Niyati · Structural Intelligence Engine · Causalor Labs
Real 6-asset portfolio. 12-month HRP mandate.
Goal path closed at month 11.
The exact step your goalbecomes structurally unreachable.
Not a simulation. Not a probability. A structural proof computed from the exact geometry of your system, before any metric moves or alarm fires.
point_of_no_return · timeline
t = 0 → 10
Action window open
Valid sequences exist. Intervention changes outcome.
t = 11 · PNR
goal_omega = 0
No remaining sequence reaches goal. Structurally sealed.
Iberian blackout
Three recovery paths existed. 12-minute window.
Archegos margin call
Exit window closed in mid-2020. Invisible.
ERCOT EMS alarm
One decision sequence remained out of seven.
A New Category of Intelligence
A new layer of intelligence.
Not a better version of an existing one.
Optimization tells you what to do next. Simulation samples from distributions. Neither tells you whether your goal is still structurally reachable from the exact current state, or names the precise moment that changes.
The Founding Insight
Failure appears as the moment where you are still making locally valid decisions, but certain outcomes have already become structurally impossible. Niyati computes that boundary before any metric moves.
What existing tools give you
What Niyati gives you
Three real events · Three structural signals · All computable before the failure
Real system states, reconstructed and run through Niyati. The signal was in the structure.
Iberian Grid Blackout · April 28, 2025
The grid had a 28% power surplus. Alarms were silent. Three viable operator paths existed. All required acting in the first 12 minutes.
ERCOT Texas Blackout · February 2021
Every metric was within normal range. The joint state had entered a structural corridor. One viable decision sequence remained. Root cause: gas pressure, not weather.
Archegos Capital Collapse · March 2021
Goldman Sachs exited profitably. Credit Suisse lost $4.7B. The structural analysis: zero valid exit sequences. Not improbable. Structurally impossible.
Core Outputs
Four Answers Only Structure Can Give
Optionality
How many futures your system still has.
Every action cuts the volume of reachable states. Niyati counts what remains. When the count reaches zero, recovery is not difficult. It is impossible.
Point of No Return
The step when recovery closed.
Not when the alarm fired. The precise step when the structural geometry closed. Niyati computes it before you get there.
Collapse Velocity
How fast your futures are disappearing.
Dashboards measure what happened. Niyati measures how fast recoverable futures are narrowing, before any surface metric moves.
Last Reversible Decision
The moment before a choice becomes permanent.
Every decision has a lock-in step. After it, structurally irreversible. Most find this in the post-mortem. Niyati finds it before.
Advanced Signals
The Layer Beneath Probability
Simulation tells you what is likely. Structure tells you what is possible. These are not the same instrument.
Breaking Point
Minimum force to make recovery impossible.
The exact perturbation that closes your goal permanently. Know this number before an adversary finds it, or a cascade does.
Hidden Fragility
Capacity in one direction. Nothing in another.
A system can appear robust along every visible axis and be structurally fragile along an invisible one. Niyati computes the blind side.
Path Concentration
48,000 paths through a single bottleneck.
Most trajectories converge through the same structural chokepoint. Forty-eight thousand options or one: those are not the same situation.
Structural Honesty
What we computed. Where we approximated.
Trillion-state systems exceed exact computation. When Niyati approximates, it declares precisely where and by how much. Never implied.
Portfolio OS · Live Production Integration
We built on it ourselves.
Portfolio OS is a live financial management system running the full 6-endpoint structural analysis on a real 6-asset portfolio (Metal, Indian equity, US equity, ETF, mutual funds, and fixed income) against a 12-month HRP rebalancing mandate.
The portfolio is structurally alive. The goal path is not.
Why this is different from a case study
Historical forensics
We analyzed what went wrong for others. Ground truth: known disasters.
Portfolio OS
We run the engine on our own portfolio. The intervention clock is live.
6 Endpoints · Live API Outputs
/v1/simulate
goal_impossible. System survives, goal path closes at month 12. Intervention deadline: month 11.
/v1/solve/adversarial-allocation
Metal collapses first. Rational adversary puts 91.7% of budget on Metal (κ_base = τ* = 0.55).
/v1/solve/saddle-point
8 rounds to collapse under sustained pressure. Both players converge on Metal every round.
/v1/solve/adversarial
Survival verdict: CRITICAL. Crash of 60bps destroys 75% of goal space (4 → 1 reachable state).
/v1/solve/fragility
Collapse risk: 7.4%. Anisotropy: 13.3x. Drawdown axis is locked; Sharpe axis has room.
/v1/solve/competition
SSI = 67×. A bad month under adversarial conditions is 67× worse than a calm month.
Structural Status Summary · Current Portfolio
| Signal | Value | Status |
|---|---|---|
| Runway verdict | GOAL IMPOSSIBLE | Act before month 11 |
| ε* (epsilon star) | 0.017 | Marginal |
| Danger fraction | 23% | Elevated |
| Survival verdict | CRITICAL | On the boundary |
| Stressed goal states | 1 of 4 | Thin |
| Collapse risk | 7.4% | Gate open ✓ |
| Anisotropy | 13.3× | Watch drawdown axis |
| Rounds to collapse | 8 | Below safe horizon |
| Adversarial multiplier (SSI) | 67× | High |
| Named weak link | Metal | First to collapse under adversarial budget |
| Root constraint | max_drawdown | Primary lever to relax |
Bottom Line
Structurally alive but thin. Fragility gate open; rebalancing proceeds normally. Metal is the named weak link. 8 months is the structural clock under sustained pressure. Survival margin under a single crash event is essentially zero. Intervention deadline: month 11. The recommended path is still open.
