Notes on reasoning, governance, and agents
Field notes from the team building the reasoning layer for enterprise AI.
Why every agent decision needs a paper trail
Autonomy without an audit trail is a liability, not a feature. Here's how we designed Axiom's reasoning traces so every action an agent takes can be reviewed, replayed, and, if needed, reversed.
Routing reasoning depth: when to think longer
Not every task deserves the same compute. A look at how Axiom's router decides when to escalate to extended reasoning.
How Quartz Bank cut manual review 71%
Inside the rollout of an Axiom variance-review agent across four regional finance teams.
Introducing policy guardrails for tool calls
A new primitive for constraining what an agent is allowed to do, before it does it — not after.
Benchmarking multi-model failover in production
What we learned running 40M requests through automatic failover across five frontier models.
A field guide to agent access control
Role-based permissions weren't built with autonomous software in mind. Here's how we adapted them.
Nimbus Logistics: agents that replan mid-shipment
How a logistics operator uses Axiom to reroute shipments automatically when conditions change.