AI Coding Assistants

Today's Stories

1. 1

   Microsoft's Copilot adoption stalls: just 1% of 365 users actively engaging weekly

   Microsoft revealed that Copilot 365 has more than 20 million paid seats out of its more than 450 million paid commercial Microsoft 365 seats—fewer than 4.5%. Enterprise surveys show weekly usage among licensed Copilot seats is only 20% to 30%, translating to roughly 4 million to 6 million weekly active users. Despite massive distribution across Windows 11, Edge, Word, and other Microsoft products—including a dedicated Copilot key on new laptops—the company's push to make AI a daily habit has not resonated with the vast majority of its existing customer base. Most employees who have access to paid Copilot do not use it regularly, suggesting that ubiquitous placement alone does not drive engagement.

   Microsoft raised Microsoft 365 prices in the US this month, increasing Business Basic from $6 to $7 and Business Standard from $12.50 to $14 monthly. The company also introduced new paid subscriptions for Microsoft 365 Business Standard and Premium with Copilot at $23.50 and $32 per user each month, even as adoption struggles. Microsoft has begun offering users the option to hide the Copilot button and allowing some organizations to uninstall the Windows app.

2. 2

   Musk's SpaceXAI, Cursor merge to speed AI model deployment

   Cursor and SpaceXAI announced they are working together and plan to complete a $60 billion（約9.6兆円） acquisition. The two companies began collaborating in April. Grok 4.5, a new model designed to work within Cursor, has been released as part of the partnership. Musk realized his AI models need a platform to be useful for agentic purposes (where AI systems make decisions and take action independently). Competitors like OpenAI, Anthropic, and Google have their own platforms for this purpose. By acquiring Cursor, SpaceXAI gains immediate access to a developer platform, rather than building one from scratch.

   Once SpaceXAI likely merges with Tesla, the question is whether Musk can maintain his usual pace across all his ambitious projects. He spun up xAI's data center in record time and nearly caught up to the frontier by last year, but then had to play catch-up again when he realized the platform gap.

3. 3

   GitHub improved Copilot code review by rewriting tool instructions, not swapping tools

   GitHub migrated Copilot code review to use shared code exploration tools (grep, glob, view) from the Copilot CLI harness, expecting better performance. Instead, benchmarks showed higher costs and fewer issues caught. After rewriting the tool instructions to match how reviewers actually read pull requests—starting from the diff and using narrowly targeted searches—the regression flipped into a win: roughly 20% lower average review cost while maintaining the same review quality. The fix reveals that agent performance depends less on tool quality than on how those tools are instructed to behave. The shared tools were sound, but their original instructions guided the agent toward broad repository exploration instead of focused diff-based review, wasting context tokens. For teams relying on Copilot for code review, this means the product should now catch issues more efficiently without adding cost.

   GitHub's reframing of tool instructions for code review—from generic coding-assistant guidance to review-specific workflows—suggests a broader principle for agent design: the same tools can fail or succeed depending on the instructions shaping their use. The company framed this as a lesson in agent-framework design, indicating it may apply these insights across other Copilot products that share the CLI harness.

4. 4

   GitHub's Copilot code review cuts cost by reshaping agent workflows

   GitHub migrated Copilot's code review feature to use shared Unix-style code exploration tools, which reduced review cost by reshaping how the agent performs pull request analysis. The shift shows that smarter tooling can sometimes force a rethink of how AI agents work — moving away from a direct approach to one centered on the actual evidence in pull requests. For development teams, this suggests code review assistance may become more efficient without sacrificing quality.

   The change demonstrates a design principle: better tools don't automatically improve outcomes; they require rethinking the agent's workflow to extract full value from them.

5. 5

   Ollama raises $65M in Series B to expand open AI model platform

   Ollama Inc. announced a $65 million（約100億円） Series B funding round led by Theory Ventures, with participation from Benchmark, 8VC, Y Combinator, Pace Capital, 49 Palms, GTMFund, and other investors and angels. Ollama positions itself as the largest AI platform connecting developers to open models, meaning this funding supports infrastructure that helps developers access and work with non-proprietary AI tools rather than relying solely on closed commercial offerings.

   The funding brings Ollama's total raised to a new level, though the body does not disclose the company's valuation or prior funding amounts, so the comparative scale of this round remains unclear.