AI-Native Developer
Copilot, then Codex, then Claude — so why does my workflow still look like 2022?
How a working developer should actually think about AI: cost structures, verification design, and the skills that gain value as code generation gets cheap.
Copilot, Then Codex, Then Claude — So Why Does My Workflow Still Look Like 2022?
Three tool migrations, zero workflow changes. What METR, DORA 2025, and my E2EE work taught me about using AI vs. redesigning around it.
Prompt Engineering Got You Here. So Why Doesn't It Scale?
Coding agents got strong; the bottleneck moved from the prompt to the workflow around it. Eight practices big tech converged on, mapped onto one 7-stage loop.
The Agent Says All Tests Pass. So Why Does QC Still Reject the Build?
A green build and a rejected QA build answer different questions. One verification contract, invariants to impact scope, serves agent, reviewer, and QC alike.
Which Mobile Testing Tool Do I Need? Ask the Boundary, Not the Platform
JUnit, Espresso, XCUITest, Maestro, Appium: a map of mobile test tools by the boundary they verify and the realism they need, with what Google and Apple say.
How Do I Test an E2EE Chat Between Two Devices Without a UI?
A headless two-device harness for E2EE chat: real keys, real store, real backend, no screen: reproducing exhausted one-time keys and a lost session on demand.