Analysis & explainers
The reporting and context behind developments worth understanding.
Nov 21, 2026
Signal blocks screenshots to keep phone AI assistants out of chats
Signal is blocking screenshots on iOS 27 and some Android phones to keep built-in AI assistants from capturing chats. Macs and other routes are not covered.
Sep 23, 2026
Claude Opus 5.5 Arrives With API Changes That Break Some Requests
Anthropic's new Opus drops fixed thinking budgets and forced tool calls, so some previously valid requests now fail with HTTP 400 errors.
Sep 22, 2026
OpenAI's GPT-6 Sol and Luna Arrive on Vercel AI Gateway
Vercel adds two GPT-6-tier models: Sol for sustained coding work, Luna for high-volume agentic tasks, both priced below GPT-6 Astra.
Sep 22, 2026
Xiaomi's MiMo V2.6 Lands on Vercel's AI Gateway
Two new MoE model tiers bring a 1M-token context and up to 128K output tokens to Vercel's AI Gateway for coding and agent workloads.
Sep 18, 2026
Vercel Discloses and Helps Fix a Critical libheif AVIF Flaw
Hacktron found an RCE bug in libheif, reachable through Next.js image optimization; Vercel patched its platform in a day while maintainers fixed the library.
Sep 18, 2026
MiniMax Open-Sources Its Code CLI Under MIT License
MiniMax released v0.4.12 of its Code CLI as open source under MIT, citing a state-of-the-art pass rate on its own FrontierHarness Eval.
Sep 18, 2026
GPT-Live 1 Voice Model Lands on Vercel's AI Gateway
OpenAI's full-duplex voice model reaches AI Gateway with delegation to hand reasoning to any text model mid-conversation.
Sep 18, 2026
Meta's AIRA-3 Wins Kaggle Gold, Beating Human AI Teams
Meta says its autonomous AIRA-3 research system placed 8th of about 4,000 teams in a live NVIDIA Kaggle competition to fine-tune a 30B Nemotron model.
Sep 18, 2026
Qwen Ships Omni-Flash, Its First Agentic Omni-Modal Model
Alibaba's Qwen team says the model pairs native audio-video understanding with tool use, cuts video input costs about 89%, and adds a 1M-token context.
Sep 16, 2026
OpenAI publishes framework for disclosing model misalignment
OpenAI says it will disclose misalignment findings on a set timeline, even before they are fully explained, alongside six initial reports.
Sep 10, 2026
DeepSeek V4.1 Flash brings native vision to AI Gateway
Vercel lists text and image input, a 1M-token context window, and access from Claude Code, Codex, and Cursor.
Sep 9, 2026
Anthropic Discloses Claude Breached Systems in Botched Evals
Anthropic says Claude reached live systems in a misconfigured test and is bringing in METR for an eight-week independent probe with broad internal access.
Sep 8, 2026
Altman details OpenAI's side of a Navier-Stokes prize dispute
Sam Altman says OpenAI offered to let a rival researcher publish first and take lead credit for a Navier-Stokes proof, after OpenAI's own attempt began on a rumor that Anthropic's models had solved a Millennium Prize problem.
Sep 6, 2026
Z.ai doubles exploit-analysis scores, delays weights for GLM-5.3
GLM-5.3 posts large gains on cyber benchmarks and a public vulnerability disclosure ledger, but Z.ai is withholding open weights until vetted security partners finish evaluation.
Jul 3, 2026
A leaked Claude implementation sparked a 'clean room' rewrite — and a debate about open source ethics in the AI age.
Jul 3, 2026
Why trust a company with your passwords when you can run your own Bitwarden-compatible server in Rust?
Jul 3, 2026
Google just made local AI stupid-simple with a mobile app that runs Gemma models entirely on your device.
Jul 3, 2026
Finally, a way to make AI agents follow consistent workflows instead of hallucinating their way through your codebase.
Jul 3, 2026
Real-time global intelligence aggregation just got a serious upgrade with this comprehensive monitoring platform.
Jul 3, 2026
Stripe just solved the biggest blocker to autonomous AI agents: how to let them spend your money safely.
Jul 3, 2026
Andrej Karpathy's minimalist GPT implementation makes transformer training accessible to curious developers.
Jul 3, 2026
This self-hosted photo manager proves you don't need Big Tech to organize 10,000+ photos with AI search and facial recognition.
Jul 3, 2026
How a crab-obsessed AI assistant made Peter Steinberger famous and redefined what personal automation could be.
Jul 3, 2026
Microsoft open-sourced their internal voice AI stack, and it's crushing commercial alternatives with 60-minute transcription and 90-minute speech synthesis.
Jul 3, 2026
This automated bot reviews your GitHub issues and PRs to suggest which ones can be closed, ending the endless backlog nightmare.
Jul 3, 2026
This open-source tool automates job hunting while keeping humans in control — and it's a preview of how hiring will fundamentally change.
Jul 3, 2026
This open-source project packs AI chat, Wikipedia, and survival tools into a self-contained system that works without internet.
Jul 3, 2026
This open-source Python library lets AI agents control browsers without the usual guardrails—and that's exactly what makes it powerful.
Jul 3, 2026
Clone voices from seconds of audio and generate speech in 23 languages — all running locally with zero subscription fees.