Microsoft Takes on Frontier AI with Project Solara, OpenClaw, and More at Build 2026 | Diet TBPN

TL;DR

• 1

  Microsoft is now clearly in the foundation model race: The company introduced MAI Code-1 Flash and MAI Thinking-1, positioning them around efficiency, ROI, and clean enterprise deployment rather than headline frontier benchmarks.

• 2

  Project Solara is Microsoft's boldest and weirdest Build reveal: Satya Nadella showed an Android-based, agent-first OS and badge-like hardware that offloads work to cloud agents, which Ben Thompson called compelling in theory but still basically vaporware.

• 3

  The core Solara thesis is 'cloud hub, devices as spokes': Instead of making the phone the center, Microsoft is arguing that agents work better when context and compute live in Azure and Microsoft 365, especially for enterprises already living in Teams, Outlook, OneDrive, and SharePoint.

• 4

  OpenClaw gives Microsoft an agent framework without owning every layer: Scout, Microsoft's proactive Copilot agent, will run on OpenClaw this summer, and Microsoft plans to contribute its security guardrails back to the open source project.

• 5

  Microsoft is selling trust as much as model quality: The presenters stressed sanitized training data, no distillation from rival labs, and fine-tuning options on Azure so enterprises can customize models without worrying about future IP blowback.

• 6

  The hosts stayed skeptical about frontier status: Alex Heath's read was that Microsoft did not show benchmark chest-thumping against the best OpenAI or Anthropic systems, and the stronger comparison may be Meta's recent smaller-model efforts rather than the absolute frontier.