🛎️ First Brain LLM

Good Morning, AI Enthusiasts!

Cognition abstracts upward. Infrastructure consolidates downward. Between them, a new layer forms.

NEW LAUNCH

First Brain LLM

👀 What’s happening: Zyphra released ZUNA, positioning it as the world’s first large scale foundation model trained specifically on EEG brain data with open weights and tooling. At 380M parameters and trained on roughly 2 million channel hours, it treats neural signals as a pretraining domain, not a narrow experiment.

🌍 How this hits reality: Unlike language models that learn from structured text scraped at internet scale, ZUNA trains on noisy, heterogeneous brain signals collected under incompatible protocols. There is no natural tokenization, no shared grammar, and far less data density. That makes generalization across devices and layouts technically harder than scaling text models, yet potentially more defensible.

🛎️ Key takeaway: Language models digitized human knowledge. Brain models attempt to digitize human states. If this approach scales, the leap is not better autocomplete, but a new interface layer between cognition and machines.

TOGETHER WITH MYCLAW

Join the OpenClaw Revolution

You don't need a Mac Mini to run.

Use MyClaw.ai instead. With it you can safely run OpenClaw in the cloud without any installation. It's always on, so you can finally close your laptop without killing your agent.

You can join the OpenClaw revolution with a managed OpenClaw.

TRAINING

Distillation Is the New Theft

👀 What’s happening: Anthropic says DeepSeek, MiniMax, and Kimi ran industrial scale distillation campaigns against Claude, citing 24,000 fake accounts and 16 million exchanges. This follows OpenAI’s similar accusation weeks earlier. In parallel, OpenRouter data shows MiniMax and Kimi API calls surging, with Kimi claiming the past 20 days exceeded its entire prior year usage.

🌍 How this hits reality: The growth spike changes the tone. When API volumes jump that fast, incumbents assume extraction, not organic demand. The debate shifts from scraped training data to behavior cloning through endpoints. Instead of copying the internet, rivals can query frontier systems millions of times and compress reasoning gains into cheap replicas. Monitoring replaces dataset policing.

🛎️ Key takeaway: AI competition is moving from data ownership to model output defense. Expect tighter API gating, usage forensics, and export pressure around chips and access. The new moat is preventing your model from becoming someone else’s training set.

STARGATE

Stargate Ambitions Stall

👀 What’s happening: The Stargate joint venture between OpenAI, Oracle and SoftBank has effectively stalled. It has not staffed up meaningfully and has not begun constructing the large scale OpenAI data centers it promised. Negotiations over control, site ownership and financing dragged on for months, including marathon sessions in Tokyo. Lenders pushed back on risk. OpenAI has quietly stepped back from building its own facilities for now.

🌍 How this hits reality: Stargate was framed as a multi hundred billion dollar push toward roughly 10 gigawatts of AI compute in the United States. That capacity was supposed to anchor long term model training and reduce reliance on hyperscalers. With no active buildout, OpenAI remains dependent on external cloud partners. Capital markets are signaling caution about underwriting speculative AI infrastructure at that scale.

🛎️ Key takeaway: If this pause continues, control over frontier compute consolidates with existing cloud giants, and OpenAI’s infrastructure independence thesis weakens, at least in the near term.

DAILY TL;DR

• Amazon will invest $12 billion in AI data centers in Louisiana and says it will cover energy and water infrastructure costs to address local concerns.
• xAI has reached a deal with the Pentagon to deploy Grok in classified military systems under an “all lawful use” standard.
• IBM shares fell nearly 13% after Anthropic said Claude Code can modernize COBOL systems, sparking fears over its mainframe business.
• Goldman Sachs said AI investment contributed “basically zero” to U.S. GDP growth in 2025, as imports offset spending and productivity gains remain limited.
• Guide Labs open sourced Steerling-8B, an interpretable LLM with a built-in concept layer that enables token-level traceability and greater model control.

TRENDING

Daily AI Launches

• Straion has launched a centralized rule manager that automatically feeds the correct context to AI coding agents for faster, enterprise-grade development.
• Tidy has launched a cloud-hosted personal agent (like OpenClaw) that learns to navigate websites without code and reports back via iMessage.
• Wordy has launched a language learning app that tracks your vocabulary growth using real movie clips and built-in quizzes.
• PulseKit has launched an Apple device widget app that puts live key metrics from your business tools directly onto your Home Screen.

Featured AI Tools

• ✨ MuseMail.ai makes creating stunning, on-brand emails from one prompt with AI precision.
• 🪶 CopyOwl is the First AI Research Agent, deep research on any topic in one click.
• 🦾 Flot AI writes, reads, and remembers across any apps and webs.

READ MORE

Let the Future Come to Your Inbox

Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.

• AI Secret uncovers what really matters in AI
• Bay Area Letters decodes tech and business shifts from Silicon Valley
• Robotics Herald tracks how robots move from labs into daily life
• Marketing Secret breaks down real growth and go-to-market playbooks
• The Hardwire explores hardware, consumer tech, and what’s coming next
• Oncely helps you discover the next wave of AI products and launches

TOGETHER WITH US

AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.

We've helped promote over 500 Tech Brands. Will yours be the next?

Email our co-founder Mark directly at mark@aisecret.us if the button fails.