Insights

Field notes from the people doing the work.

No SEO chum, no AI-generated filler. Writing from our engineers, designers and PMs about how things actually break, deliver and scale.

Featured · AI & ML

RAG in production: the seven failure modes nobody warns you about

Most RAG demos work because the demo's question matches the document. Production traffic doesn't. Here's what breaks first, and how to instrument for it.

Aarav Mehta · Mar 12, 2026

9 min

Engineering

The boring case for event-driven backends in 2026

Kafka isn't fashionable anymore. That's exactly why it's a great default.

Priya Shah · Mar 04, 2026

Design

Why we deleted our design system component library, and delivered faster

We replaced 240 components with 38 primitives. Velocity went up. So did designer happiness.

Rohan Iyer · Feb 27, 2026

Product

Discovery sprints, post-mortem: what worked across 14 client engagements

Two patterns separated the successful sprints from the politely-archived ones.

Sneha Kapoor · Feb 19, 2026

Cloud

Our $42k/month AWS bill, dissected line-by-line

What we cut, what we kept, and the one item that was 31% of the spend for no good reason.

Anish Verma · Feb 11, 2026

AI & ML

Evals are the product. The model is the implementation detail.

An opinionated case for treating eval harnesses as first-class artefacts.

Priya Shah · Feb 02, 2026

Engineering

TypeScript at scale: the four tsconfig flags that paid for themselves

Strict isn't enough. Here's what we turned on after a year of regret-driven development.

Karan Patel · Jan 23, 2026