Research

Research that wanders on purpose.

We're working out what Enterprise AI becomes next — in the open, in plain language, and honestly enough to say what we don't know yet.

Flagship program

Learning-Augmented Generation

Retrieval answers questions. Learning accumulates capability.

Today's enterprise AI mostly retrieves: you ask, it finds, it answers, it forgets. LAG is our wager on what comes after — systems where every interaction leaves something behind. A correction becomes a rule. An outcome becomes a preference. A mistake becomes a boundary.

The organization doesn't just use intelligence — it accumulates it. We're building LAG into the learning layer of our platform and writing up what we find as we go, including the parts that don't work.

Compound interest applies: every month the loop runs is learning your competitors don't have.

Abstract & first paper — publishing 2026

The manifesto

“The first generation digitized business. The second automated work. The next will continuously learn.

The long-form version — why we think the learning enterprise is inevitable, what it demands, and what we're doing about it — is being written now, with the same care we'd give a client's brand.

Publishing soon — subscribe below to read it first

What we're pulling on

Five threads, one direction.

i.

Organizational intelligence

What does it mean for a company — not a model — to be intelligent? Where does that intelligence live, and who tends it?

ii.

Enterprise learning

How feedback becomes improvement in production systems: what to keep, what to forget, and how to know the difference.

iii.

AI architecture

The shapes that hold up under real enterprise load — knowledge, memory, agents, and the seams between them.

iv.

Reference architectures

Patterns we’ve built more than once, written down so the next team — ours or yours — doesn’t start from zero.

v.

Engineering research

The unglamorous findings from the workshop floor: evals, guardrails, failure modes, and what actually broke.

Essays

Recent writing

A response · July 2026

After the descent, the only moat is memory

Chamath Palihapitiya says the price of intelligence is collapsing. Our response: what lands at the bottom of the curve is brilliant, amnesiac — and the moat moves to memory.

Read the essay

Research · July 2026

Learning-Augmented Generation

The program behind our wager: systems where every action becomes memory, every outcome becomes learning, and March is measurably better than January.

Read the essay

Argument · July 2026

You can't configure a moat

Low-code agent builders optimise for the median demo. Your edge lives in the twenty percent they abstract away — which is why the agentic layer has to be built and owned, not configured.

Read the essay

Position · July 2026

Build the thing that fits

The cost of building software has collapsed — so the old bargain of generic tools plus an army of managed services is over. Build the thing that fits, own it, and let it stay with you.

Read the essay

Thesis · July 2026

Own the weights

Open models have closed the agentic gap. The enterprise that fine-tunes its own weights turns institutional memory into an asset nobody can rent back to it.

Read the essay

Stress test · July 2026

If the bubble bursts

Maybe AI is a bubble. Maybe the labs go bust. The enterprise that built — rather than rented — keeps everything: the code keeps running, the weights don't expire, and the bill goes down.

Read the essay

Inside view · July 2026

The end of the billable hour

AI broke the maths of the billable hour. What replaces T&M — outcomes, subscriptions, virtual FTEs — why the giants are already pivoting, and the bet we placed ourselves.

Read the essay

Read it while the ink
is still wet.

One email when we publish. No newsletters, no nurture sequences — we're too busy building.