The Convergence
What pattern governs how your phone call gets routed, how your payment settles, how your sensor reports its data, AND how your attention gets sold?
The same one.
Three events. One week. March 2026.
Stripe and Tempo launched the Machine Payments Protocol — agents paying agents, settling on-chain, a hundred services compatible at launch.
Adeniyi Abiodun, co-founder of Mysten Labs, published the thesis his team has been building toward: optimization functions route to the lowest-friction path. Sui's object model is designed to be that path.
Zhejiang University released SkillNet — 150,000 curated AI skills with a three-layer ontology and five evaluation dimensions. The skills aren't the product. The accumulated intelligence about which skills work is the product.
Three announcements. Same architecture underneath.
The Pattern
Five stages. Every flow.
INTENT → ROUTE → INFRASTRUCTURE → SETTLE → FEEDBACK
Intent. What you want to happen. Call this person. Pay this merchant. Report this reading. Capture this attention. The system doesn't care about the how — only the what.
Routing. Finding the optimal path. Cost, speed, reliability, compliance. Someone — or something — evaluates options and picks the best route. In telecom, this was MEV before MEV had a name.
Infrastructure. The physical layer that carries the signal. Towers, cables, satellites. Rails, chains, bridges. Sensors, hotspots, nodes. Screens, feeds, algorithms.
Settlement. Proof it happened. The CDR that bills you. The on-chain receipt. The timestamped attestation. The impression log. Without settlement, there's no trust.
Feedback. The loop that improves routing. This path was congested. This rail was expensive. This sensor was unreliable. This ad didn't convert. Next time, route differently.
The pattern recurses at every scale. A single payment. A treasury operation. An entire financial system. Same architecture, different zoom level.
Pattern Origin
I spent years building routing algorithms for international voice traffic. Route inbound calls from carriers to destinations across 40+ countries. Every call was a decision — which carrier, which route, which price, which quality threshold.
The routing table was the business. Not the switches. Not the fiber. The intelligence about which path to choose, given the constraints, in real time.
Then the same pattern appeared in crypto. Solvers routing trades. Bridges routing value. Oracles routing data. The abstraction was identical. The settlement layer had changed.
The convergence isn't a theory. It's a biographical fact dressed as infrastructure analysis.
The Fractal
The same five steps govern every scale:
| Domain | Intent | Route | Infrastructure | Settle | Feedback |
|---|---|---|---|---|---|
| Messages | Call this person | Least-cost routing | Towers, fiber, satellites | CDR billing | Quality metrics |
| Value | Pay this merchant | Solvers, bridges | Rails, chains, stablecoins | On-chain receipt | Price, speed, cost |
| Data | Report this reading | Edge AI, oracles | Sensors, hotspots, nodes | Timestamped attestation | Reliability scoring |
| Attention | Reach this audience | Ad exchange, algorithm | Screens, feeds, platforms | Impression log | Conversion rate |
| Agent commerce | Discover capability | A2A, MCP, UCP | Compute, models, tools | SPTs, x402, AP2 | Accumulated skill graph |
| Personal | What do I want? | Priorities, values | Time, energy, relationships | Results, evidence | Reflection, correction |
Six rows. Same columns. The personal row is not metaphor — it's the same optimization problem running on biological hardware.
The Signal
Stripe MPP
Stripe and Tempo launched the Machine Payments Protocol. An agent calls the PaymentIntents API. The protocol routes to the optimal rail — stablecoins, cards, Bitcoin Lightning via Lightspark. Settlement happens on Tempo's blockchain, backed by Stripe and Paradigm.
Shared Payment Tokens handle authorization. Sessions enable streaming micropayments. Rail-agnostic by design.
Design partners: Anthropic, OpenAI, Visa, Mastercard, DoorDash, Shopify, Revolut, Ramp.
Map it:
- Intent: Agent needs to pay for a service
- Route: MPP selects optimal rail (card, stablecoin, Lightning)
- Infrastructure: Tempo blockchain + existing payment rails
- Settle: Shared Payment Tokens prove the transfer
- Feedback: Transaction data improves future routing
The Value flow, automated, agent-to-agent. The agent commerce protocol stack predicted this — A2A discovers, MCP communicates, MPP settles.
