Insights

Notes from inside the work.

Written by the team that ships the systems: what works in production, what things actually cost, and the honest trade-offs behind AI engineering decisions. No syndicated fluff.

AI Agents

14 August 2026 : 7 min

AI agents that actually finish the job

The gap between an agent demo and an agent in production is not model quality. It is the engineering around the loop: what the agent may touch, how it verifies its own work, and when it must ask a human.

Read the note

Automation

22 July 2026 : 8 min

What AI automation actually costs

Automation budgets fail in both directions: initiatives killed over costs that would have paid back in a quarter, and subscriptions that quietly cost more than the salaries they replaced. Here is how the math actually works.

Read the note

RAG

30 June 2026 : 7 min

RAG: when your business needs its own memory

A language model has read the internet and none of your documents. RAG fixes that: it retrieves the relevant passages from your knowledge before answering, then cites them. The idea is simple; doing it well is engineering.

Read the note

Voice AI

5 June 2026 : 6 min

Voice agents: from novelty to front desk

Voice AI quietly crossed a line: latency dropped below conversation speed and the voices stopped being creepy. The businesses shipping it well treat it as staff training, not software installation.

Read the note

AI Product Development

18 May 2026 : 7 min

How to validate an AI product idea before you build it

The most expensive mistake in AI products is building the thing in full before anyone has proven they want it. Validation is a sequence, and it starts before code exists.

Read the note

GEO / AI Search

29 April 2026 : 6 min

How AI search engines choose who to quote

Answer engines do not rank pages the way classic search does. They assemble answers from sources they can parse, verify and attribute. Being quotable is a design decision, and most company websites fail it quietly.

Read the note