Claude Code shipped two releases in two days (Aug 3-4): v2.1.221 adds a VSCode Focus view and real permission-check bypass fixes, v2.1.222 closes a gap where worktree-isolated subagents could still run destructive git commands against the main checkout — and quietly removes the Ultraplan feature entirely.
A 79.2% SWE-bench Pro figure for Claude Opus 5 has been circulating since launch — but Anthropic’s own announcement never cites one, and Scale AI’s actual public leaderboard doesn’t list Opus 5, Fable 5, Mythos 5, or even Opus 4.8. The number traces back to a secondary aggregator whose own methodology page warns against treating it as verified.
Alibaba’s Qwen3.8-Max went from a bare parameter count to a full benchmark table on August 3 — $2/$6 per million tokens, 67.7% SWE-bench Pro, and a two-point Terminal-Bench 2.1 win over Claude Fable 5. Every number is still self-reported: no independent lab has scored it yet, and open weights remain a promise for the week of August 10.
GitHub added xAI’s Grok 4.5 to Copilot on July 28 — a 500K-context model bolted onto an already multi-model picker. A day later, Satya Nadella told investors Microsoft is building one unified app to merge Copilot Chat, GitHub Copilot, Copilot Cowork, and Autopilot, sometime in 2026. That roadmap gap is the whole story.
The Model Context Protocol’s 2026-07-28 specification finalized on schedule, locking in a stateless protocol core, Multi Round-Trip Requests, header-based gateway routing, and hardened OAuth — with Tier 1 SDKs already at general availability and close to half a billion downloads a month across them.
Moonshot AI shipped Kimi K3’s full weights on July 27 — 2.8T total parameters, 104B active, 1M context, the largest open-weight release to date. But the license isn’t the ‘Modified MIT’ terms that circulated before launch: it’s a bespoke document with a $20M MaaS revenue gate and a mandatory UI-branding clause above 100M MAU, and several outlets reported the permissive version anyway.
Security firm Accomplish disclosed SharedRoot on July 23: a six-step exploit chain that escapes Claude Cowork’s local macOS sandbox and reads/writes anywhere on the host filesystem, affecting an estimated 500,000 local-execution users. Anthropic closed the report as ‘Informative’ without shipping a fix — the exposure is only sidestepped because Cowork now defaults to cloud execution, leaving anyone who opts into local mode still exposed.
Anthropic shipped Claude Opus 5 on July 24, positioning it as near-Fable-5 coding intelligence at half the price ($5/$25 per million tokens) and making it the new default across Claude Max, Pro, and Claude Code v2.1.219. Cursor and JetBrains both issued launch-day endorsement quotes.
A new Theori study built 28 apps with five frontier AI models and confirmed 434 exploitable vulnerabilities after deduplication and PoC testing. SQL injection and XSS were nearly absent — but resource-exhaustion bugs, hardcoded secrets, and authorization gaps that double in larger codebases (11% to 28%) are the new failure modes agentic coding tools haven’t solved.
Claude Code v2.1.215 stops auto-running /verify and /code-review, putting review invocation back under explicit developer control. v2.1.216 follows a day later with a real isolation fix — worktree-isolated subagents could redirect git into the shared checkout via git -C or GIT_DIR — plus a quadratic-slowdown fix and a new sandbox.filesystem.disabled setting.