Sui Thesis
Sui's object model is built on one conviction: optimization functions route to the lowest-friction path. Make the object model the lowest-friction path.
Objects on Sui aren't abstractions. They ARE the things they represent — the token, the NFT, the permission, the mandate. When an agent authorizes a payment, it doesn't call an API that calls a database that updates a ledger. It touches the object directly. Sixteen published packages. Programmable Transaction Blocks composing multiple operations in a single ~390ms settlement.
The mandate module implements AP2 — agent-to-agent payments in Move. Not theoretical. Deployed.
Map it:
- Intent: Agent expresses what needs to happen
- Route: PTBs compose the optimal execution path
- Infrastructure: Sui validators, object store
- Settle: ~390ms finality, object ownership transfers
- Feedback: On-chain state feeds next decision
Infrastructure and settlement, fused. Machine coordination at machine tempo.
SkillNet
Zhejiang University's SkillNet paper mapped 150,000 curated AI skills into a three-layer ontology: Taxonomy, Relational, Package. Four relation types connect skills — composition, dependency, similarity, hierarchy.
Five evaluation dimensions: Safety, Completeness, Executability, Maintainability, Cost-awareness. Results: 40% reward improvement, 30% fewer execution steps.
The individual skills aren't the breakthrough. The ontology is. It's a routing table for capability — which skill to compose, in what order, given the constraints.
Map it:
- Intent: Agent needs capability X
- Route: Ontology finds optimal skill composition
- Infrastructure: Compute, models, tool APIs
- Settle: Execution result verified against five dimensions
- Feedback: Accumulated graph improves routing for every agent
The Data flow applied to intelligence. The valuable output isn't any single skill — it's the graph that routes to the right combination.
Attention
The $1 trillion advertising industry spends roughly 60% proving attention happened. Impressions without verification. Clicks without intent. Conversions attributed by last-touch guesswork.
Attention follows the same five stages. But its settlement layer is broken.
| Stage | Messages | Value | Data | Attention |
|---|---|---|---|---|
| Intent | Call someone | Pay someone | Report data | Reach audience |
| Route | Least-cost path | Solver/bridge | Edge AI | Ad exchange |
| Infra | Towers, fiber | Rails, chains | Sensors, nodes | Screens, feeds |
| Settle | CDR billing | On-chain receipt | Attestation | ??? (impression log) |
| Feedback | Quality metrics | Price data | Reliability | ??? (attribution) |
Messages settle in CDRs. Value settles on-chain. Data settles in attestations. Attention settles in self-reported impression logs from the platform selling the attention.
The same infrastructure that settles payments in 390ms could settle attention proofs. 375ai is building DePIN billboards — physical proof of display, cryptographically attested. The missing settlement layer for the fourth flow.
The Telco Trap
Rails are crucial early. Then they commoditize.
| Era | Built Rails | Captured Value |
|---|---|---|
| Railroads | Train companies | Retailers, cities |
| Internet | ISPs | Google, Meta |
| Mobile | Telcos | Apple, Uber |
| Crypto | Blockchains | Wallets, agents, platforms |
The trillion-dollar companies won't validate blocks. The value accrues to whoever owns the experience — whoever turns intent into outcome without the user thinking about the infrastructure.
Don't optimize to be the telco. Optimize to be the Uber.
The Inner Loop
The same pattern runs inside you.
| Stage | Infrastructure | Personal |
|---|---|---|
| Intent | What the system must do | What do I want? |
| Route | Optimal path selection | Which priorities, which values? |
| Infrastructure | Physical layer | Time, energy, relationships |
| Settle | Proof it happened | Evidence, results, outcomes |
| Feedback | Improve next routing | Reflection, correction, growth |
A congested route in telecom means calls drop. A congested route in your life means you said yes to everything and finished nothing.
