Wall Street says: Google Wins

Plus: Amazon’s Agents Go Wild


Welcome back! If you’ve ever wanted to edit a Wikipedia page but feared the wrath of Very Serious Internet People™️, good news: Grokipedia exists, and literally no one is guarding the door.

xAI just opened its AI-generated encyclopedia to public edits…except the “editor-in-chief” is Grok itself. The result? A swirling grab-bag of half-truths, Musk memes, out there medical advice, and whatever chaos the internet feels like proposing today.

Amazon’s New Agents Don’t Need You Anymore

Via Newsweek

Amazon just unveiled its most ambitious bet on autonomous AI to-date: a new class of “frontier agents” that can work for hours or even days without human input—tackling software development, security, and operations end-to-end.

The three long-running agents are designed to behave like full-fledged teammates:

  • Kiro, an autonomous coding agent that keeps context across tasks; learns from your repos, PRs, and docs; and can work independently until it needs human review

  • AWS Security Agent, which embeds security checks throughout development and can run penetration tests in hours instead of weeks

  • AWS DevOps Agent, which watches incidents in real time, traces root causes, and acts like a veteran SRE plugged directly into your observability stack

What this means for engineering teams: Amazon insists these agents won’t replace developers—but they will change the craft. The focus shifts from writing every line of code to orchestrating agents, setting up repositories and rules, and curating knowledge bases the agents depend on. Early internal projects reportedly finished in 78 days instead of 18 months.

  • The next frontier: formal verification, deeper multi-agent coordination, and automated property-based testing.

  • Kiro already extracts testable properties from specs and generates thousands of scenarios automatically—far beyond what a human team could manually write.

The bigger picture: If autonomous agents can learn software development, Amazon believes they can learn other complex domains, too. These three preview agents are just the starting point.

Wall Street Just Picked a New AI Winner

The AI hype cycle isn’t lifting every boat anymore—and the stock market is making that painfully clear. Two stories this week show how quickly sentiment can swing when Wall Street thinks a company is either riding the right wave…or falling behind.

First: Microsoft stumbled. A report from The Information claimed Microsoft had quietly lowered growth targets for its AI enterprise products after many Azure sales teams missed their numbers—especially for Foundry, Microsoft’s platform for building and running AI agents. 

But then: Google surged. Alphabet is suddenly the market’s AI darling. Since releasing Gemini 3 in mid-November (and scoring huge benchmark wins) Google’s stock has outperformed nearly all of the Magnificent 7. Its partner, Broadcom, is rallying too, thanks to rising interest in Google’s custom TPUs. 

What’s driving the shift?

  • Google’s TPUs are looking increasingly competitive with Nvidia’s GPUs.

  • Gemini 3’s performance surprised everyone, including OpenAI.

  • Rumors that Meta may test Google’s chips spooked investors.

  • OpenAI’s “code red” push to improve ChatGPT reads like a reaction to losing ground.

The bigger picture: AI is no longer a monolithic trade where everything goes up. The Nasdaq 100’s average pairwise stock correlation just hit 14%—the lowest ever recorded. Translation: Investors are finally distinguishing between leaders, laggards, and hype.

The Fight Over Who Gets to Regulate AI Isn’t Over

Another attempt to jam a federal ban on state-level AI regulation into a must-pass defense bill just failed…again. GOP leaders had tried to preempt states from writing their own AI rules, a move Silicon Valley strongly supports, but bipartisan pushback shut it down this week.

Why it’s contentious: States have been filling the AI regulation vacuum left by Congress, passing laws focused on safety, transparency, and consumer protections. Critics say blocking them without a federal replacement would hand Big Tech a regulatory free pass. Supporters argue a patchwork of rules will crush innovation.

What’s next: House Majority Leader Steve Scalise says legislators will try to introduce the ban elsewhere, and a leaked draft executive order suggests President Trump may attempt to act on his own—though those efforts appear paused, for now.

AI-powered browser tests for every PR

Via DebuggAI

DebuggAI is an AI-powered, zero-config testing platform that scans every pull request, explores your app with autonomous agents, and generates targeted end-to-end browser tests automatically. 

How you can use it:

  • Auto-generate test coverage for new features

  • Catch UI/UX regressions before review

  • Speed up code reviews with instant test insights

Pricing: free and paid

Real-time collaboration features in one SDK

Via Velt

Velt is a collaboration SDK that lets developers instantly add multiplayer features—comments, live cursors, presence, recordings, and more—directly into their apps.

How you can use it:

  • Build real-time collaborative editing and annotation

  • Capture product feedback with AI-generated summaries

  • Increase engagement with live cursors and shared presence

  • Accelerate development with plug-and-play collaboration tools

Pricing: free and paid

Jobs, announcements, and big ideas

  • Kling AI introduces Avatar 2.0 with more lifelike expressions and character realism.

  • Google rolls out Gemini 3 Deep Think mode for AI Ultra subscribers in the Gemini app.

  • Yahoo debuts an AI system that delivers real-time football recaps far beyond standard box scores.

  • Anthropic releases Interviewer, sharing insights from 1,250 professionals working with AI.

  • Google unveils Workspace Studio to automate workflows and tasks using Gemini 3 AI agents.

  • Amazon adds an Alexa Plus feature that lets Fire TV users jump directly to specific movie scenes.

OpenAI sounds the alarm: Follow along as I break down OpenAI’s code red motion and what it means for the future of AI.

That’s a wrap! See you next week.

—Matt (FutureTools.io)