Meta's model turn brain waves into words

Welcome back, Superhuman. Last year, Meta launched a brain-to-text tool that monitors and decodes brain waves. Yesterday, it dropped the upgraded version — and it’s trained on 10X more data than its predecessor. With breakthroughs like this, it feels like the singularity is getting closer by the day.

Today: OpenClaw gets an app, how to audit website SEO with Replit, and get the latest prompts and trending social posts.

1. Meta updates its research tool that turns brain waves into sentences: Brain2Qwerty v2 can analyze raw brain signals captured via a noninvasive scanner and decode sentences in real-time at 61% accuracy. It’s Meta’s highest-performing brain-to-text tool yet, trained on roughly 22,000 sentences from 9 volunteers each recorded for 10 hours. The stated goal: to benefit people who can no longer speak after a stroke, accident, or brain disorder. Learn more here.

2. OpenClaw and Cursor go mobile: OpenClaw — the platform that popularized personal AI assistants in January — is now available on the iOS and Android stores, letting you command your AI assistant from a phone-native app. The AI coding platform Cursor also launched an iOS app so you can kick off always-on cloud agents at any time. This viral release video breaks down how it works (3M views).

3. Samsung, SK Hynix, and Micron face class-action lawsuit over price fixing: A new California lawsuit accuses all three memory giants — who together control roughly 90% of the global DRAM market — of illegally coordinating to restrict supply to inflate prices. This comes as memory prices have soared in the past year, driving up costs for everything from Apple Macs to Microsoft's Xbox.

Messy PDFs, broken tables, nested charts… Most AI workflows fall apart before the model even gets context. 

How do leading AI teams do it? They use LlamaParse to turn hours of manual document processing into seconds of automation, with VLM-powered document understanding agents.

Stop patching brittle OCR pipelines—feed your agents clean context at scale.

Get 20,000 free credits today to use before July 31st with coupon code: SUPER20

The growing debate. For years, there wasn’t much competition when it came to using a closed or open-weight model. Proprietary closed models, like Claude, ChatGPT, or Gemini, were more powerful than open-weight models by some distance, so the decision was easy. But that’s changed abruptly in the past few months, with several open-weight contenders closing the gap — and giving corporate America a decision to make.

What’s the difference? Open-weight models are published freely for anyone to download, so you can host them privately, customize them, and keep data local on your own computer. Proprietary models like GPT-5.5 and Claude are closed: you can only access the model through the company that owns it, usually through a chatbot platform or API.

Why go open-weight? Open-weight models offer three benefits: more control, privacy, and often lower long-term costs than proprietary models. For years, none of that mattered because the models weren’t as strong. But today’s open-weight models have grown highly competitive, with Zhipu AI’s GLM-5.2 ranking 6th on Artificial Analysis’ leaderboard — closely followed by MiniMax-M3 and DeepSeek V4 Pro.

The decision. Some companies, like Uber and ServiceNow, blew through their entire AI budgets just months into 2026. One option some are weighing is shifting workloads to cheaper, open-weight models that match performance while cutting costs. That said, the trade-offs in performance, security, and engineering overhead mean it isn't the right fit for every team.

The Work AI Index 2026 from Glean’s Work AI Institute surveyed 6,000 digital workers and found that AI saves time, but much of it goes back into cleanup. Read the report to see where AI time savings go and what high AI achievers do differently.

Download the report

📝 Viral Post: It’s frustrating when AI blindly follow your instructions. One Redditor found a solution: adding a rule to “always suggest a better alternative” when doing tasks for you. See the full prompt (2.5K upvotes)

🤔 Role Evolution: Claude Code’s creator sent out a high-signal post on what the future of tech roles could look like. He expects 5 role archetypes to replace today’s domain-specific jobs (2M views).

🤦 LLM Writing: The internet is calling out “fake profound statements” as another obvious sign of LLM writing — consider rephrasing this structure if you’re borrowing work from AI.

😺 Live Action: Today’s AI video-generation tools make it easy to convert 2D animation to 3D live action. This viral Tom and Jerry clip shows how (3.5M views).

🔒 Protecting Privacy: National Design Studio released Rampart, an on-device tool that helps keep your personal info separate from AI chatbots.

^{* indicates a promoted tool, if any}

Readers loved this AI rendering of what famous historical figures might look like today, including Socrates, Mozart, Cleopatra, and more.

Grow customers & revenue: Join companies like Amazon, HubSpot, and Salesforce. Showcase your product to our 1M+ readers and 2M+ followers on socials. Get in touch.

Until next time — Zain, Theodore, & the Superhuman AI team.