Google says Search is being upgraded into an AI-powered query box where users can follow up directly and let information agents handle more of the searching, summarizing, and organizing behind the scenes. Read alongside TechCrunch’s framing, the key shift is that Google is trying to move AI into the default search flow itself, not keep it in a separate mode.

OpenAI launched Guaranteed Capacity with one-to-three-year commitments, spend-based access certainty, and drawdown across multiple product surfaces. This is not a model launch story. What matters is the product shape itself: access certainty is now being sold as a defined part of the enterprise offer.

According to The Information, Anthropic and OpenAI now generate 89% of revenue across the 34 leading AI startups it tracks. If that figure holds, it sharpens an uncomfortable point for the application layer: a large share of revenue in that cohort is still flowing toward the model layer, which means app builders will need to show much more clearly what workflow or distribution advantage they actually own.

Cursor now lets users assign work items directly in Jira or mention @Cursor so the cloud agent can scope the task from the ticket, comments, and repo context. The notable move here is that the agent is being pushed into the ticketing layer, not kept inside the IDE.

OpenAI previewed a public verification tool for provenance signals like Content Credentials and SynthID across images generated by ChatGPT, the API, and Codex. What stands out is that provenance is being turned into a public-facing tool that users and platforms can touch directly.

xAI says SuperGrok and X Premium subscriptions can now be used inside OpenClaw, a local-first open-source agent that runs on users’ own hardware. The notable signal is that a model provider’s subscription product is being pulled into an agent wrapper that lives outside the first-party shell.

Google says Managed Agents can spin up agents inside isolated Linux sandboxes, browse the web, execute code, and define custom behavior through AGENTS.md and SKILL.md. For builders, the signal is clear: sandboxing and orchestration are being packaged as their own API surface.

Google revealed Android XR eyewear with Gemini built into glasses from Gentle Monster and Warby Parker. The key thing to note is that Google is still pushing the assistant into a hardware form factor that does not require a clear app-launch moment.

TechCrunch describes Gemini Spark as a cloud-based personal AI agent with Gmail, Docs, Workspace integrations, and Chrome-powered web access. If that description holds, the notable angle is how tightly Google is tying its personal-agent story to Workspace distribution.

Anthropic lays out four operational priorities for computer-use systems: click accuracy, thinking effort, long-session context control, and recording demonstrations agents can replay.

The practical value is straightforward: this gives you a production checklist instead of another browser-agent demo to admire from a distance.

If you are building with browser-use or computer-use workflows, the four-part pass is simple:

The most useful part of this checklist is that it forces a team to review a UI agent like a real system: are the clicks reliable, is the reasoning worth the cost, does context drift, and can a good run be replayed.

For a small team, that is often enough to separate an impressive demo from a workflow that is actually starting to work.

Managing your cap table doesn’t have to be complex. Pulley simplifies equity management for Founders and CFOs with intuitive workflows, accurate, audit-ready reporting, and predictable pricing—so you can plan and scale without surprises. Onboard in days, not weeks, and rely on responsive, expert support every step of the way.  

From issuing grants to 409A valuations or ASC 718 reporting, Pulley gives you the clarity to manage equity, make decisions, and get back to work. Experience a platform built for business owners and finance teams: transparent, reliable, and designed to put your company first.

See Pulley in action

TechCrunch reads Google’s latest updates as a fairly clear end to the “ten blue links” era, with the query box, follow-up flow, and information agents becoming the main interaction pattern. This is the reading to keep if you want an external lens on today’s lead story.

This paper measures more than 55,000 queries to track how AI Overviews change source selection, claim fidelity, and publisher impact. Its value is that it adds data to a topic that is often discussed through intuition and rhetoric.

OpenAI argues that provenance needs to be verified through multiple signals and that perfect detection is not realistic. The reading matters because it shows the vendor itself framing provenance as a product and operations problem, not just a policy ideal.

WIRED uses its conversation with Hassabis to push a fairly clear labor thesis: AI should expand what people can do, not simply cut headcount. Next to the Search, agent, and managed-runtime stories, this helps clarify the message frontier labs want the market to hear.