Game-Theoretic Lenses
What does a rational adversary do to your system?
Four endpoints that reframe your system as a two-player game. One player optimises for survival. One player optimises for collapse. The engine computes the equilibrium.
Adversarial Budget
/v1/solve/adversarial-allocation
Question
If a rational adversary had a fixed budget to break your system, where would they spend it?
Output
Named weak link + optimal attack allocation across all variables.
Portfolio OS Result
Metal collapses first. Rational adversary puts 91.7% of budget on Metal (kappa_base = 0.55).
Saddle-Point Equilibrium
/v1/solve/saddle-point
Question
Under sustained, rational adversarial pressure, how many rounds before the system collapses?
Output
Rounds-to-collapse at equilibrium. Both players converge on the same lever.
Portfolio OS Result
8 rounds to collapse under sustained pressure. Both players converge on Metal every round.
Adversarial Crash
/v1/solve/adversarial
Question
What does a single worst-case shock event do to the reachable goal space?
Output
Survival verdict and fraction of goal states destroyed by a single crash.
Portfolio OS Result
CRITICAL. A 60bps crash destroys 75% of goal space (4 to 1 reachable state).
Nash Equilibrium
/v1/solve/nash
Question
When multiple agents compete for the same constrained resource, what does each agent do at equilibrium, and what does coordination failure cost?
Output
Per-agent optionality at Nash equilibrium + cooperation_deficit: the structural cost of not coordinating.
Portfolio OS Result
At Nash equilibrium, Metal and fixed income desks each independently reduce the other's goal space. Cooperation deficit: 31% optionality loss.
Historical Forensics · Real Events · Known Outcomes
We ran it backward on real events.
Here is what the engine found.
Each case: a known disaster, a conventional explanation, and what structural analysis computed before it happened. Ground truth is known.
All outputs produced by the live engine against historical data · Ground truth known
Structural Verdicts · Engine Output
Four states the engine returns.
No probability range. No confidence interval. A precise determination from the exact geometry of your system: what it means and what to do next.
At least one structurally valid path to the goal exists from the current state.
Action: Act before the intervention deadline. point_of_no_return is computable and approaching.
Recovery was structurally possible. The point of no return has passed. No sequence of remaining actions restores reachability.
Action: Pivot strategy. The current path cannot recover. Surface metrics will continue to look normal.
The dangerous gap: system appears healthy, mandate is gone
The initial state was structurally incompatible with the goal. The system operates; the mandate was always foreclosed.
Action: The goal cannot be achieved on any path from this start. Reframe or restart from a structurally viable state.
Omega = 0. No action sequence leads to any reachable state. This is total structural collapse. The visible end, not the cause.
Action: Post-mortem. The structural seal happened earlier, at the point_of_no_return, not here.
Vertical Applications
One engine.
Every domain where futures close quietly.
The structural question is the same across every domain. The constraints, variables, and goals change. The geometry does not.
Upcoming sectors
- Healthcare · ICU capacity corridors and clinical pathway reachability
- Strategic Planning · M&A due diligence and mandate feasibility
- Infrastructure · Civil project constraint geometry
- Defence · Mission planning under resource bounds
The structural question
“Is this position still structurally exitable? Is this mandate still reachable from today's state, or are we operating in a structural mirage?”
What existing tools miss
VaR is backward-looking. Monte Carlo samples probability distributions. Neither maps the geometry of what is structurally possible from the current state.
Bloomberg Risk, FactSet, VaR, Monte Carlo stress tests
Use cases
- PMS Sharpe and drawdown mandate feasibility
- Hedge fund leveraged position exit windows
- Credit book orderly liquidation corridors
- Adversarial stress: named weak link identification
- Multi-desk coordination deficit under shared liquidity
- Portfolio OS: 6-endpoint structural audit
Solver endpoints
- /v1/simulate· Full structural runway: verdict, epsilon-star, intervention deadline
- /v1/solve/fragility· Distance-to-collapse and anisotropy: which axis is locked
- /v1/solve/adversarial-allocation· Rational attacker budget: which asset class breaks first
- /v1/solve/saddle-point· Sustained stress clock: rounds to collapse under pressure
Forensic proof
Archegos Capital · Portfolio OS
Impossibility Proof
When recovery is impossible,
Niyati proves it.
Not an alert. Not a score. The exact constraint, the exact arithmetic, and the exact reason, in a certificate your team can act on.
STRUCTURAL IMPOSSIBILITY CERTIFICATE
INITIAL STATE
PROOF OF STRUCTURAL IMPOSSIBILITY
Condition: systemic_pressure > 0.90. At t = 0: systemic_pressure = 1.55.
Each injection reduces pressure by 0.10. cascade_spike adds 0.10/step. Net: 0.
To reach systemic_pressure <= 0.60: reduction of 0.95 required.
Max injections: floor(12.0 / 3.0) = 4. Max achievable reduction: 4 × 0.10 = 0.40.
0.40 < 0.95. The goal constraint cannot be satisfied regardless of action sequence.
Omega = 0. No reachable state satisfies both goal conditions simultaneously.
This is not an estimate. It is a proof.
The Core Observation
“Systems don't fail because of bad decisions.They fail because the space of good decisions disappears.”
Pilot Access
Run it on your system.
Active pilots in finance, energy, and logistics. Not demos. Production integrations on real data. Tell us your domain.
nischay@causalorlabs.com