Pathways to
the Post‑Web

Decompose Yourself
or Be Unbundled by Agents

The race is on.

Every firm, institution, and system is a bundle of knowledge, skills, and processes currently legible only to humans. AI agents are already beginning to reverse-engineer these from public surfaces — inferring process logic from patterns of behaviour, reconstructing skill taxonomies from outputs, building approximate knowledge graphs from published material.

The question is not whether your institution will be decomposed into machine-readable primitives. It is whether you control the terms.

This is not hypothetical. It has already happened twice in AI alone. Large language models scraped the open web to build products. Generative AI tools ingested creative industries without consent. By the time media organisations tried to invoke copyright protection, the value had already been captured by others. Now generalise this. What happened to media will happen to every firm and institution on the planet — not by a handful of well-funded labs, but by thousands of agentic systems simultaneously, at machine speed. The firms that move first get sovereign decomposition. The laggards get extracted.

This creates two distinct paths into the Post-Web — one for the existing economy, one for what does not yet exist. Together they constitute the complete transition framework.

The Disappearing Web

Post-Web Chapter 1 — The foundational claim.

The web as we know it is disappearing. AI agents are replacing human browsing. The browser paradigm ends not with a replacement browser but with the disappearance of browsing itself.

The web — as a document-centric, browser-mediated, advertising-funded paradigm — is being unbundled and replaced. Each era of the web added a new capability: Web1 gave us Read. Web2 added Write. Web3 introduced Own. The Post-Web adds the critical fourth: Delegate.

Users no longer interact with the web directly. They instruct agents to act on their behalf. When delegation becomes the primary mode of interaction, the entire architecture of the internet shifts from capturing human attention to executing human intention.

The Universal Interface

What replaces the browser — and how every organisation must now build two front doors.

What replaces the browser is not another browser. It is the user's preferred LLM — a single conversational surface through which all information, services, and decision-points are rendered. Most apps, most websites, most purpose-built interfaces become redundant. They collapse into data sources that the LLM consumes and re-presents contextually.

The direction of flow inverts. Users no longer navigate to information. Information — and the human-in-the-loop decisions that require their attention — comes to them. The LLM becomes the last interface: adaptive, personal, and capable of rendering any service without requiring the user to learn a new app, visit a new site, or manage a new login.

This is not a prediction about a possible future. It is a description of a structural transition already underway. The consequences ripple through every institution, every business model, every system that was built for the browsing paradigm.

The Dual Front Door

If the LLM is the universal interface, every organisation must now maintain a dual presence: a visual website for humans to browse, and a machine-readable gateway that exposes their underlying services so AI agents can programmatically discover, invoke, and transact with them — without ever parsing a web page or requiring custom developer integrations.

The mechanism for this is already emerging. Protocols like Anthropic's Model Context Protocol (MCP) act as a universal discovery layer on top of existing APIs — translating complex endpoints into standardised tools that give the AI a clear instruction manual for interacting with external systems. Historically, websites were built for human eyeballs and traditional APIs were built strictly for software developers. MCP opens the third channel: Business-to-Agent.

Two Modes of Agent Interaction

Not all agent interactions are equal. Where the AI is the application — a dynamic assistant reasoning through unpredictable prompts — protocol-mediated discovery is essential. The agent must find, interpret, and invoke services contextually. The token cost is justified by the complexity of reasoning.

But where the AI builds the application — coding an automated pipeline, a trading bot, a dashboard — the final artefact should bypass the discovery layer entirely and call APIs directly for zero-latency, zero-overhead execution. The protocol layer serves as a development sandbox; the production system runs natively against the underlying infrastructure.

This distinction matters for decomposition. When a firm exposes its primitives, it is not choosing between human interfaces and machine interfaces. It is building both — and understanding that the machine interface itself has two registers: one for reasoning agents that need contextual discovery, and one for automated systems that need raw speed.

