🤖 The Blockbrain Brief #13

Agents everywhere: Gemini goes personal, Claude goes hands-on, OpenAI boosts low-latency compute, and the AI boom gears up for mega-IPO scale.

📩 High-level Summary

• Personal AI Gemini: Google’s new Personal Intelligence connects Gemini to Gmail, Photos, YouTube, Search so it can reason across your data.
• Claude Cowork: Anthropic’s Cowork brings Claude Code-style agent actions to non-technical users to read/edit/create files in a folder, queue multi-step tasks, and run work more like a “coworker” than a chat thread.
• Anthropic Labs: Anthropic is doubling down on an internal incubator model prototype fast, ship frontier features, then scale
• OpenAI × Cerebras: OpenAI’s partnership adds massive ultra low-latency capacity, another signal that speed and inference scale are becoming competitive features, not backend details.
• PayPal × Copilot: Checkout is moving inside assistants: PayPal powering inventory & payment flows within Copilot, pushing “agentic commerce” from demo to distribution

Product Releases

Gemini launches "Personal Intelligence" 👀

Google launched Personal Intelligence (US beta for now) that lets Gemini reason across connected Google apps (Gmail, Photos, YouTube, Search). It’s designed for multimodal recall and reasoning. For example, Gemini can now pull a license plate from Photos, or find a vehicle trim detail via Gmail, then combine that with recommendations.

Privacy-wise, Google emphasizes it’s off by default, users choose which apps to connect, and it doesn’t train directly on your Gmail/Photos content. Instead, it may train on limited Gemini interaction data with steps to filter personal info.

Anthropic launches Cowork (Claude Desktop)

Anthropic introduced Cowork, inside the Claude macOS app for Claude Max subscribers. It is compared to Claude Code, but without the "code". Cowork can give Claude access to a specific folder, and it can read/edit/create files there turning agentic workflows (reports, spreadsheets, organizing files) into a chat-driven experience.

Cowork lets you queue tasks and run them in parallel, and it’s meant to feel less like back-and-forth and more like “leaving messages for a coworker.” On the safety side, Anthropic explicitly warns about destructive actions (like deletion) and prompt injection risk in agentic settings.

PayPal checkout inside Microsoft Copilot

PayPal and Microsoft are pushing “agentic commerce” forward: PayPal will power merchant inventory surfacing, branded checkout, guest checkout, and card payments inside Copilot.com, so shoppers can complete purchases without leaving the Copilot experience.

Microsoft also shared early performance signals for Copilot shopping journeys (e.g., higher likelihood of purchase when Copilot is part of the journey) and noted rollout begins in the U.S. with partner activation across PayPal, Shopify, and Stripe.

Infrastructure and Partnerships

Anthropic expands “Labs” incubator

Anthropic is scaling its internal incubator, Labs, to rapidly prototype Claude-powered products. In their announcement, they cite Labs-style experimentation as the engine behind Claude Code, MCP (Model Context Protocol), Skills, and Cowork; they specifically call out Claude Code scaling from research preview to a “billion-dollar product” in six months, plus MCP hitting “100M monthly downloads.”

OpenAI partners with Cerebras for low-latency inference at massive scale

OpenAI announced a partnership with Cerebras to add 750MW of “ultra low-latency” compute into OpenAI’s platform and integrate it into their inference stack in phases across workloads. They aim for faster responses, that organically lead to high-value workloads.

Cerebras describes it as a multi-year agreement rolling out in stages starting in 2026, framing speed as a key driver of AI adoption. Reuters additionally reported the deal could be worth more than $10B and deploy capacity in phases through 2028.

Markets and Funding

2026 may be the year of the Mega IPO

"If SpaceX, OpenAI and Anthropic go public, they will unleash gushers of cash for Silicon Valley and Wall Street"

NYT believes 2026 as a possible mega-IPO year, led by OpenAI, Anthropic, and SpaceX. OpenAI and Anthropic are reportedly in early IPO prep, while SpaceX is furthest along. It cites huge private valuations: Anthropic ~$350B, OpenAI ~$500B, SpaceX ~$800B.

However, IPOs are still recovering ($142B raised in 2021 vs $44B last year), and NYT argues one blockbuster could reopen the window especially since AI firms need massive capital for compute. There are developments of OpenAI with $60B+ raised and planning $115B spend through 2029; Anthropic $40B+ raised and pursuing $10B more.

xAI raises $20B Series E

xAI announced it completed an upsized $20B Series E (above a targeted $15B), naming major investors and noting strategic participation (including NVIDIA and Cisco Investments) to support scaling compute infrastructure.

A separate recap notes the raise supports building massive GPU clusters and continued Grok product development (including voice and upcoming models). On xAI’s roadmap, the big focus areas are:

• Grok 4 model line: Continuing their next-gen model releases and improvements.
• Grok Voice: Low-latency speech in dozens of languages, with tool-calling + real-time data access—used across the Grok app, Tesla vehicles, and X.
• Grok Imagine: Image and video generation capabilities.
• Grok 5: Already in training.

OpenAI invests in Merge Labs (BCI)

OpenAI is participating in Merge Labs’ seed round, framing brain-computer interfaces as a next interface frontier. OpenAI says it plans to collaborate on scientific foundation models and AI systems that can interpret intent from noisy signals. Merge is pursuing non-invasive approaches and situates it in the broader neurotech trend (with other players exploring similar brain/AI interfaces).