Agentic Coding
Top tech stories today across software, hardware, AI, and product launches. Senior engineer audience — skip rumour churn and pre-announcement leaks. Lead with shipping-now stories and what changed for builders.
← Zur aktuellen AusgabeDan Luu's agentic testing playbook, Opus 5.5 context jumps, skill curation grows
Dan Luu Agentic Testing
Someone finally documented what happens when you let agents loose on real test suites.
Dan Luu's latest deep dive covers months of running agentic test processes at scale, including the uncomfortable truth that agents routinely do things that would get a human fired—yet spinning up a thousand of them still produces net value [Source: Dan Luu]. The piece walks through benchmarking methodology, how LLM performance variance affects test reliability, and the practical tradeoffs of parallel agent execution. It's experiential rather than theoretical—exactly the kind of workflow evaluation the Vibe Code Pro benchmarks have been missing.
Worth the full read if you're scaling beyond single-agent sessions.
Opus 5.5 Model Card
Your context window just got official documentation.
Anthropic published the full Opus 5.5 model card covering context limits, output caps, pricing tiers, and platform-specific model IDs [Source: Claude Docs]. The overview consolidates what was previously scattered across release notes and API docs into one reference. You'll find availability details for each platform and direct links to implementation guides. If you've been configuring your Claude Code setup based on guesswork or outdated threads, this is the canonical source.
Bookmark it before your next AGENTS.md revision.
Awesome Claude Skills
The community skill catalog keeps growing quietly.
The awesome-claude-skills repository now organizes contributions across document handling, development tools, data analysis, and more [Source: GitHub]. It's a curated directory rather than a news feed, but the categorization helps you find specialized skills faster than searching the main catalog. Pairs well with the 2,400+ entry AAS Core catalog covered two days ago—use one for discovery, the other for production manifests.
Check what's been added since your last visit.
Agentic test processes, LLM benchmarks, and other notes ... - Dan Luu21 hours ago ... This is also generally the case with vibe coding and/or agentic loops because you have agents doing who-knows-what and you sometimes get a window into what's ...danluu.com
GitHub - BehiSecc/awesome-claude-skills23 hours ago ... ... code, generating optimized test suites with pairwise coverage. ... Claude Code plugins with schemas, templates, validation workflows, and troubleshooting.github.com