Something broke. Something cost you $50,000. Something got breached.
Ask any tool why - and it shows you what moved.
Not modules. The machinery every wing runs on.
The Causal Oracle
discovery“It learns cause from your telemetry, then answers what if?”
Structure discovery (CUTS+ / NOTEARS), do-calculus, counterfactual simulation, a digital twin of your infrastructure.
The Symbolic Guardian
safety“Every action is proven safe before it runs, and signed.”
Cryptographically attested, fail-closed safety gating on every intervention the system proposes.
The Active Inference Agent
decision“It chooses the move that resolves the most uncertainty, safely.”
Expected Free Energy decisions that balance acting on the graph against learning about it.
Show the engine once - every wing inherits its credibility.
Eight rooms hang off the one graph.
A curator never gives every work the same wall. One room is installed and live. Seven are commissioned studies. The ambition is the full platform; the depth is earned where it can be proven.
Will it break before users notice?
On the spine - Sheaf early-warning - drift reads as a rising edge on the graph
Why did the bill spike - in dollars?
Performance - the reveal
deploy checkout-v2.3
one commit · merged 14:02
memory leak
+0.4 GB/hr
OOM kills
first casualties
pod thrash
47 restarts
EBS churn
+$1,140/day
$0
traced through the chain
87%
of the spike · one cause
1 commit
checkout-v2.3
Instrument - operate it yourself
Every line correlates with cost. None of them explains it.
$26,310
of $43,850 avoided
counterfactual on the shared graph
We start most partners on Cost - it pays for itself in the first spike. The rest of the graph comes with it. See the offer →
Will this deploy break prod?
On the spine - Counterfactual - simulate the ship before you ship it
How did the breach actually happen?
On the spine - Attack chain reconstructed via the Inflation Technique
Can it fix itself - provably safely?
On the spine - Guardian-gated action loop
Prove why this was allowed.
On the spine - Causal provenance plus a signed attestation
What one experiment teaches us most?
On the spine - Expected Free Energy picks the highest-information failure to inject
What should we do next - across all of it?
On the spine - The whole graph, zoomed out to risk and portfolio
sig 9f2c…a41e · fail-closed · attested
Autonomy you can't trust is a demo. Autonomy that's provably safe is a product.
- Fail-closedif safety can't be proven, nothing runs
- Cryptographically attestedevery gate leaves a signed receipt
- Proven before it runsthe proof precedes the action, always
You met the Guardian in the Engine Room. You will meet it again in every room that acts - deployment gates, remediation, compliance. It is the museum's structural steel: autonomy you can sell is autonomy you can prove.
One causal graph. Eight ways to operate smarter.
Start where it hurts most.
Book 30 minutes →Wing I · CausalOps
Why did it break?
enter through this pain ↗
Wing II · CausalCost
Why did it cost?
enter through this pain ↗
Wing III · CausalDev
Will this deploy break prod?
enter through this pain ↗
Wing IV · CausalGuard
How were we breached?
enter through this pain ↗
Wing V · CausalRemedy
Can it fix itself?
enter through this pain ↗
Wing VI · CausalCompliance
Can you prove why?
enter through this pain ↗
Wing VII · CausalTest
What should we test?
enter through this pain ↗
Wing VIII · CausalInsight
What should we do next?
enter through this pain ↗
We start most partners on Cost - it pays for itself in the first spike. The rest of the graph comes with it.