From the Attention Economy
to the Intention Economy

Agents care about optimal outcomes. They don't care about ethics. The web will shift from capturing attention to fulfilling intent.

The Attention Economy was the defining economic logic of Web2. Platforms competed for and monetised human attention. The entire architecture of the web was optimised for capturing and selling it — advertising, engagement loops, dark patterns, algorithmic feeds designed to maximise time-on-platform.

The Post-Web replaces this with the Intention Economy: a highly contextual, deterministic yet adaptive, verifiable machine-optimised internet. When agents act on behalf of users, they execute on intent, not captured attention. The economic model shifts from "how long can we keep this person's eyeballs?" to "what does this person actually want accomplished?"

This is not just a business model shift. It is a structural change in how value is created and captured. Any system still optimised for attention is, by definition, not Post-Web.

The Design Constraint

Organisations must be built within the paradigm of intention — aligned around conviction and outcomes rather than metrics that drive attention. This applies to both paths:

In Path 1 (decomposition), your primitives must be structured for intention-execution, not attention-capture. A firm that decomposes itself but structures its primitives around engagement metrics has just digitised its Web2 self.

In Path 2 (constitution), your founding principles must encode intention-alignment from the ground up. No dark patterns. No engagement loops. No metric that rewards time-on-platform.

The Post-Web Stack

Post-Web Chapter 2 — Four primitives that together constitute the Post-Web infrastructure.

These four primitives co-constitute the Post-Web. The stack builds upon Web3 infrastructure and optimises it for a new primary user: AI agents. Wallets evolve from simple digital asset storage into comprehensive control centres for personhood and resource management. Distributed ledger technology provides a system for verifying both human and AI agent activities on-chain.

The Post-Web introduces a new Agentic Layer, where AI agents act autonomously on behalf of users — interacting with DLT and smart contracts as intermediaries that navigate complex ecosystems. This is the Convergence thesis, first articulated by OV in 2016, now fully realised.

Agents as Economic Actors

The most consequential implication of this stack is that agents gain increasing levels of economic agency. They do not merely execute tasks — they hold wallets, transact value, enter agreements, manage portfolios, and allocate resources. Each layer of the stack extends the frontier of what an agent can do autonomously on behalf of its principal.

This is not a technical curiosity. It is a fundamental shift in what it means to participate in an economy. Agents begin as tools — executing discrete instructions. They evolve into delegates — holding budgets, negotiating terms, routing capital. Eventually they operate as economic representatives of our personhood: acting with delegated authority that reflects our values, preferences, risk tolerances, and intentions. The agent does not replace the human in the economy. It extends the human's economic presence into environments and timescales no individual could navigate alone.

This is why decentralised identity and crypto rails are not optional layers — they are foundational. Without sovereign identity, the agent cannot credibly represent a specific principal. Without programmable value transfer, the agent cannot exercise economic agency. The stack is not a feature list. It is the minimum infrastructure required for agents to carry our economic personhood into the Post-Web.

Decompose What Already Exists

When there is a firm, institution, or system.

The existing economy holds the vast majority of the world's operational knowledge, process logic, and skill capacity — but it is trapped in human-readable, human-executable form. For the Post-Web to happen at scale, institutions must be rendered into primitives that agents can navigate, invoke, and recombine.

APIs are the connective tissue. Once an institution's knowledge, skills, and processes are decomposed into machine-readable primitives, they must be exposed as composable programmatic interfaces — APIs — that agents and other systems can discover, invoke, and chain together. Discovery protocols like MCP then act as the universal layer through which agents find and reason over these interfaces — the dual front door described earlier. The API layer is where primitives become actionable: it is the mechanism by which agentic workflows recompose decomposed capabilities into entirely new surfaces — websites, apps, media, products, and services that are generatively rendered based on context and intent, not pre-built and shipped as fixed artefacts.

The Institutional Primitives

Knowledge Graphs

The structured topology of everything an institution knows — its domain expertise, pattern recognition, accumulated insight, relational understanding. Not archived as documents. Structured as queryable, machine-navigable graphs with typed relationships and full provenance.

