Most software teams are building wrappers around LLMs to automate their processes. That creates suboptimal solutions more often than not. Grab’s data engineering case study is a great example of crafting a well-engineered system of agents to optimise an internal bottleneck that uses LLM as a component rather than its brain.
Slow vibe coding is the only style of AI-assisted coding I can do feeling content and proud about my code quality. If anything, AI has not made me 10x faster but 5x better and maybe 2x faster.
Search and discovery will be the first thing that’s disrupted in the agentic commerce era. 7/10 store owners are actively investing in LLM discovery and traffic. Both stats match my instincts about how e-commerce is changing, which makes building what I’m building even more exciting.
Realism is a mark of seniority, optimism is a trait of visionaries, pessimism is what puts your work and career in danger. The difference is subtle and lies in how you conclude situations. The realist says, “The timeline is tight and the dependencies are unclear.” The pessimist says, “The timeline is tight and the dependencies are unclear, so this isn’t going to work.” The optimist says, “The timeline is tight and the dependencies are unclear, so here’s what we’d need to do to ship it.”
The FDE role evolving into solutions architect isn’t surprising given how sales is a profit center in any org. All top AI companies rapidly building their services businesses was inevitable.
The age of GitHub as we know and love is sadly getting over. I’m not moving away yet, so hoping for a revival. It must happen sooner than later, or it’s game over.
Is AI the long fantasized silver bullet for peak productivity? Was IDE one? Interesting and historically-rich account of silver bullets (or rather the lack of it) in software engineering.
Professionalism at work is critical, but my take is nobody is born professional. You gotto experience the 4 behaviors to know how to do better next time. Experience is the biggest teacher.
Reading about OpenClaw’s architecture is like getting familiar with a system design problem. Also, the section about AI evals is unmissable.
Focusing on supporting specialized hardware to run AI smoothly is a wise choice than clobbering the OS experience with dysfunctional AI parts, haha.