Tuesday, November 3, 2026173 stories

Models, agents, laws and the companies putting AI to work. Dated, sourced, no noise.

AI coding news: tools, teams and results

How software teams use AI coding tools: the share of code AI writes, productivity, quality and what changes for engineering jobs.

AI codingOctober 29, 2026

Anthropic AI code review: Claude writes 90% of Claude Code

Anthropic AI code review now gates a codebase that Claude mostly writes: the company says Claude writes about 90% of Claude Code, InfoQ reported on October 29, 2026. Daily merged changes in its monorepo rose from roughly 2,500 in May to more than 30,000 in September, so Anthropic rebuilt source control and sorted changes into risk tiers. For engineering leaders, it shows where AI coding moves the bottleneck: from writing code to reviewing and shipping it.

5 min read · 4 sources · By Mike Mozg

October 28, 20264 min
GitHub Octoverse 2026 report: agents start over 40% of PRs
The GitHub Octoverse 2026 report says AI agents started over 40% of pull requests in September 2026, up from 17% in March, with 220 million merged PRs a month.
September 28, 20264 min
Linear StyleX migration makes navigation about 30% faster
The Linear StyleX migration took more than 1,000 pull requests over about five months; Linear says navigation got about 30% faster on a mid-tier machine.
September 18, 20264 min
Copilot runtime Rust migration cost $120,000 in AI tokens
The Copilot runtime Rust migration produced over 800,000 lines of Rust in 14.5 weeks; startup fell from 5.25 seconds to 292 milliseconds.
July 10, 20265 min
Datadog AI-assisted migration cuts database costs by 90%
The Datadog AI-assisted migration of Stream Router to PostgreSQL cut a 45-minute operation to about a second and database costs by 90%, using Claude and Cursor.
June 2, 20264 min
Uber AI spending cap sets $1,500 a month per coding tool
The Uber AI spending cap limits each employee to $1,500 a month per agentic coding tool, after Uber used up its 2026 AI budget in four months.
May 8, 20264 min
AI-written code at Airbnb reached 60% in the first quarter
AI-written code at Airbnb made up 60% of engineers' output in Q1 2026, the company says, while its support bot settled 40% of issues alone.
February 19, 20264 min
Stripe Minions coding agents produce over 1,300 PRs a week
Stripe Minions coding agents now produce over 1,300 merged pull requests a week with no human-written code, each one reviewed by an engineer.
December 2, 20254 min
Claude productivity study finds a 50% self-reported gain
Anthropic's Claude productivity study surveyed 132 engineers: they use Claude in 60% of their work and report a 50% productivity boost, but delegate little.
November 24, 20254 min
Claude Opus 4.5 SWE-bench score passes 80% for coding agents
First model over 80% on SWE-bench Verified: the Claude Opus 4.5 SWE-bench result comes with pricing of $5/$25 per million tokens for agents.
November 18, 20254 min
Gemini 3 Pro scores 54.2% on Terminal-Bench 2.0 coding test
Google released Gemini 3 Pro on November 18, 2025, with a 54.2% Terminal-Bench 2.0 score and preview pricing of $2 per million input tokens for developers.

Anthropic AI code review: Claude writes 90% of Claude Code

GitHub Octoverse 2026 report: agents start over 40% of PRs

Linear StyleX migration makes navigation about 30% faster

Copilot runtime Rust migration cost $120,000 in AI tokens

Datadog AI-assisted migration cuts database costs by 90%

Uber AI spending cap sets $1,500 a month per coding tool

AI-written code at Airbnb reached 60% in the first quarter

Stripe Minions coding agents produce over 1,300 PRs a week

Claude productivity study finds a 50% self-reported gain

Claude Opus 4.5 SWE-bench score passes 80% for coding agents

Gemini 3 Pro scores 54.2% on Terminal-Bench 2.0 coding test