Observations from the trenches: product strategy, AI in production, customer signal, and systems that earn trust.
What is linkblogging?
A linkblog entry is commentary on something published elsewhere. Start with the original source, then come back here for the why.
Chesterton's Fence is a product leader's test for legacy systems: before you remove the ugly process, code path, or ritual, understand the failure it was built to stop.
Linkblog: commentary on ted.com
If you've worked with me and heard me talk about the first follower problem, this is probably the video I meant. Derek Sivers explains movement-building in under three minutes.
Most PM discourse draws a hard line: PMs shouldn't code. But when you're the only person on the team, building is user research, domain modelling, and competitive analysis compressed into the same motion.
My go-to Dublin potluck dish. A highly scalable, naturally vegan Cowboy Caviar featuring a warm caramelized dressing.
Why I stopped styling RSS feeds and started pre-compiling them at the edge. A D1-to-KV fan-out pattern that drops feed latency to single-digit milliseconds.
The enterprise AI rush is exposing a decade of organizational debt. AI guardrails are explicit context, strict boundaries, and unified taxonomies — the exact documentation we should have had all along.
Why Bentley Systems in Dublin keeps pulling my attention: a source-led read on smarter-city work, Cloud Connect, and sustainability that stays practical.
Why I integrated ActivityPub into ryanw.eu, adopted a POSSE publishing model, and treat the Fediverse as distribution infrastructure rather than rented audience.
LinkedIn is a database. A personal site is a product decision. What maintaining one taught me about editorial judgment.
Domain-Driven Design, ubiquitous language, and bounded contexts matter because product teams ship the wrong workflow when legal, product, and engineering mean different things by the same word.
A Zendesk case study in customer feedback prioritization: how a simple weighted scorecard separated signal from noise, surfaced hidden issues, and changed the roadmap.
A production-readiness release gate for AI features: representative evals, structured outputs, tool safety, observability, rollback, and the evidence I expect before GA.
High engagement in a product employees have to use doesn't mean good UX. It means no alternative. Here's how to pick HEART categories that matter.
RICE still works as a shared prioritisation language. On internal platforms, the failure starts when teams treat raw reach math as the answer instead of adding a domain-risk pass and documenting the rationale.
Knowledge shouldn't be a side project. Teams that treat docs as a first-class feature ship faster and break less.
Product teams drown in feedback: tickets, sales calls, NPS, stakeholder opinions. Without a scoring system, the loudest voice wins. Here's what replaces gut feeling.
