Month: September 2025

Evolving what development means

The shift to agentic code isn’t about replacing developers. It’s about evolving what development means. There are valid concerns AND solvable problems. Leaving engineering efficiency on the table is not an option.

“I Don’t Trust AI-Generated Code”

  • When was the last time you yolo’d a review for by a co-worker?
  • Good possibility you have transitive dependencies that you a) don’t know about and/or b) haven’t audited.
  • Trust isn’t binary, it’s earned through verification. In the kernel performance work I used to do, we’d compile 40 different kernel versions to isolate an issue. Each build was an experiment, each test a verification. We didn’t trust without validation.

Agentic code works the same way. You don’t trust the code because an AI wrote it. You trust it because:

  • It passes comprehensive tests you defined
  • It meets specifications you controlled
  • It survived review by humans who understand the business context
  • It performs correctly in production (the ultimate judge)
Continue reading “Evolving what development means” →

The Path to Vibe Coding for the Enterprise

Hey folks,

I’ve been spending a lot of time thinking about how we write software. Actually, this was spurred on by a recent project: I thought, how can I drive inference workloads to llm-d? How can I keep vLLM “hot”? codegen is an inference-heavy workload, so let me look deeper.

What I found was a series of opportunities that we’ll need to tackle in order to make “vibe coding for the enterprise” possible… Opportunities that I feel uniquely positioned and experienced to effect 😉

First, we have to face the problem. The need for software…truly secure software is obvious and growing in urgency. The solution to this is to stop writing software entirely.

Like every other career path that transforms and evolves, often suddenly and without warning, your entire career is about to transform and evolve.

Continue reading “The Path to Vibe Coding for the Enterprise” →