Thursday, May 7 May 7, 2026
- Google shut down Project Mariner and appears to be folding its agent tech into Gemini Agent and AI Mode. - Google also updated AI Mode and AI Overviews with forum/community previews, deeper follow-on links, and subscription-aware source labels. - Anthropic signed a major compute deal with SpaceX for Colossus 1, unlocking more than 300MW of capacity and higher Claude usage limits. - Snap ended its $400M Perplexity partnership, a reminder that AI distribution deals are still fragile. - DeepSeek is reportedly discussing a first outside round that could value it at up to $45B. - New courtroom reporting revived concerns about OpenAI governance, disclosure, and board oversight. Business idea: an AI visibility and citation intelligence platform for publishers and brands to track and improve how they appear inside AI search, chatbot answers, and agentic browsing flows.
Good morning. It’s Thursday, May 7th, and here’s your AI morning briefing.
Today’s theme is consolidation. Experimental agents are getting folded into default products, compute is becoming a product of its own, and flashy AI distribution deals are proving much less durable than they first appeared.
First, Google has shut down Project Mariner, its experimental web agent. The Verge reports the Mariner landing page now says the feature was shut down on May 4th and that its technology “voyaged to other Google products.” In practice, that means the work appears to have been absorbed into Gemini Agent and into AI Mode in Search. Google is signaling that it no longer wants a scattered collection of AI demos. It wants one integrated agent layer inside products people already use.
Second, Google is also changing how AI search itself works. In an official Google blog post published May 6th, the company announced new updates to AI Mode and AI Overviews, including deeper follow-on links, labels for content from a user’s news subscriptions, and previews from public discussions, social posts, and other firsthand sources. TechCrunch framed this as Google adding more context from Reddit, forums, and creator communities. The key shift is that AI search is becoming less like a single answer box and more like a guided discovery interface.
Third, the infrastructure story is enormous. Anthropic announced an agreement with SpaceX to use all of the compute capacity at the Colossus 1 data center, giving it access to more than 300 megawatts of new capacity, or over 220,000 Nvidia GPUs, within the month. Anthropic says that extra capacity is already allowing it to raise usage limits for Claude Code and the Claude API. TechCrunch’s broader read is important too: compute itself is now the product. If frontier labs are renting giant GPU blocks from each other instead of only hoarding them for model training, the AI market starts to look more like an energy market.
Fourth, Snap says its Perplexity partnership is over. TechCrunch reports that the companies amicably ended the relationship in the first quarter, and Snap said in its quarterly earnings materials that guidance assumes no contribution from Perplexity. The original deal, announced last November, was supposed to bring Perplexity’s conversational search into Snapchat and was valued at about 400 million dollars over a year. The takeaway is that AI distribution deals remain fragile even when the use case sounds obvious.
Fifth, DeepSeek may be headed for a major valuation jump. TechCrunch reports that the Chinese lab is in talks for its first outside investment round and could be valued at as much as 45 billion dollars, up sharply from figures closer to 20 billion just weeks earlier. The report, citing the Financial Times and Bloomberg, says state-linked capital may lead the round, with Tencent and Alibaba also reportedly in talks. If that holds, it reinforces two trends at once: open-weight credibility can command massive value, and China is pushing hard to build a more self-sufficient AI stack around domestic capital and hardware.
Sixth, governance is back in focus. The Verge’s court coverage on May 6th surfaced more detail on Sam Altman’s 2023 removal from OpenAI, with reporting that board concerns centered on candor, disclosure, and oversight failures, including the board learning about ChatGPT through public channels. This is not just recycled drama. It is a reminder that in frontier AI, governance failures can quickly become product risk, funding risk, and regulatory risk.
So the short read on the market this morning is this: Google is consolidating agents into search, Anthropic is buying industrial-scale capacity to keep usage growth alive, distribution partnerships are getting stress-tested in public, and governance questions are staying attached to the biggest labs.
One business idea stands out today: build an AI visibility and citation intelligence service for publishers, brands, and large content owners. The product would track how a company appears inside Google AI Overviews, AI Mode, public-forum citation blocks, chatbot answers, and agentic browsing flows, then recommend specific content, markup, community, and subscription tactics to improve inclusion and click-through. The customers would be publishers, ecommerce brands, and high-consideration B2B companies already losing traffic to AI answer layers. Why now is simple: Google is redesigning AI discovery, distribution partnerships are breaking, and everyone needs new instrumentation for machine-mediated attention. The defensibility comes from proprietary longitudinal data on prompt surfaces, citation patterns, and conversion outcomes across multiple AI channels.
That’s the briefing for Thursday, May 7th.
Sources used: TechCrunch AI category; The Verge AI coverage; Google Keyword blog, “5 new ways to explore the web with generative AI in Search,” May 6, 2026; Anthropic announcement, “Higher usage limits for Claude and a compute deal with SpaceX,” May 6, 2026; Snap Q1 2026 investor materials referenced by The Verge and TechCrunch.