The What-Next Algorithm

What next?

Every system faces this question at every moment. The telco switch selecting a carrier for a voice call. The factory scheduler assigning a work order to a work center. The robot negotiating a handoff with a stranger it has never met. The person standing at a fork, deciding by principle or by convenience.

The answer is always the same structure.

Find the intent. Find the route. Move through infrastructure. Settle the result. Feed it back.

Five steps. Every system. Every scale. Strip the domain away and the algorithm is identical whether it runs in a switching rack, on a factory floor, across a fleet of physical AI agents, or inside a single human making a decision under pressure.

The Question Hiding

Most systems do not know they are running an algorithm. They call it "operations" or "process" or "the way we do things." That is how the algorithm stays implicit — and an implicit algorithm cannot be tuned. The competitor who makes it explicit can adjust the constants. The one who leaves it unnamed gets adjusted by them.

The telcos that extracted maximum margin per route won the quarter and lost the network. The carriers that invested in quality — maintaining thresholds that served both sides, sharing reference data, honoring trade agreements — built the goodwill that compounded. Same routing algorithm. Different north star. The north star is not part of the algorithm. It is the input the algorithm cannot supply for itself.

Three North Stars

There are only three versions of the what-next algorithm. The difference between them is not the steps — the steps are identical. The difference is what the setpoint serves.

A loop without a setpoint amplifies. No correction, no target, no gauge. Doom scrolling, panic selling, a growing god-file nobody can navigate — all runaway loops. The algorithm runs. Output compounds. Nothing improves.

A loop with a setpoint corrects. The routing algorithm finds the cheaper carrier, hits the QoS threshold, stops. It delivers what it was told to optimize. But a setpoint can serve the wrong thing. Optimizing for margin per route while the network degrades is a corrective loop pointed at extraction.

The third version adds two constraints the others lack: the setpoint must be tested against reality, and it must serve beyond the system running it. Not just "we hit the metric" — did the metric measure what mattered? Not just "we optimized the route" — does the route strengthen the network or weaken it? This is the validated virtuous feedback loop — the algorithm that runs on top of the other algorithms to improve them.

The Stations Named

The VVFL diagram is the essential algorithm with all stations labeled. Capture and Priorities are where intent forms. Attention and Value are where routing happens — where energy goes and what it transforms. Systems is infrastructure, the self-correcting layer that reduces decision fatigue. Standards and Distribute are settlement — the gauge reads reality, the output releases. Reflect and Evolve are feedback — the controller reads the gap, the outer loop updates the setpoint.

The center of the diagram is Flow. That is not a station. It is what happens when all five steps run clean — when the algorithm is explicit, the setpoint is right, and the gauge is real. Flow is not a reward for working hard. It is the output of good architecture.

What Gets Lost

The routing algorithm was a decision-making system. The most valuable output was not the routes — it was the traces showing why routes were chosen. Every exception, every override, every blocklisted carrier encoded judgment that no rule book captured. The telcos that fixed a routing arbitrage and moved on got exploited again when reference data drifted. The ones that persisted the decision trace built compound intelligence.

Rules tell you what should happen. Decision traces capture what did happen and why. That gap is where enterprises bleed margin — and where individuals repeat the same mistakes at different scales.

The bridge is the trace made structural. Not documentation. Not communication. A template improved so the next traveller inherits the answer without needing to find it again. That is what makes the loop virtuous rather than merely corrective.

The Same Algorithm

A manufacturing job routing sheet. A telco switching algorithm. A physical AI commerce flow running across nine interoperability pillars. A work delegation matrix that routes jobs between human judgment and AI execution. A person deciding at a fork whether to act by principle or convenience.

Same five steps. Different matter. The routing intelligence — the quality of the reference data, the clarity of the setpoint, the fidelity of the gauge — determines whether the system extracts from the network or enables it.

The what-next algorithm is not something you install. It is already running. The question is whether you know what it is optimizing.

Context

• Essential Algorithm — The five-step pattern in full, with domain table and the XV as coded algorithms
• Validate Virtuous Feedback Loop — The VVFL diagram: all stations named, three loop types, the setpoint as the critical element
• The North Star — The fixed reference that determines which loop type runs: extraction, correction, or virtuous compounding
• Scoreboard — The SETTLE + FEEDBACK half: instruments that read whether the routing decision produced the intended state
• Reality — The gauge layer: instruments that confirm the route delivered what was promised
• Routes — Fork / Obstacle / Sign / Bridge: the four moments in any routing decision
• Telco MEV — How the routing algorithm extracts or enables, and why decision traces compound
• Manufacturing Routing — Physical job routing: the same algorithm applied to work centers, capacity, and throughput
• Intercognitive Protocol — Nine pillars that let robots run the what-next algorithm without prior agreements
• Work Charts — The JTBD routing table: which fork goes to human judgment, which to AI execution
• Intelligence Arbitrage — The arbitrageur sees the routing gap before the market closes it

Questions

What is your system optimizing for — and did you choose that, or did it choose itself?

• Which of the three loop types is your primary system running right now — and how would you know?
• If the most valuable output of a routing algorithm is the decision trace, not the route, what are you currently preserving and what are you discarding?
• When the setpoint is wrong, the loop is perfectly efficient at moving in the wrong direction. What would tell you your setpoint needs updating?
• Where in your system is the gap between what the gauge measures and what actually matters — and who has permission to close it?