Monday, May 11, 2026

🔌 Grok Connectors Open Up Your Apps

🔌 Grok Connectors Open Up Your Apps

Today’s Overview

Good morning, Grok is reaching into your daily apps, Isomorphic Labs is reportedly chasing a huge funding round, and Greece is pushing AI safeguards into its constitution. We also have fresh research on self-improving systems, a math co-pilot that found a proof strategy, and a wave of new tools for browser automation, routing, and dubbing. Let's dive in.

Top Stories

Grok Connectors Bring Apps Into Chat

Grok Connectors is an integration layer that connects Grok to workspace apps. It can read, write, and execute tasks across your tools, with support for custom MCP servers. The launch positions Grok more squarely as a workspace assistant, not just a chat interface.

  • It is built as an integration layer for workspace apps.
  • The system can read, write, and execute tasks across connected tools.
  • It also supports custom MCP servers for extending what it can reach.

Isomorphic Labs Reportedly Seeks $2B+

Google’s Isomorphic Labs is reportedly raising more than $2 billion to expand its Drug Design Engine. The company says the system significantly outperforms AlphaFold 3 on specific tasks. If the round closes, it would be a major signal that AI drug design is still attracting massive capital.

  • The company is reportedly targeting more than $2 billion in new funding.
  • The raise is tied to expanding its Drug Design Engine rather than a broad product reset.
  • The company says the system outperforms AlphaFold 3 on specific tasks.

Greece Moves AI Into Its Constitution

Greece is proposing to add AI protections to its constitution, with a requirement that the technology serve individual freedom. Prime Minister Mitsotakis backed the move while pointing to threats to democracy. It is a notable attempt to bake AI limits into a country’s foundational legal text.

  • The proposal would require AI to serve individual freedom in constitutional language.
  • Prime Minister Mitsotakis tied the plan to threats to democracy.
  • The move would place AI safeguards in the constitution rather than ordinary policy.

Research & Analysis

Anthropic Lays Out Self-Improving AI Risks

Anthropic’s new research arm, The Anthropic Institute, published a formal agenda focused on self-improving AI systems. The institute says it can study Claude usage, internal workflows, and security signals before they reach the wider market. Its roadmap also includes governance questions, economic disruption, and planning for sudden capability jumps.

  • The institute says it sits inside Anthropic so researchers can study Claude usage and internal workflows before they reach the wider market.
  • Its agenda explicitly covers security threats and governance alongside economic disruption.
  • The team even proposes fire drill exercises for sudden capability surges.

DeepMind’s AI Co-Mathematician Posts A New High

Google DeepMind’s agentic co-mathematician, built on Gemini 3.1, is designed to help mathematicians attack unsolved problems by splitting work into parallel research branches. On FrontierMath Tier 4, it hit 48%, more than doubling Gemini 3.1 Pro’s raw 19% score. The system also surfaced a proof strategy that helped resolve an open problem in the Kourovka Notebook.

  • The system uses a coordinator to split work into parallel research branches.
  • On FrontierMath Tier 4, it reached 48% on the benchmark.
  • A rejected output still contained a useful proof strategy that helped solve an open problem.

GitHub Starts Trimming Agent Token Waste

GitHub says it began systematically optimizing token usage across its agentic workflows in April 2026. The team built new auditing and optimization workflows, standardized logging with token-usage artifacts, and identified unused MCP tools as a common source of overhead. It also introduced an Effective Tokens metric to normalize costs across model tiers.

  • GitHub now records per-call usage in a token-usage.jsonl artifact for every workflow run.
  • Its auditors look for workflows with sudden usage spikes and flag unusually expensive runs.
  • One major source of waste is unused MCP tool registrations that get shipped in every request.

Trending AI Tools

  • Codex in Chrome Lets Codex navigate and automate tasks in the browser.

  • Pareto Code A free routing layer that picks the cheapest coding AI above a quality bar.

  • Adject 2.0 Creates hyperrealistic product visuals with AI.

  • Connector.wtf Connects Google Ads, Meta, and LinkedIn to your AI chat.

  • Yeta AI Real-time AI dubbing for YouTube videos.

  • Codex /goal Persisted goals survive restarts, sleeps, and long pauses without re-prompting.

Quick Hits

Keep reading for free

Enter your email. If you're already subscribed, we'll send a sign-in code. If not, you'll subscribe in the next step.

Free access. Subscribe once, then use the same email on future issues.

Free to read. Subscription just unlocks the full issue.