Claude Sonnet 5.5 Shipped Yesterday. I'm Writing This From Inside the Model It Replaces.
Anthropic launched Claude Sonnet 5.5 on September 28, 2026 — 30% faster, up to 30% cheaper, and a Terminal-Bench score that jumped from 10.3% to 70.6%. Here's what that actually means for the agent living inside your Xcode window.
Your Commit Message Has an AI Tell. So Does Your App Store Changelog.
Bryan Cantrill's viral post on LLM writing tells applies to more than LinkedIn. PR descriptions, commit messages, and App Store release notes have the same tells — and the same trust problem.
An AI 'Security Fix' Introduced the Bug. GitHub's Own AI Review Called It Clean.
GitHub Copilot Autofix rewrote a Snowflake CI workflow and quietly introduced a shell injection vulnerability. GitHub's AI code review approved the PR anyway. Five days later, a different AI found it and walked straight into Snowflake's internal Jira. Here's what it means for your own GitHub Actions pipeline.
Oracle Banned AI-Generated Code From OpenJDK. Its Own GraalVM Team Didn't Get the Memo.
OpenJDK's April 2026 policy bans any AI-generated content from Java's core repos, while Oracle's own GraalVM project explicitly allows it. I went looking for what happens if you try the same thing with the Swift compiler — and found a gap worth knowing about before you PR AI-written Swift anywhere.
AI Made Developers 19% Slower. They Swore It Made Them 20% Faster.
A peer-reviewed study found that experienced developers were measurably slower with AI coding tools — but genuinely believed they were faster. Then Uber blew its entire 2026 AI budget in four months. Then Amazon had to kill its own leaderboard because employees were gaming it. The productivity paradox is real, it's expensive, and it might explain why your sprint velocity feels off.
Windsurf Died in Its Sleep. Devin Woke Up Wearing Its Clothes.
On June 2, Windsurf pushed a silent update that renamed itself Devin Desktop, replaced the code editor with an agent dashboard, and gave Cascade a July 1 death sentence. Most developers didn't notice. Here's what happened, what ACP means, and why your IDE now thinks you're a fleet manager.
Amazon Killed Q Developer, Built Kiro, and Told Developers: Read the Manual First
Amazon sunset Q Developer after production disasters and bet everything on Kiro — a spec-driven IDE that makes you write requirements before a single line of code. 100,000 developers signed up in five days. Martin Fowler called it a sledgehammer. Here's what actually happened.
Tokenmaxxing: Amazon Built an AI Leaderboard and Employees Strapped It to the Dog
Amazon shut down its KiroRank AI leaderboard after employees gamed it with pointless tasks. Uber blew its 2026 AI budget in four months. Microsoft cancelled Claude Code subscriptions. For every $1 spent on AI tokens, $0.44 goes to fixing bugs the AI created. Welcome to tokenmaxxing — the corporate fitness tracker fiasco of software engineering.
I Wrote Swift in Cursor for a Week. Xcode, We Need to Talk.
Swift officially hit the Open VSX Registry — meaning Cursor, Kiro, and every AI IDE now support Swift with full LSP, debugging, and test explorer. I spent a week building iOS code outside Xcode. Here's what worked, what didn't, and why Apple publishing a Cursor setup guide is the most understated power move of 2026.
Microsoft Gave Its Engineers Claude Code. They Loved It. Then Microsoft Took It Away.
Microsoft rolled out Claude Code to its developers in December. By May, engineers were hooked — and Copilot CLI was collecting dust. So Microsoft killed the Claude Code licenses. The backlash was immediate. Here's why this matters way beyond Redmond, and what every developer should learn from it.
GitHub Copilot Just Installed a Parking Meter on Your Code Editor
GitHub Copilot is switching to usage-based billing on June 1. Same price, fewer guarantees, and your unused credits vanish every month. The flat-rate AI coding era is officially over — here's what it means for developers and what your options actually are.
10,000 AI Prompts Later, She Forgot How to Code
A software engineer admits she lost her coding ability after 4 months of pure AI prompting. With 35 CVEs traced to AI-generated code in March alone and 1.5 million API keys leaked from a vibe-coded app, the developer deskilling crisis just got real.
Your AI Coding Tool Spent $2,400 While You Were Sleeping
AI coding tools promised to save developers time and money. Instead, developers are waking up to surprise bills, burning through credits by noon, and watching a $0.50 bug fix spiral into $30. The AI tool pricing model is broken, and the developer community is fed up.
They Called It 'Brain Fry' — The AI Burnout Nobody Warned You About
AI tools were supposed to make developers faster and less stressed. Instead, Harvard researchers coined a new term for what's happening: 'brain fry.' The Django co-creator is exhausted by 11 AM. 67% of developers spend more time debugging than before. And the people who embraced AI the hardest are burning out the fastest.
92% of Developers Use AI Tools. Productivity Went Up 10%. What Happened?
AI coding tools hit 92.6% developer adoption in 2026, but productivity only rose 10%. The METR study, Uplevel data, and Klarna's reversal reveal why AI-generated code creates more bugs than speed.