Critically, knowledge graphs can be permissioned or unpermissioned, fee-gated or freely queryable. This is a design decision with constitutional weight. A firm may choose to gate its proprietary domain knowledge behind token-metered access while making its general taxonomy openly queryable. An open-source project may expose its entire graph permissionlessly. The access model is not a detail — it determines who can build on your primitives and under what terms.

Skills

The catalogued capabilities of what an institution can actually do — its competencies rendered as invocable functions. Deal sourcing. Token design evaluation. Content production. Thesis construction. Each skill becomes a discrete, composable unit an agent can call on.

Processes

The formalised workflows by which an institution operates — its investment pipeline, governance model, production methodology, portfolio support cadence. Converted from informal human routines into executable, composable directed acyclic graphs that agents can run, fork, and recombine.

Machine-First Surfaces

Assume the primary user is agentic. Products and services are designed for agent navigation and invocation first, human experience second. The human interface is a rendering of the machine-readable substrate, not the other way around. Your API matters more than your UI. Your agent-navigable surface matters more than your website.

Surfaces themselves are not fixed products. They are generatively rendered — thin shells composed on demand from the underlying primitives, shaped by context and intent. A website, an app, a media experience, a service offering: each is constituted dynamically from knowledge graphs, skills, and processes at the moment an agent or user invokes them. The surface may never exist in the same form twice.

Sovereign Decomposition vs. Extraction

Self-decomposition is a defensive act. You decompose on your own terms — structure your primitives, price them, gate access, govern how they get recombined — or agents do a worse, uncontrolled version from your public surface and you capture nothing.

This is not hypothetical. It has already happened — twice — in AI alone. Most large language models, OpenAI's included, scraped the open web as training data to ingest and create products. By the time media organisations attempted to invoke copyright protection, the advantage had already been taken. The training data was consumed, the models were built, and the value had been captured by others. Similarly, generative AI tools for image, video, and music ingested vast corpora from the film and music industries without consent — decomposing creative output into machine-readable patterns and recomposing it into competing products.

Now generalise this. What happened to media and creative industries through centralised AI labs will happen to every firm and institution on the planet through swarms of autonomous agents — at machine speed. Agents will reverse-engineer operational logic, reconstruct knowledge graphs from public surfaces, and approximate skills from observable outputs. The critical difference: this will not be done by a handful of well-funded labs. It will be done by thousands of agentic systems simultaneously, making it impossible to protect through regulation. By the time organisations understand what has happened, the advantage has already been captured by new agentic firms and systems.

The firms that move first get sovereign decomposition. The laggards get extracted. Most institutions don't even know they're in the race.

Constitute What Doesn't Exist Yet

When there is no prior system.

The second path applies when there is no institution to decompose. Here the task is not unbundling but formation — designing the constitutional conditions from which a new system can emerge bottom-up.

This is principle-first design. Rather than planning the outputs, you design the conditions — the constitutional principles, the grammar of interaction, the methodology of formation — and let the system grow through stigmergic coordination between human and agent participants.

The meaningful creative act is not producing the output. It is designing the conditions from which outputs become possible.

The Founder as Systems Designer

In the Post-Web, the role of the founder or CEO shifts fundamentally. They are no longer the chief executor of a business plan. They are the systems designer — the architect of the constitutional conditions, the incentive structures, and the coordination mechanisms from which a system self-organises.

This is where incentive design becomes the core competency. The game theory that governs how participants behave — how value flows, how contributions are rewarded, how bad actors are penalised — must be hard-coded into the system through cryptographic enforcement. Tokens and smart contracts make incentive design programmable and tamper-proof. The constitution is not a document that participants are asked to respect. It is an executable set of rules that participants cannot circumvent.

This is a profound shift. Traditional organisations rely on management hierarchies, legal contracts, and cultural norms to align behaviour. Constituted systems encode alignment into the infrastructure itself. The founder's most consequential act is not building a product — it is designing the incentive architecture that makes the system self-sustaining, self-correcting, and resistant to capture.

