Anthropic Daily

Monday, July 6, 2026

Alberta scanned 466 million lines of provincial government code in 20 hours with Claude Code, and v2.1.202 shipped a workflow-size config and 14 bug fixes.

The Lead

466 million lines. 20 hours.

Alberta's Ministry of Technology and Innovation built a two-agent security review system and ran it against the province's entire government codebase. A red team agent probed from outside as an attacker would, mapping exploit paths. A blue team agent assessed defenses against international security standards and wrote remediation plans pointing to exact files. Result: 466 million lines scanned, vulnerabilities remediated, new tooling built. Time elapsed: 20 hours.

The province says the system sped up work by as much as 20 times and cut modernization timelines by up to 95 per cent. An industry day in Edmonton this month will let other governments copy the playbook. Province-wide scale-out begins fall 2026. Anthropic published the case study on July 6.

This is not a headline about AI potential. It is a headline about a government that shipped.

Back of Page

The Release Log

Confirmed items from the 24-hour window ending July 6, 2026, 21:00 ET.

Claude Code

1 release

Jul 06

CODE

Claude Code v2.1.202

Dynamic workflow size config, two new OpenTelemetry attributes, and 14 bug fixes. Mostly maintenance, with one operator-facing addition.

How to use claude update to get the release. New /config option workflow.size accepts small, medium, or large to guide agent-count in workflow runs. Advisory, not a hard cap. Workflow-spawned agents now emit workflow.run_id and workflow.name in OpenTelemetry traces for better activity reconstruction.

Fixes: Ctrl+R history search crash on accept or cancel; /rename reversion on job restart; transient mTLS handshake failures during certificate rotation; Remote Control commands failing with "Unknown command" in interactive sessions; images and files from the Remote Control app dropped without captions; unreliable clickable sign-in URLs over SSH; chat-opening crash in claude agents with a "currently running as a background agent" loop; unicode quote corruption in workflow scripts; unbounded voice dictation retry loop on audio failures; wrong permission mode shown in Remote Control sessions; session resume slowness in repos with many git worktrees; installer retry on transient connection drops; skill instruction duplication on re-invoke; MCP error clarity for servers with url but no type.

Also: /review <pr> reverts to fast single-pass review. Use /code-review <level> <pr#> for multi-agent. Workflows list gets wider titles, a dedicated time column, and shorter model names.

News

1 item

Jul 06

NEWS

Government of Alberta: 466 million lines in 20 hours

Anthropic published a case study from Alberta's Ministry of Technology and Innovation. The team used Claude Code with Opus and Sonnet models to scan 466 million lines of provincial government code in 20 hours, remediate security vulnerabilities, and build tooling for ongoing security review. Two-agent approach: red team probing for exploits, blue team writing remediation plans to exact files. An industry day in Edmonton this month. Province-wide scale-out begins fall 2026.

Why it matters Governments scanning their own codebases is not new. Twenty hours for 466 million lines is. This is the first public government-scale case study with Claude Code as the primary tool, and Alberta is sharing the playbook openly.

466 million lines. 20 hours.

The Release Log

Get Shipped. in your inbox.