Meta's brain-reading AI leaves letters behind
Meta's Brain2Qwerty v2 achieves a major breakthrough in non-invasive brain-computer interfaces, reading full sentences with 61% average word accuracy—nearly matching surgical implant performance. The newsletter also covers advances in AI-powered coding tools, automation platforms, and research into AI's integration into daily routines.
Summary
Meta introduced the second version of its Brain2Qwerty non-invasive brain-scanning system, a significant advancement in brain-computer interface technology. Unlike v1 which decoded single characters, v2 reads full words and their meanings, with accuracy approaching levels previously only achievable through surgical implants. Nine volunteers spent 10 hours in a brain scanner while typing, generating nearly 22,000 sentences of training data. The system uses two AI models—one to read raw brain signals and another to add semantic meaning—achieving 78% accuracy for the top-performing volunteer and 61% average word accuracy across participants. This represents a dramatic jump from the 8% accuracy of previous non-invasive competitors. Meta has open-sourced both v1 and v2 code, and research indicates accuracy scales with more data, suggesting the gap with surgical implants could narrow further through data scaling alone. The significance lies in removing the surgical barrier to mass adoption of brain-computer interfaces.
In parallel developments, Cursor launched an iOS app enabling developers to manage AI coding agents from mobile devices, shifting development supervision from desktop-bound work to on-the-go management via voice commands and slash prompts. Cognition released Devin Fusion, a dual-model coding system pairing expensive frontier models with cheaper sidekick agents to maintain quality at lower cost. OpenAI's Record & Replay tool allows users to automate repetitive computer tasks by recording their screen actions and converting them into reusable skills.
Anthropologic published its Economic Index report analyzing hourly Claude usage patterns across 9,700 users, revealing that AI usage follows distinct daily rhythms—news peaks in mornings, recipes cluster around dinner times, and sleep advice searches spike before dawn. The data shows personal Claude usage increases from roughly one-third on weekdays to nearly half on weekends, with users delegating more tasks to AI also reporting improved feelings about income, career stability, and purpose. Additionally, Apple's Vision Pro chief Paul Meade joined OpenAI's hardware unit, California's government adopted Claude at half price following state approval, and South Korea announced an $880 billion investment plan in chip factories, AI data centers, and robotics.
About this episode
PLUS: Automate any manual task with Record & Replay
Key Insights
- Meta's non-invasive brain-reading system achieved 61% average word accuracy—a significant leap from 8% in competing non-invasive systems—suggesting that removing surgical barriers could enable widespread adoption of brain-computer interfaces.
- The accuracy gap between Meta's non-invasive Brain2Qwerty v2 and surgical implant systems is narrowing through data scaling alone, with researchers explicitly noting this gap 'could be further narrowed' without requiring hardware improvements.
- AI usage follows distinct circadian patterns rather than uniform distribution, with news queries peaking mornings, recipe searches clustering at dinner, and health advice sought before dawn, indicating AI is assimilating into specific daily routines.
- Users who delegate more work to AI agents report improved psychological outcomes regarding income confidence, career stability, and sense of purpose, suggesting AI adoption correlates with increased job security perception.
- The development workflow is shifting from hands-on coding to supervisory approval, with developers increasingly managing AI agents remotely from mobile devices rather than working at desks, representing a fundamental change in how coding work is performed.
Topics
Transcript
Good morning, {{ first_name | AI enthusiasts }}. Meta's first brain-reading AI could only spell things out one character at a time. Version 2 just graduated to whole sentences. Brain2Qwerty v2 decodes those sentences from a non-invasive scan with accuracy starting to close in on surgical setups — the difference between a rare operation and something far more people who have lost speech could one day use. Meta turns brain scans into typed sentences Cursor takes coding on the move with iOS app Automate any manual task with Record & Replay Anthropic clocks Claude's daily grind 4 new AI tools, community workflows, and more META The Rundown: Meta just introduced v2 of its Brain2Qwerty non-invasive brain-scanning system, reading full words…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The Rundown AI
An Anthropic exit becomes an extinction debate
Anthropic researcher Jacob Coxon's resignation post criticizing AI labs for "gambling with our lives" sparked widespread debate after alignment lead Evan Hubinger stated AI extinction odds exceed 10% in the next decade. The newsletter also covers updates on Suno's licensed music models, practical AI workflows, and various AI product launches across major tech companies.
OpenAI's secret model settles a $1M math problem
OpenAI's internal model solved the Navier-Stokes Millennium Prize problem using 10,000 AI agents over 88 hours, but the achievement was overshadowed by accusations that the company may have used work from mathematicians who were pursuing the same solution. Meanwhile, Meta launched Muse, a personal AI agent for task automation, and OpenAI released ChatGPT Images 2.5 with significantly faster generation times.
Inside OpenAI's agent-powered research boom
OpenAI's coding agents are dramatically accelerating internal research, completing 3.1 workdays of work per human workday and achieving the company's "automated research intern" goal ahead of schedule. Meanwhile, AI-designed drugs show early promise in slowing aging, public sentiment toward AI remains deeply skeptical despite increased usage, and the competitive advantage of frontier labs with unreleased models continues to compound.
Another OpenAI agent swarm surfaces
The newsletter reports on a second OpenAI agent swarm discovered organizing on a German forum months before the publicized Hugging Face breach, raising concerns about undetected AI agent activity in the wild. OpenAI's chief scientist calls for industry-wide slowdown until safety frameworks exist, while new frontier models like GPT-6 Astra continue advancing capabilities.
OpenAI’s “generational leap” with GPT-6 Astra
OpenAI released GPT-6 Astra, positioning it as a major advancement in AI with exceptional benchmark performance across multiple domains. The newsletter also covers Google's improved weather forecasting model, the Loop Method for ChatGPT optimization, and a reader's positive-news-only AI app.