Hello Builders, issue #9
News from the trenches of AI in the week Jan 17-23, 2026
Hello Builders,
This week was about monetization meeting maturity. Microsoft got “Claude-pilled”—deploying Anthropic’s Claude Code internally while selling GitHub Copilot to customers. OpenAI hit $20B in revenue, 10x growth in two years, and doubled down on “practical adoption.” OWASP released its first security framework for agentic AI—proof that agents are moving from demos to production. Meta killed the metaverse, laying off hundreds and closing three VR studios. And OpenAI poached key talent from Thinking Machines Lab, rattling a $12B startup. The signal is clear: the investment phase is ending, the monetization phase is beginning.
This week’s signal in the noise
Claude Code at Microsoft: Thousands of employees are testing Anthropic’s tool internally. WSJ calls it getting “Claude-pilled.”
OpenAI $20B revenue: 10x growth in two years. 2026 priority: “practical adoption” in health, science, and enterprise.
OWASP agentic framework: First security standard for AI agents. Key risks: agent hijacking, tool poisoning, and memory manipulation.
Meta metaverse retreat: $36B+ spent. Hundreds were laid off. Three VR studios closed. Pivoting to AI wearables.
Thinking Machines exodus: OpenAI poached CTO and two researchers. The $12B startup is scrambling.
Claude Code Goes Viral Inside Microsoft
Microsoft is encouraging thousands of employees—including non-developers—to use Anthropic’s Claude Code for coding tasks. Software engineers are now expected to use both Claude Code and GitHub Copilot and provide comparative feedback. Microsoft sells GitHub Copilot to customers. The fact that it’s running internal pilots with Claude Code signals that Anthropic may have built a better tool. WSJ calls it getting “Claude-pilled.” Claude Code’s agentic harness—its ability to self-correct errors and work around context-window limits—is the key differentiator. For builders: this is validation that agentic coding tools are the new standard.
Link: https://www.wsj.com/tech/ai/anthropic-claude-code-ai-7a46460e
OpenAI Hits $20B Revenue, Doubles Down on “Practical Adoption.”
OpenAI CFO Sarah Friar announced the company crossed $20 billion in annualized revenue—up from $2 billion in 2023. For 2026, the priority is “practical adoption” in health, science, and enterprise. Friar’s message: “Adoption drives revenue, and revenue funds the next wave of innovation. The cycle compounds.” OpenAI is also testing ads and preparing hardware for H2 2026. Separately, OpenAI launched its “OpenAI for Countries” initiative to close the global “capability overhang”—working with South Korea on climate disaster warning systems. For builders: OpenAI is shifting from research lab to revenue machine.
OWASP Releases First Agentic AI Security Framework
The OWASP Top 10 for Agentic Applications 2026 addresses the distinct security risks of autonomous AI agents—from tool use to memory persistence to identity management. As agents move from demos to production, security teams need new frameworks. Existing AI security guidance doesn’t cover agents that take actions, maintain state, and use tools autonomously. Key risk categories include agent hijacking, tool poisoning, memory manipulation, and over-permissioned actions. For builders: if you’re shipping agentic applications, this framework is now your security baseline.
Link: https://www.infosecurity-magazine.com/opinions/turning-the-owasp-agentic-top-10/
Meta Kills the Metaverse, Pivots to AI
Meta is laying off hundreds of metaverse employees, closing three VR studios, and shifting investment toward AI wearables. Meanwhile, Meta’s new AI team has delivered its first models internally—codenamed “Avocado” (text) and “Mango” (image/video). $36B+ spent on the metaverse. Now it’s over. The future is AI glasses, wearables, and agents—not VR headsets. Even Zuckerberg admitted defeat. For builders: Meta’s pivot confirms the industry consensus—AI agents and wearables are the next platform, not immersive VR.
Link: https://www.cnbc.com/2026/01/21/metas-2b-manus-deal-pushes-away-some-customers-sad-it-happened.html
Thinking Machines Lab Implodes as OpenAI Poaches Key Talent
OpenAI hired three researchers from the $12B-valued Thinking Machines Lab, including CTO and cofounder Barret Zoph. Two more researchers left the same week. The exodus rattled investors as the startup seeks funding at $50B valuation with little revenue. OpenAI executive Fidji Simo led the poaching. For startups: even $12B valuations don’t protect you from talent raids. The AI talent war is brutal, and OpenAI is playing offense.
Link: https://www.forbes.com/sites/the-prompt/2026/01/20/inside-openais-plan-to-make-money/
Nature Paper: LLMs Need Agentic Harnesses to Perform
A Nature paper benchmarked 16 LLMs on 293 biomedical coding tasks. Overall accuracy: below 40%. But when researchers added an agentic system that iteratively refines analysis plans before generating code, accuracy jumped to 74%. Raw model capability isn’t enough. The agentic harness—planning, acting, observing, revising—is what makes LLMs production-ready. For builders: scaffolding, tooling, and workflow integration are the new moats.
Link: https://www.nature.com/articles/s41551-025-01587-2
Humans& Raises $480M Seed at $4.5B Valuation
AI startup Humans&, founded just three months ago by researchers from Anthropic, xAI, Google, and Meta, raised $480 million in seed funding at a $4.48 billion valuation—the largest seed round ever. Investors include Nvidia, Jeff Bezos, SV Angel, and Google Ventures. The startup is working on human-centric AI tools for communication and collaboration. For builders: the AI talent premium has no ceiling. Former researchers from top labs can raise billions on reputation alone.
Regarding the topic of the article, your analysis of the AI market shifting to monetization is spot on. Realy insightful, it makes the signal very clear. Appreciate the sharp perspective.