What I'm doing now.

Shipped Paperline (github.com/vscarpenter/paperline), a warm editorial design system: paper-toned surfaces, plum accents, Instrument Serif paired with Instrument Sans, and around 30 accessible React primitives with dark mode. Now live across GSD (gsd.vinny.dev) and Cascade (cascade.vinny.dev). Came out of a single Claude Design sitting and is the cleanest evidence I have so far that the LLM-as-pair-designer workflow can ship production-quality artifacts.

All in on OpenClaw: hardening the production setup, tightening security, and optimizing model routing and usage patterns.

Consolidated my earlier multi-instance experiments into one focused effort to make it genuinely reliable.

Shipped a NotebookLM-powered audio and video overview pipeline for the blog.

Each long-form post now carries a ~20-minute podcast and a short video explainer, rendered from frontmatter with proper RSS enclosures and JSON-LD. Using my own feed as the test bed.

Experimenting with home automation projects.

Finally jumped into the Home Assistant space with a new Home Assistant Green. Whole new world opened up for me, integrating AI assistants with smart home devices to build something actually useful.

"What to Make of a Life" by Jim Collins, a reflective read from the author of Good to Great on the questions worth organizing a life around.

"The Obstacle Is the Way" by Ryan Holiday, a Stoic playbook for turning friction into the path forward.

A useful counterweight to a quarter spent shipping fast and second-guessing harder.

What a small design system needs to feel like when LLMs are first-class consumers: token naming that survives a code-gen pass, component APIs that don't leak Tailwind specifics, docs that double as agent prompts.

Paperline is the test bed; the broader question is what this does to the apprenticeship path that made senior designers senior in the first place.

Agentic AI for the enterprise SDLC: separating genuine productivity from tool sprawl, and figuring out which patterns actually help teams ship.

Claude Code skills as a pattern for encoding senior-engineer workflows: where they outperform raw prompting, and where they just add friction.

How to measure and improve developer experience without adding more surveys to people's plates.

The balance between standardization and team autonomy in large engineering organizations.

What the Anthropic release cadence is doing to enterprise adoption: Design, Cowork, Chrome, Excel, Mac, Opus 4.7 all landing inside a single planning cycle.

Where does governance even attach when the surface area changes every two weeks?

Content-Signal, llms.txt, and what it actually means to declare a 'no training' posture when your site is the training data, both as a writer and as someone advising teams on AI strategy.

The future of learning in an AI-native world: when models can explain anything on demand, what does it actually mean to build expertise, and how should engineers (and kids) learn differently from here?