Thinking from 18 years of shipping software.
Articles on AI, SaaS, product engineering, and building software that scales. Written by practitioners — people who build the things they write about.
Latest articles
Showing 6 of 14 results
Why 80% of enterprise RAG implementations fail — and what to do instead
The retrieval layer is designed last. The knowledge architecture is an afterthought. The evaluation framework doesn't exist. Here's what production RAG actually requires.
The four architecture decisions that kill SaaS products at scale
Multi-tenancy assumptions, authentication shortcuts, billing complexity underestimation, and observability as an afterthought. We've seen all four sink SaaS products that should have scaled.
TechBlueprint in practice — how structured discovery prevents six-figure rewrites
Three projects from the last 12 months where TechBlueprint discovery identified a fundamental architectural problem before a line of code was written. The cost of catching it early vs late.
AI agents in production — what works, what doesn't, and why
We've shipped AI agents for procurement, compliance, customer service, and research workflows. The patterns that make agents reliable in production are different from the ones that make demos impressive.
What 600 products taught us about the build vs buy decision
After 18 years and 600+ products, the patterns are clear. The build vs buy decision is almost never about cost — it's about competitive differentiation, integration complexity, and long-term ownership.
LLM fine-tuning is not always the answer — and here's when it is
Prompt engineering, RAG, and fine-tuning solve different problems. Most organisations reach for fine-tuning when RAG would be faster, cheaper, and more maintainable. Here's how to tell the difference.
Browse by topic
Hover a topic to learn more.
Get these in your inbox.
Join 210,000+ readers. New articles every week — no fluff, no filler.