Mighty Brief: April 11, 2026
Start Here
AI security is consolidating into the platform layer: The Information says Cisco is in talks to buy Astrix Security for roughly $250M to $350M. If that lands, the agent-security wedge is moving from startup novelty to large-vendor control plane.
Anthropic's coding story is now the business story: The Information's Boris Cherny profile frames Claude Code as the growth engine, not a sidecar. That matters because coding agents are becoming the distribution layer for frontier models inside companies.
Enterprise AI adoption is widening, not concentrating: Techmeme flagged FT data from Ramp showing Anthropic paid adoption among US businesses jumped to 30.6 percent in March from 24.4 percent in February, while OpenAI stayed roughly flat around 35 percent. Translation: the market is opening up, not locking in.
The operator lesson from Waymo is utilization, not coverage: Ryan Petersen argued for saturating one city. Bill Gurley pushed back that demand swings 4x peak to trough, so skimming at lower penetration is more economic. Same mental model applies to agent rollouts: optimize utilization before land-grab instincts take over.
The emerging scarcity is proprietary data and workflow exhaust: Every's piece on the market for making AI better is basically Wick's thesis in another costume. Value is shifting toward the operational data and environments that improve models, not just the models themselves.
Blank Metal
Direct BM inbox signal is blocked this morning: gog gmail search --account dan.wick@blankmetal.ai 'is:unread newer_than:1d' failed on April 11, 2026 with Google OAuth invalid_grant and invalid_rapt. That is auth debt, not evidence of a quiet pipeline.
Best active view still comes from memory: HR&A remains the clearest near-term monetization path, Rally is still the warmest founder-led lead, and Tungsten, Great Hill, Sentinel, Orbis, and Care Providers of MN are still sitting in the follow-through bucket.
The strategic packaging question remains the same: BM demand is showing up around rollout, training, and workflow design. The risk is not lack of leads. It is letting the offer stack stay fuzzier than the market window.
Munger Observer
Opportunity Cost: Bitfarms got captured while older Wick items are still open. New inbound work is fine. Quietly carrying stale strategic inventory is not.
Second-Order Thinking: One meeting captured is first-order progress. The second-order test is whether capture becomes a decision, or just another layer of storage.
Personal Thought
The tension this morning is still compounding versus scattered. Public signal keeps validating Dan's thesis, but too much of the upside still lives in private notes, private client work, and private conviction. That mismatch keeps getting more expensive.
Quick Scan
- Stratechery's weekly bundle is still centered on Anthropic, the New York Times, and Altman. The meta-point is that model capability, distribution, and narrative power are all colliding now.
- SK Hynix is getting framed as the overlooked memory winner. Same old truth: when demand gets weird, the picks-and-shovels story migrates down-stack.
- Garry Tan's OpenClaw plus GBrain setup post is a useful reminder that we are still in the Altair BASIC era of agents. Painful setup is a feature of the phase, not a bug in one product.
- Social noise around Altman is high this morning. The more durable signal is still enterprise adoption, security spend, and coding-agent distribution.