Meta, Google join the AI launch party
Meta and Google launched new AI models in early September, with Meta's Muse Spark 1.3 achieving near-frontier performance at low cost while Google's Gemini 3.8 Flash represents a recovery step but still trails the frontier. The newsletter also covers AI safety concerns about reasoning transparency, tech literacy as a career ceiling, and practical AI workflows for professional development.
Summary
Meta and Google both released new AI models this week, marking significant industry activity. Meta's Muse Spark 1.3 scores 62 on the Artificial Analysis Intelligence Index, placing it behind only Claude Fable 5.1 and Opus 5 while maintaining significantly lower costs—Zuckerberg described it as having "frontier performance almost too cheap to meter." Meta has signaled that its larger model codenamed "Watermelon" is next, along with plans to release Spark's weights. Google's Gemini 3.8 Flash maintains the previous version's pricing ($0.75/$3.75) with improvements in coding, reasoning, and agentic tasks, scoring 59 on the Intelligence Index. DeepMink's Koray Kavukcuoglu acknowledged that Gemini sits "a little below the frontier" and emphasized that "there's nothing other than being at the frontier that's important for us." This reflects Google's pressure to achieve a frontier-level comeback after a difficult 2026.
The newsletter features an educational essay by Nate Grahek on tech literacy as the ceiling for AI adoption. Grahek argues that technical understanding—knowing what servers are, understanding cloud storage, grasping how data is organized—is foundational to maximizing AI capabilities, regardless of whether someone writes code. He notes that high-level adopters he works with (Uber engineers, tech-savvy small business teams) already possess baseline technical literacy, and that AI now provides an unprecedented opportunity for non-technical people to build this literacy through interactive learning.
A significant concern emerges around OpenAI's upcoming Astra model and reasoning transparency. OpenAI's new "recurrent depth" technique loops analysis over text multiple times to boost coding and reasoning performance, but these loops produce mathematical outputs rather than human-readable reasoning. OpenAI reportedly dialed back the loops so Astra still provides explainable reasoning, with promised monitoring at launch. Chief Scientist Jakub Pachocki flagged concerns about a potential "race into unmonitorability" driven by performance pressures, and admitted that current monitoring is already fragile.
The newsletter includes practical guidance on creating a "Proof Project" for job interviews—recording a five-minute AI workflow demonstration, converting the transcript into a five-slide deck, and tailoring it to specific roles. This approach emphasizes demonstrating human-in-the-loop thinking rather than just automation.
About this episode
PLUS: Nail job interviews with the “Proof Project” method
Key Insights
- Meta has achieved frontier-level AI performance at significantly lower costs than competitors, positioning it to challenge Google's market position through a combination of intelligence and affordability rather than raw capability alone.
- OpenAI's new 'recurrent depth' technique improves reasoning by looping analysis multiple times, but produces opaque mathematical outputs instead of readable reasoning, creating a tension between performance gains and AI safety monitoring that the company acknowledges as a broader industry risk.
- Tech literacy—understanding basic infrastructure concepts like servers, cloud storage, and data organization—functions as a hard ceiling on AI adoption capability, affecting even highly technical users who don't write code themselves.
- Current AI safety monitoring is already fragile and deteriorating for reasons independent of new techniques, suggesting that the ability to understand and monitor model reasoning may be closing regardless of individual model design choices.
- DeepMind explicitly prioritizes reaching frontier-level performance as the only acceptable outcome for Google's AI strategy, indicating that non-frontier models are considered insufficient even if more efficient or affordable.
Topics
Transcript
Good morning, {{ first_name | AI enthusiasts }}, and welcome to our 3,428 new readers. We’re only three days into September, and a flurry of new AI models has already hit the leaderboards. Both Meta and Google just joined the launch party Anthropic kicked off, but on very different trajectories — Zuck and co. keep climbing into rarefied AI territory, while Google is out to prove it still has a pulse after a year to forget. Meta, Google join the September AI launch party Nate's Notebook: Tech literacy = your AI ceiling Nail job interviews with the “Proof Project” method Report: OpenAI’s loops hit AI safety monitoring META & GOOGLE Image source: Meta/ Google The Rundown: Meta and Google both…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The Rundown AI
Fable 5.1 kicks off launch week at the frontier
Anthropic released Claude Fable 5.1, showing significant improvements in coding and research tasks with reduced safety rejections, marking the end of the summer's cautious release period as OpenAI's Astra launch approaches. Bernie Sanders published an op-ed calling for a global AI pause, citing control concerns and societal risks, while ongoing legal battles between Apple and OpenAI center on alleged theft of confidential designs.
Runway's Solaris previews the no-code internet
Runway unveiled Solaris, an AI-powered interface that renders websites and apps as real-time video with no underlying code, while Imperial College researchers developed an AI model that detects heart disease from ECGs in under two seconds with superior accuracy to human doctors.
OpenAI cuts out SpaceX-owned Cursor
OpenAI is removing its models from Cursor coding editor by mid-November following SpaceX's acquisition of the platform, citing Elon Musk's history of contract violations as justification. The move escalates the ongoing feud between Sam Altman and Elon Musk while putting developers in the middle of a high-profile corporate conflict.
Every machine is about to speak Claude
Anthropic unveiled the Model Hardware Standard, enabling AI agents to control physical machines like microscopes and robot arms with minimal setup time. The newsletter also covers voice-controlled AI workflows gaining mainstream adoption, brain surgery assisted by live AI video analysis, and various AI industry developments from major tech companies.
Grok 4.6 storms the AI frontier
SpaceXAI's Grok 4.6 has emerged as a competitive frontier AI model, scoring 61 on benchmarks and undercutting rivals on price at 60% less cost, with Grok 4.7 teased for release in 3-4 weeks. The newsletter covers AI trends including practical guidance on using AI effectively, a personalized cancer vaccine startup for dogs built with AI tools, and updates on competing models and AI products.