AI & Engineering

Perspectives on building AI products, data and domain advantages, engineering talent, and turning internal tools into production-grade platforms.

Diagram showing AI-generated code passing code review with lint and unit tests green, then failing in production three weeks later with boundary and contract failures

AI & Engineering July 7, 2026 12 min read

How to Test AI-Generated Code: A QA Guide for 2026

AI-generated code looks clean and passes review — then fails at boundaries and integrations. A QA playbook for how to test AI-generated code in 2026.

Read article →

Four-step pipeline of an AI failure explainer in a Playwright test framework: a test fails, a fixture gathers page or API context, a local Ollama LLM reasons over it with JSON Schema structured output, and a verdict with root cause, suggested fix, and confidence is attached to the HTML report

AI & Engineering July 6, 2026 12 min read

AI-Native Test Framework: Playwright + a Local LLM That Explains Failures

Build an AI-native Playwright test automation framework: prompt-engineered Gherkin test cases, plus a local Ollama LLM that explains every failure inside the HTML report — no API key needed.

Read article →

A 2x2 matrix of the four AI-usage strategies for writing tests — C1 (AI ideates and implements), C2 (AI idea, developer implements), C3 (developer idea, AI implements), and C4 (fully manual) — mapped against test idea and implementation ownership

AI & Engineering July 1, 2026 9 min read

Generative AI for Software Testing: What a Study Found

A study of developers writing unit tests with ChatGPT: the four AI strategies, one-shot vs iterative prompting, and why AI gets ideas right but assertions wrong.

Read article →