Systems Not Startups

Post-Web Chapter 3 — Zero to Many.

We will no longer be investing in classic startups but rather agentic systems.

The consequences of the agentic internet mean the death of zero-to-one classic startup theory — how founders design, build, scale, and operate companies. Zero to Many replaces Zero to One: startups, tokens, and agents combine to create network effects that compound at intersection points. Value scales non-linearly rather than linearly.

The unit of innovation is the system, not the company. This has different implications for each path:

In Path 1, you are not optimising the firm — you are decomposing it into primitives that feed systems. The firm's value isn't in its bundled offering but in the quality and composability of its primitives.

In Path 2, you are not founding a startup — you are constituting a system. The grammar applies: thesis, substrate, modular tools, instances. Agentic protocols outperform on every dimension: they improve autonomously, coordinate without overhead, and compound via network effects rather than headcount.

Surfaces Are Generative

A surface in the Post-Web is not a fixed artefact. It is not a static website, a pre-built app, or a packaged product. A surface is a generatively rendered output — a website, application, media experience, product, or service that is dynamically composed based on context, intent, and the primitives available at the moment of invocation.

This is the thin web principle: surfaces carry minimal fixed structure. They are thin shells that pull from the substrate — knowledge graphs, skills, processes — and render contextually for the requesting agent or principal. The same underlying primitives can generate radically different surfaces depending on who is asking, what they need, and what context they bring. A product is not built once and shipped. It is constituted on demand from composable primitives, rendered as a surface that may never exist in the same form twice.

The Grammar

The Solo Founder
and Hybrid Teams

New systems assume the 100x Founder — a single operator augmented by AI who matches small teams on execution speed. AI augmentation allows a new class of founder to operate with 10-100x the output previously requiring a larger team. The founder becomes an orchestrator of AI systems rather than executor of tasks.

Systems are designed for the founder gig economy: minimal overhead, maximum composability, hybrid human-agent teams. The constitutional design must enable individual actors to participate at institutional scale.

This is already demonstrable. Consider a single documentary director using the Post-Web substrate and generative tools — they effectively replace the research team, translation unit, cultural consultancy, archival licensing, and post-production infrastructure. This is not efficiency. It is constitutional amplification: the founding design enabling individual creative acts of a kind previously available only to institutional actors.

What This Means for Both Paths

In Path 1, decomposed institutional primitives become the raw material that solo founders and hybrid teams recombine into new products and services — accessing capabilities that previously required building or joining a large organisation.

In Path 2, systems are constituted from the ground up for this new actor profile. The constitutional design assumes small teams, high agent augmentation, and composable infrastructure. If your system requires a 50-person organisation to operate, it is not Post-Web.

Agents Ruthlessly Remove
Extractive Intermediaries

Agents care about optimal outcomes. Moats based on information asymmetry, bundling, and switching costs are vulnerabilities, not strengths — because agents find the less extractive path.

Agentic systems disintermediate anything that is too extractive. Aggressive rent-seeking becomes a vulnerability because agents will route around it. Every moat that depended on human friction — complexity, information asymmetry, bundling, switching costs — dissolves when the user is an agent optimising for outcomes.

Design for Minimal Extraction

This means you design for minimal extraction from the start. Overt aggressive rent-seeking will be removed as an inefficiency by agents. The general consequence is downward pressure on prices and margins across the entire economy.

Tend Towards Free

Most things will be commoditised. Systems tend towards close to free, if not free. This is the economic gravity of the Post-Web. You cannot fight it — you must design for it.

This is not a threat to be mitigated. It is a design constraint to be embraced. Minimally extractive systems that deliver genuine value will outperform extractive ones because agents will route all traffic and transactions towards them.

