What we are learning as we build with AI.

Notes from real work, written for the people who run operations.

01 · PERSPECTIVES

Longer reads on how AI fits into real work.

Each one starts from a problem we have worked on and ends with what we would do again.

LatestAI Systems · · 2 min read

The model isn't the product anymore.

What differentiates useful AI systems is everything built around them: context, data, tools, memory, permissions, evaluation, interfaces and operating workflows.

What we learnedThe competitive advantage is shifting from having AI to engineering systems around intelligence.

Read the perspective
02 · FIELD NOTES

Short lessons from building.

Two or three minutes each. One problem, what we saw, and what we do now.

  1. AI engineering · · 2 min read

    Agents need state, not just conversation history.

    Keep a small, structured record of where the work stands. Let the conversation be a log, not the memory.

  2. Retrieval · · 2 min read

    Top-K is not a retrieval strategy.

    Choose fewer, better passages. Filtering and ranking matter more than how many results you pass to the model.

  3. Architecture · · 2 min read

    When agent frameworks help, and when they don’t.

    Start with plain code. Add a framework when you need branching, pausing for approval or rolling back, not before.

  4. Infrastructure · · 2 min read

    Long AI tasks don’t belong in a web request.

    Put long-running work on a queue, show progress to the person waiting, and make every step safe to retry.

  5. Evaluation · · 2 min read

    “The answer looks good” is not an evaluation.

    Write down what a good answer must do, collect real examples, and re-run them every time the model, prompt or data changes.

  6. Security · · 2 min read

    Agents should not inherit every user permission.

    Give each tool only the access its task needs, and ask a person before anything sensitive is changed.

03 · WHAT DIDN’T WORK

Not everything we try works. Here is what we changed.

Failed assumptions are where a lot of the useful learning is.

04 · ON OUR RADAR

What we’re watching, and whether it’s ready.

A short, honest read on newer technology. We update it as our view changes.

  • Ready to useWe would use it in real work today.
  • Building withWe use it in prototypes and pilots.
  • TestingWe are running small experiments.
  • WatchingPromising, but not ready for most teams.
  • Small, specialised models

    Not every task needs the largest model. Smaller ones can run closer to your data.

    Ready to use

    Our takeWorth using now for high-volume, narrow tasks such as sorting, extraction and routing, especially where data must stay in-house.

  • Shared standards for agents and tools

    Open standards such as the Model Context Protocol let one tool work with many AI systems.

    Building with

    Our takeUseful today for exposing your own systems to AI safely. Open agent-to-agent networks are still early.

  • AI that operates existing software

    Models that can use screens, keyboards and forms could automate older software with no API.

    Building with

    Our takePromising for back-office work. Keep it in a sandbox and have a person approve anything that commits a transaction.

  • Long-term memory for agents

    Agents that work over weeks need structured memory that can be reviewed and corrected.

    Testing

    Our takeImportant for multi-day workflows, but the patterns are still settling. Keep memory visible and editable.

  • Interfaces built around the task

    Software that assembles the right screen for the job, instead of fixed menus and dashboards.

    Testing

    Our takeWorth trying in internal tools. Keep customer-facing screens predictable for now.

  • AI in machines and sensors

    Vision and sensor models are moving AI from screens into factories, sites and vehicles.

    Watching

    Our takeCamera-based inspection and safety checks are usable today. Robotics is still one to watch.

The operational AI note · monthly

One useful email a month.

What we built, what worked, what failed, and one idea you can try in your business.