AxisOps Harbour
Labs

Topics

ai

micro-zero: own the harness

Agentic sessions are where our features get built now, which makes the harness they run in a piece of infrastructure. We are building our own, called micro-zero, and connecting it to the organisation's memory. The zero is where every session starts.

The context problem is a retrieval problem

An assistant session starts knowing nothing about the organisation, and no context window is ever going to hold fourteen years of decisions. We pointed retrieval at our own vault instead. The idea is a weekend's work to try; whether it holds up in daily use is what we intend to find out.

The model fits; the memory bus decides

Quantised models running on a workstation, a Pi or a phone. The interesting constraint is memory bandwidth rather than FLOPS, which changes which hardware is worth your money, and a capable model on a machine you own means no prompt ever leaves the building.

The expensive part stopped being the munition

A first-person-view quadcopter with a warhead taped to it costs a few hundred pounds and destroys a vehicle worth several million. The airframe is a toy, the radio is a toy, and the only component that was ever hard is now a model that fits on a board you can buy. That is a cost inversion, and cost inversions do not reverse.

The protein problem became a data problem

A neural network won CASP13 outright, and a sequencer that runs off a laptop's USB port costs less than a server. Biology is becoming a field where the expensive instrument is a model and the cheap instrument fits in a coat pocket, which changes who is able to do the work and who ends up holding the data.