AI and developer tool updates (2026-07-26) #89

July 26, 2026 — 3 min read

Today's Letter

GitHub said Claude Opus 5, Anthropic's newest Opus model, is now available in GitHub Copilot as of July 24, 2026.
The model is positioned for complex long-running coding tasks, including multi-step reasoning, tool use, autonomous code changes, and regression verification.
GitHub said early testing showed targeted edits, stronger self-validation, and lower unnecessary execution overhead on agentic coding workflows.
Access is planned for Copilot Pro+, Max, Business, and Enterprise users across VS Code, Visual Studio, Copilot CLI, the Copilot cloud agent, github.com, mobile apps, JetBrains, Xcode, and Eclipse.
Rollout is gradual, Enterprise and Business admins must enable the Claude Opus 5 policy in Copilot settings, and usage is billed at provider API list price under usage-based billing.
GitHub also said the model includes enhanced safeguards for high-harm cyber content, which may block some security-adjacent requests.

Source: github.blog

Anthropic announced Claude Opus 5 on July 24, 2026 and made it available immediately.
The company said Opus 5 approaches Claude Fable 5 on frontier intelligence at about half the price, and it becomes the default model on Claude Max and the top-tier model on Claude Pro.
Anthropic said Opus 5 matches Opus 4.8 pricing while improving performance on coding and knowledge-work evaluations including Frontier-Bench v0.1 and CursorBench 3.2.
On CursorBench 3.2 at max effort, Anthropic said Opus 5 comes within 0.5% of Fable 5's peak score at roughly half the cost per task.
The company also reported stronger results on ARC-AGI 3, Zapier AutomationBench, and OSWorld 2.0, with Opus 5 leading its internal cost-performance comparisons across those benchmarks.
Anthropic said the model improves scientific research tasks over Opus 4.8, including internal gains of 10.2 percentage points on an organic chemistry benchmark and 7.7 points on a protein-function task.
The release also emphasizes agentic reliability, with Anthropic highlighting examples in debugging, computer vision-assisted CAD reconstruction, and long multi-step workflow execution.
Anthropic noted that Opus 5 still trails Mythos 5 on cybersecurity tasks despite broader state-of-the-art claims in coding and knowledge work.

Source: anthropic.com
More: decrypt.co

Vercel announced Vercel WAF for Blob in beta on July 24, 2026, extending WAF protection to Vercel Blob stores.
Existing WAF actions including deny, challenge, rate limit, redirect, and log now apply to blob traffic without changes to code, blob URLs, or `@vercel/blob`.
Protection is enabled from Blob store settings because every blob is already served through Vercel's CDN, with rules evaluated at the edge by IP, country, path, and related signals.
Deny returns HTTP 403 before data transfer begins, while rate limiting returns HTTP 429 when a client exceeds the configured threshold.
Vercel said the OWASP Core Ruleset is not supported because it targets dynamic application traffic rather than object delivery.
Setup is currently dashboard-only and connects protected stores to a shared `vercel-blob-default-project` under the team.
One rule set covers every protected store, so rules cannot be scoped per store during the current beta.
Browser challenge rules require a browser to solve, which means server-side `@vercel/blob` requests matching a challenge rule are blocked.

Source: vercel.com

Jocoletter curates AI, software, and product trends for developers and builders.

#Anthropic #GitHub #Vercel