Gemini 2.5 goes generally available with a thinking dial
Pro and Flash hit GA, Flash-Lite arrives, and every tier exposes a thinking budget.
page 10
older dispatches from the README archive.
Meta buys half of Scale AI and most of its leadership
$14.3 billion for 49%, and Alexandr Wang moves to run a new superintelligence lab. The data layer just got picked.
WWDC 2025: Liquid Glass, a version number reset, and a local model API
Apple renumbers every OS to 26, redesigns everything, and quietly ships the most developer-relevant thing in years.
The Illusion of Thinking, and the argument about what reasoning is
An Apple paper finds reasoning models collapse past a complexity threshold. The rebuttals are as instructive as the paper.
Your logs are a product and you are shipping a bad one
Structured events, one line per unit of work, and the specific reason your grep-based debugging is slow.
OpenAI buys a hardware company that hasn't shipped anything
$6.5 billion for io, Jony Ive's design studio. The bet is that the phone is the wrong shape for this.
Claude 4 and the agent that runs for hours
Opus 4 and Sonnet 4 ship with a focus on long-horizon work, and Claude Code goes generally available.
I/O 2025: Google puts AI in the search box and means it
AI Mode, Veo 3 with audio, Jules, and an Android XR headset. The distribution advantage, deployed.
Build 2025: Microsoft open-sources WSL and puts MCP in Windows
A decade-old request granted, a coding agent in GitHub, and an agent protocol shipped at the OS layer.
Codex, and the agent that opens pull requests
OpenAI ships a cloud software engineering agent that works in a sandbox and hands you a diff.
Rust 1.87, and the standard library grows an anonymous pipe
std::io::pipe, asm! jumping to Rust code, and the tenth anniversary of 1.0.
The test suite is a design document
If your tests are hard to write, your design is wrong. That's the signal, and most teams ignore it.
Node.js 24 and the slow reinvention of the runtime
V8 13.6, npm 11, fetch no longer experimental, and the permission model losing its dashes.
Stack Overflow's traffic fell off a cliff and it is not coming back
Question volume down roughly three quarters from peak. The knowledge commons has a serious problem.
Qwen3 ships a whole family under Apache 2.0
Eight models from 0.6B to 235B, hybrid reasoning modes, and a genuinely permissive license across all of them.
Shopify's memo and the new hiring question
"Prove AI can't do this job" as a requirement before headcount. A reasonable policy with an unreasonable failure mode.
Chrome keeps third-party cookies after all
Five years, a Privacy Sandbox, a regulatory process, and the answer is: nothing changes.
o3 and o4-mini put tools inside the reasoning loop
The models can now search, run Python, and look at images while they think. That's an architecture change, not a feature.
Vibe coding is real, and it is not what you think
The term went viral as a joke about not reading the code. The actual practice is a skill, and it has a failure mode.
Google Cloud Next: Ironwood, A2A, and a very cheap Gemini
A seventh-gen TPU aimed squarely at inference, plus a protocol for agents talking to other agents.