Reading

Orosz digs into how companies like Google, Stripe, and Shopify actually measure productivity — not just DORA metrics but the qualitative signals they track. As we think more carefully about AI's impact on engineering velocity at Coursera, the frameworks here are useful. The key insight: the best teams measure outcomes, not activity. Lines of code and PRs merged are lagging indicators of work quality, not leading ones.

Simon Willison has the most consistently honest takes on what LLMs can and can't do in practice. This post on using LLMs for code is refreshingly pragmatic — no hype, just careful observation of where they genuinely accelerate and where they introduce subtle bugs you need to watch for. His point about LLMs being great at 'known unknowns' but risky for novel problems aligns exactly with my experience vibe coding side projects.

Will Larson continues to be one of the clearest thinkers on engineering leadership. This piece on strategy cuts through the abstraction — diagnosis, policies, actions. The hardest part he identifies (and gets right) is that strategy must explicitly acknowledge the tradeoffs you're not making. Most 'strategies' I've seen are just goals dressed up in strategy language. Useful framing for CTO-level planning work.

Mollick is doing the most rigorous empirical work I've seen on AI in education. His findings challenge several assumptions — including that AI tutoring straightforwardly improves outcomes. The nuance around when AI assistance helps vs. creates learned helplessness is directly relevant to decisions we're making at Coursera about how and where to deploy AI in the learning experience.

A classic that holds up. Allspaw's framing of 'mature engineers' vs just 'senior' resonates with how I think about building engineering culture. The emphasis on not just solving problems but understanding the human and organizational context around them — that's exactly what separates strong IC leaders from the rest. Still share this with every engineer I mentor.

One of the best practical guides I've found for LLM system design. Eugene breaks down patterns like evals, RAG, fine-tuning, caching, and guardrails with real-world tradeoffs. The section on context management maps directly to challenges we face at Coursera when building AI features at scale — particularly the tension between retrieval quality and latency.

Excellent breakdown of agentic AI patterns - from simple tool use to complex multi-agent systems. The progression from ReAct to more sophisticated patterns like planning and multi-agent collaboration mirrors what we're seeing in production systems. Particularly relevant as we build more autonomous engineering tools.

A sobering look at how urban development threatens centuries-old tea cultivation in Uji. The parallels to tech disrupting traditional industries are striking - sometimes 'progress' comes at the cost of irreplaceable cultural heritage.