Aitina Tech R&D

The Software Lens.

Aitina Tech's research and development wing. One question per piece: what actually changes when software begins to reason at call time, which of our existing patterns survive that change, which quietly stop working — and what all of that asks of the people running the organisation. Watch the episode, then read the writing behind it.

7 video episodes10 articles and posts Free, no signup
Currently in the lab

Two pieces are in active research right now: Hire the AI Agent, Part 2 and Talk to Data. Published when they are right rather than when they are due — no dates, and no signup to read them.

What we are researching →
Currently researching

What the wing is working on now.

Published when it is right rather than when it is due, so there are no dates here. These are the two pieces in active work — and when they land, they land here first, free.

In progress

Hire the AI Agent, Part 2

The first part argued that software stops being licensed and starts being hired. Part 2 is the harder half: what hiring actually looks like in a contract, how an agent is paid, measured and released, and what a buyer signs when the thing they are buying is a capability rather than a seat.

Follows Hire the AI Agent — SaaC and the end of SaaS
In progress

Talk to Data

How to let people ask questions of their own data reliably — without handing a model the keys to the warehouse. Permission-aware retrieval, the queries an agent may and may not compose, what it is allowed to see versus what it is allowed to say, and the harness that makes the whole thing auditable.

Builds on our harness engineering work

Want to know when one lands? Tell us which and we will send it — no list, no newsletter, just the piece.

Part 1

Episodes — watch first

The video series. Most episodes have a written companion in Part 2, so you can watch the argument and then read the detail.

Subscribe on YouTubeNew episodes are announced on the channel first.

Part 2

Writing — articles and posts

Everything we have written, grouped by subject. Long-form research and shorter pieces from delivery work sit together, because the distinction only ever mattered to us.

Foundations — Start here

What actually changes when software begins to reason at call time. The seventy-year arc that led here, and the vocabulary the rest of the work uses.

AIM — The architecture

The Agentic Interface Model - where the seams go once microservices stop fitting, and what that looks like running on AWS Bedrock.

Nothing in the library matches that.

What it is

The research and development wing of Aitina Tech.

Most consultancies do their thinking inside client engagements and never write it down. We took the opposite view: the questions worth answering are bigger than any one project, so they deserve dedicated time, a public argument and somebody willing to be wrong in print.

The Software Lens is where that work happens. It is funded by the consulting practice and it feeds straight back into it — the frameworks published here are the ones our architects build to on Monday morning.

It also says something about how we hire and how we work. Engineers here are expected to read, to disagree, and to bring an argument rather than a preference. Original thinking is part of the job description, not an after-hours hobby.

  • Independent thinking. We publish conclusions that cost us work when they are the right ones — including “do not build this”.
  • Research before opinion. Every framework here came out of a real system that had to work, not a whiteboard.
  • A creative environment. Time is set aside for R&D. It is protected, not squeezed into the gaps between billable work.
  • Published in the open. No signup, no newsletter, no gated PDF. If it is useful, take it — and use it with another supplier if you like.
  • Willing to be corrected. Reasoning is shown so you can disagree with it. That is the whole point of writing it down.
Why we publish it

The library is the by-product of the work.

Everything here started as a problem on a real engagement — an architecture that would not scale under agents, a knowledge layer that returned plausible nonsense, a delivery model that stopped making sense once the caller was not a person. We write it up because writing it up is how we make sure we actually understood it.

None of it is gated. There is no signup, no newsletter and no PDF to trade an email address for. New research is published regularly, in reading order.

We used to post short industry notes here as well. Those are kept in the industry notes archive; we no longer run a rolling headline feed, because repeating someone else’s announcement is not research.

  • No gated PDFs, no signup wall, no newsletter
  • Every episode has a written companion; every article names its sources
  • Updated as the research continues
  • Reusable in your own decks — just credit it