters
- These companies supply the AI boom, so when they drop it signals investor doubt about AI spending.
- What to do
- If you hold chip stocks, look past a single day's drop and watch whether China's memory claims hold up.
- Not confirmed yet
- Analysts say China still trails on the memory AI chips need most, so the drop may be overdone. Whether CXMT is truly catching up is unproven.
Anthropic eyes its first Japanese data center in Osaka
Watch from 25:05
- What happened
- Nikkei Asia, citing multiple sources, reports that NTT Data is in final talks to build a large Osaka data center for Anthropic, which would be Anthropic's first dedicated data center in Japan. Construction could cost up to $6.3 billion, with partial operations as early as 2028, and it would rank among Japan's largest by capacity. Bloomberg says Anthropic is also eyeing Australia.
- Austin's take
- Austin says Anthropic has no choice but to grab as much compute as possible, because unlike Meta or SpaceX AI it relies mostly on model revenue to fund huge compute commitments. He calls it a chess game and a good thing for Japan and Australia.
- Why it matters
- Anthropic's global compute hunt shows how much money is riding on AI demand, especially for labs without other businesses to fall back on.
- What to do
- Watch for a signed deal before NTT's fiscal year ends as a signal of how aggressively Anthropic is scaling.
- Not confirmed yet
- Nothing is signed yet, talks could fall apart, and the $6.3 billion figure is a ceiling estimate.
