AI Daily Dev — July 7, 2026

Alberta Scans 466M Lines of Code With Claude in 20 Hours — 6.5 Years' Work

Anthropic case study out July 6: Alberta's Ministry of Technology and Innovation ran ~50 Claude Code agents in parallel across all 27 provincial ministries.
Coverage: 1,280 applications and 3,400 code repositories — every repository Alberta owns.
Two-stage routine: rules-engine flag pass, then Claude review that cites the exact file and line so developers can verify each finding.
Alberta says the same work would have taken about 6.5 years of manual review; the province is publishing white papers so other governments can copy the setup.

industry anthropic.com

ByteDance's Doubao and Alibaba's Qwen will disable all user-created AI agent features on July 15 to comply with China's new Interim Measures for Anthropomorphic AI Interaction Services.
Every existing user agent stops functioning that day; Doubao redirects builders to Maoxiang, ByteDance's separate character-app product.
Configurations and chat histories stay viewable during a transition, then get purged from platform servers on October 15.
Front-line evidence that China is willing to shut down live consumer AI features to enforce its personification rules — the first big regulatory kill of a mainstream Chinese AI product line.

industry technode.com

Two-day inaugural session runs July 6–7 in Geneva, convened by the UN General Assembly with every member state at the table.
Guterres framed AI as no longer a technology issue but a geopolitical one; independent scientific panel warned no technical guarantee of safety currently exists.
Draft priorities: shared testing and risk-measurement so 'safety travels with the technology,' plus a proposed AI Child Safety Pledge.
Session two lands in New York in May 2027; ITU's AI for Good Global Summit picks up in Geneva through July 10.

industry un.org

Pro, Max, Team, and select Enterprise plans have Fable 5 included up to 50% of weekly limits through July 7 (today).
From July 8 the model bills only through usage credits at $10 / $50 per Mtok — double Opus 4.8, ~5× Sonnet 5's intro rate.
No automatic metered fallback: if credits are not enabled in Settings → Usage before the pool empties, Fable 5 access just stops mid-session.
Standard Enterprise seats get zero grace period and were on credits from day one; teams relying on Fable 5 for July work are being told to set a cap now.

models bleepingcomputer.com

Announced July 6 — one API, one voice, every language: no per-language variants.
Voices cast for specific jobs: support, characters, commentary, advertising, education; original five upgraded for pacing and phrasing.
Inline speech tags like [pause] and <whisper> give developers explicit delivery control at generation time.
Available in the realtime Voice Agent API, the standalone TTS API, and the no-code Grok Voice Agent Builder shipped last week.

tools x.ai

Official OpenAI plugin lets Claude Code delegate reviews and tasks to Codex without leaving the terminal — one command install.
Re-took #1 on GitHub trending this weekend, ahead of JuliusBrussee/caveman (token compressor) and alibaba/page-agent.
Notable as a first-party cross-lab integration: OpenAI shipping tooling that treats Anthropic's coding agent as the shell it plugs into.
Companion piece to Anthropic's Claude for Chrome GA and background-agent PR-drafting from last week — the coding agents are converging on each other's surfaces.

open-source github.com