July 13, 2026
Merge requests, hurt feelings
A Study of Microsoft's Early 2026 Rollout of Claude Code and GitHub Copilot CLI
Microsoft says AI coding tools boosted output — commenters say, “Cool, but did anything better get built?”
TLDR: Microsoft found that engineers using AI coding assistants got more code changes approved, and that coworkers heavily influenced who started using them. But commenters argued the real question is whether this created anything actually valuable — or just a very expensive pile of extra work.
Microsoft rolled out two artificial intelligence coding helpers to tens of thousands of engineers and came back with a flashy stat: people who adopted them ended up merging about 24% more code changes over four months. The company’s big takeaway was that these tools spread mostly through office buzz and peer pressure — basically, if your teammate used it, you probably tried it too. But in the comments, the crowd was not exactly throwing confetti.
The loudest reaction was a giant, skeptical “So what?” One camp argued that more code changes getting approved does not automatically mean better products, happier customers, or more money. Commenters wanted the real scoreboard: Did bugs go up? Did features ship faster? Did profits change? One reader practically rolled their eyes at the headline-worthy framing, calling the result “not exactly headline-making material.” Ouch.
Then came the money panic. Readers zeroed in on the eye-watering token costs behind these tools, with one comment warning that usage at scale could hit millions of dollars and maybe even more than the sticker price suggests. That sparked the thread’s favorite running joke: companies paying luxury prices for machines to create more pull requests nobody asked for. The vibe was half curiosity, half budget horror, with a side of meme-worthy disbelief. In short: Microsoft says the bots aren’t a fad, but the comment section wants receipts, not vibes.
Key Points
- •The article examines Microsoft’s early-2026 rollout of Claude Code and GitHub Copilot CLI across tens of thousands of engineers.
- •It states that first use of the tools spread primarily through social networks within the organization.
- •Retention was associated more with engineers’ coding activity than with demographic characteristics.
- •The study used merged pull requests as its proxy for output and acknowledged that merged PRs do not directly measure delivered value.
- •Adopters were reported to merge roughly 24% more pull requests than they otherwise would have, with the effect persisting over four months.