Company
Sense, reason, act: the loop behind ConsciousLab
Why we built one control surface for agents, machines, money and markets instead of five dashboards.
Mira Okafor5 min read
Journal
Notes from the people building ConsciousLab: how the systems work, what we measured, and what we got wrong.
Latest,
Sense, reason, act: the loop behind ConsciousLab
Why we built one control surface for agents, machines, money and markets instead of five dashboards.
Mira Okafor, Co-founder, CEO. 5 min read.
7 articles shown
- Product
Approval gates: deciding what an agent may do unattended
A policy model for agent actions that scales from one refund to a hundred thousand runs a day, and what we learned rolling it out.
Hannah Lindqvist4 min read - Engineering
Bridging ROS 2 fleets without touching a single node
How the ConsciousLab edge agent subscribes to DDS topics, keeps traffic on the customer's network, and survives 72 hours offline.
Felix Brandt4 min read - Engineering
Why agents never touch a balance
The double-entry ledger under Payments, and the three invariants that let an agent issue refunds without finance losing sleep.
Daniel Achterberg4 min read - Research
Replaying 40,000 production traces to evaluate a model change
Offline benchmarks told us the new model was better. Replaying real agent traces told us where it was worse, and by how much.
Priya Raman5 min read - Engineering
Pre-trade risk checks in under a millisecond
Every order on ConsciousLab passes six risk checks before it routes. Here is how they stay fast enough that traders never notice.
Jonas Ehrlich4 min read - Product
One event schema for five products
How a single envelope with an entity, an actor and a timestamp lets Analytics answer questions that cross robotics, payments and agents.
Sofia Marchetti4 min read