Blog

How we are growing the team, and how we are not

The state of sovereign agent adoption, 2026

Oran in the wild, and what surprised us

What changed in Oran 3.6

Hiring engineers who have never run an agent

What we look for in Scholars

Agents that work outside English

What public sector teams ask us before anything else

Community research across regions

Introducing the Yaju Software Factory: from backlog to pull request

How we work with delivery partners

Building for AI regulation before it fully lands

What sovereign deployment actually requires

Running agents under EU data rules without pretending it is simple

Small agents, scoped tightly, run for years

Why our engineers sit with customers, and what they are not there to do

Notes from a distributed team that does not have an office

An agent can only be as good as what your corpus left out

A small agent for a narrow job, and why that is the point

The Yaju approach to AI governance

Meet the Software Factory

A checklist for evaluating an agent platform

Small models do most of the work in a well-designed system

What regulated industries ask that everyone else eventually will

A day in the life of a wealth manager, with and without agents

Rolling agents out across business units without central control

The optimisations worth doing, and the ones that only look impressive

How our early access programme works

Where agent time actually goes, and why it is rarely inference

Running agents on hardware you own

Automating fork maintenance, a job nobody wants

What an embedding actually commits you to

What breaks when a transcription system meets a language it was not tuned for

The total cost of AI ownership is not the number on the invoice

A maturity model for agent operations, with the levels that actually matter

Agents and the systems of record

Where agents fit in telecom operations

What agent workloads do to infrastructure planning

Why agent programmes stall at the pilot

Agents on the factory side of the business

The future of work debate has an evidence problem

The gap between a working demo and a working business process

Building a retrieval evaluation set from a public corpus

Running reranking inside your own cloud

Most of your knowledge is not text, and your index probably ignores it

When a bigger model is the wrong answer

Agents that plan and agents that react

Reranking is the cheapest accuracy you are probably not buying

Latency budgets for agents, and why nobody sets one

Choosing a model is an operations decision, not a benchmark decision

What air-gapped agent deployment actually involves

Chunking is the retrieval decision everyone makes by accident

Documents are not text, and parsing is where most retrieval quietly fails

Exploring agents in education, carefully

Mapping the agentic task ecosystem

Transcription is an input problem, not a transcription problem

The early footprint of automations, measured rather than predicted

What we talked about at Connect