The five archetypes map to the five stages. The Philosopher clarifies intent. The Dreamer routes to possibility. The Engineer builds infrastructure. The Realist settles — measures what actually happened. The Coach feeds back — corrects the routing table for next time.
The Tight Five maps the same way:
| Position | Tight Five | Stage |
|---|---|---|
| 1 | Why does this matter? | Intent — Purpose |
| 2 | What truths guide you? | Route — Principles |
| 3 | What do you control? | Infrastructure — Platform |
| 4 | What do you see others don't? | Settle — Perspective |
| 5 | How do you know it's working? | Feedback — Performance |
Five questions. Five stages. Same architecture. The VVFL is this loop running continuously — settlement feeds feedback feeds better routing.
Device Agency
In classic telco, the network delivered signaling and data. Your phone was a dumb endpoint on a dumb pipe.
In DePIN, that same device can:
- Prove work — coverage provided, energy generated, data captured
- Get paid — tokens for contribution, stablecoins for settlement
- Emit instructions — autonomously via smart contracts
Your base station becomes an agent. Your meter becomes an agent. Your EV charger becomes an agent. They invoice, settle, and respond to price signals in real time.
| Old Model | DePIN Model |
|---|---|
| Device reports usage | Device proves work |
| Telco bills monthly | Smart contract settles instantly |
| Fixed pricing | Dynamic price signals |
| Operator controls | Device responds autonomously |
A solar home mints tokens representing kWh produced. An EV charger sells excess capacity. A 5G small cell earns per-byte, paid in stablecoins.
The device sends money AND instructions. Same message, same network. The Intercognitive Protocol extends this to physical AI — nine pillars coordinating robots, drones, and autonomous systems through the same five-stage architecture.
What It Means
For builders: Stop building on platforms. Build instruments that compete on platforms. The best instrument wins. That's the whole game.
For investors: The value isn't in the rails. It's in the experience layer that turns intent into outcome. Don't bet on being the telco.
For operators: You can own the infrastructure you operate. The coordination protocols exist. The permissionless platforms exist. The question is whether you'll deploy.
The internet was supposed to be permissionless. Then platforms captured it.
DePIN is the second chance. Open infrastructure where the best instrument wins. Agents composing skills from a shared ontology. Payments settling at machine tempo. Attention finally getting an honest receipt.
Same architecture. Messages, money, data, attention. Five stages. Every scale.
How will you add value to a platform that can't block a good idea?
Context
- Telco MEV Algorithm — The origin: routing intelligence as value capture
- Three Flows — The companion page: five stages, three things that move
- Agent Commerce — The standards war: ACP, MPP, x402
- Agent Protocols — The full stack: A2A, MCP, UCP, AP2, VI, PCP
- Advertising Industry — The fourth flow: attention without settlement
- Sui — Sub-second settlement at machine tempo
- Sui Move Development — 16 packages, object model, PTBs
- Payments Industry — $25T B2B flows, stablecoin rails
- Agents & Instruments — The P&ID pattern: Thinker, Valve, Pipe
- The Tight Five — Same five positions, every domain
- Persuasion — Ethos, Logos, Pathos, Kairos, Topos
- VVFL — Settlement feeds feedback feeds better routing
- Intercognitive Protocol — Nine pillars for physical AI coordination
Links
- Stripe — Machine Payments Protocol — The Value flow goes live for agents
- Tempo — Mainnet is Live — The settlement blockchain behind MPP
- SkillNet — AI Skills Ontology — 150K curated skills with three-layer ontology
- MPP Documentation — Stripe's technical specification
Questions
What happens when all four flows share one settlement layer?
- The advertising industry spends $1 trillion to prove attention landed. Blockchain settles payments in 390ms. Why hasn't the same infrastructure settled the attention proof?
- The telco trap says rails commoditize and experience captures value. If agents don't care about experience — only optimal paths — who captures the value in agent commerce?
- The inner loop uses the same five steps as the infrastructure. If the pattern is identical, what does a "congested route" look like inside a person?
- SkillNet's most valuable output is the accumulated intelligence about what works, not the individual skills. What's the equivalent accumulation layer for the attention flow?