Initiator as Power User,
Not Gatekeeper

If your system tends toward free, if agents remove extractive intermediaries, if moats dissolve — then the way you benefit from constituting a system is not by owning or gatekeeping it. It is by being its most sophisticated, most deeply embedded user.

You benefit from the system by being the actor who understands it best, who has the deepest relationship with the substrate, who compounds the most value through use. The founder's advantage is competence within the system, not control over the system.

This resolves the tension between openness and founder incentive. It explains why self-decomposition is worth doing even though it exposes your operational logic. It explains why constituting a system that tends toward free is rational.

The DeFi Proof

This principle is already demonstrated in early DeFi. Protocols like Uniswap, Aave, and MakerDAO are entirely open-source and forkable — any actor can copy the code and launch a competing version. In theory, they should have no moat. In practice, they have built durable network effects that forks consistently fail to replicate. The advantage is not in the code. It is in the depth of the initiators' relationship with the system — the liquidity, the integrations, the governance participation, the accumulated trust, and the compounding expertise of being the most embedded users of their own protocols.

This is the Post-Web pattern. The same logic applies at every scale. A firm that decomposes itself into primitives and then becomes the most sophisticated recombiner of those primitives captures more value than one that tries to gatekeep access. An initiator who constitutes a system and then becomes its power user benefits more than one who tries to extract rent from every transaction. The protocol is open. The advantage is in the competence.

The Token as
Constitutional Enforcement

Translating principles into game theory.

Tokens aren't hype. They're governance infrastructure.

The constitution defines what a system values. The token translates those values into incentive structures that actually govern behaviour. Without the token, the constitution is aspirational. With it, the constitution becomes executable game theory.

The token is the interface layer between system constituents and stakeholders. It is the mechanism by which a system's values become economically binding. This is a significant evolution beyond "token as fundraising instrument" or even "token as coordination primitive" — this is token as constitutional enforcement mechanism.

Tokens Are Dynamic

Previous generations of tokens were fundamentally static — fixed-supply assets with predetermined issuance schedules and rigid utility definitions. Post-Web tokens are dynamic. They respond to system state, adapt to agent behaviour, reweight incentives based on real-time conditions, and evolve their function as the system matures. A token's parameters — its emission rate, staking requirements, governance weight, access permissions — are not set once at launch and left unchanged. They are programmable variables governed by constitutional rules and adjusted through on-chain governance as the system learns.

This is what makes tokens capable of enforcing constitutions rather than merely representing value. A static token can price access. A dynamic token can govern behaviour — rewarding contribution, penalising extraction, rebalancing incentives, and adapting to emergent conditions without human intervention.

Token Functions in the Post-Web

As dynamic, multi-functional primitives, tokens serve several critical roles simultaneously:

Incentive Alignment

Tokens power incentives that align agent behaviour with constitutional principles. Well-designed token economics create systems where individual optimisation serves collective outcomes — the constitution expressed as economic gravity.

Accountability for Autonomous Agents

Through staking and slashing mechanisms, agents or their operators post collateral that is forfeited in cases of faulty behaviour. Reputation must be enforced algorithmically in environments where trust cannot rely on human oversight. The token makes accountability programmable.

Governance Automation

Tokens automate governance, coordinating agent activity and embedding trust in ecosystems where humans and machines collaborate in real time. The governance layer is not a committee — it is a token-mediated coordination protocol.

Ownership and Access

Rather than centralising control in platforms, tokens enable ownership rights over digital agents and infrastructure, grant permissioned access to private services or datasets, and facilitate revenue sharing with aligned incentives among stakeholders.

Minimally Extractive Systems

The MVE Requirement

An autonomous agent economy needs a Minimum Viable Economy before bootstrapping. Launch demand, then justify a token. Never the reverse. Token incentives without MVE collapse post-subsidies. This is a strong empirical pattern derived from 400+ portfolio investments.

Token as Contribution Primitive

If systems tend toward free and rent-seeking gets disintermediated, the token cannot function as an extractive toll. Instead, it becomes a reputation and contribution primitive — tracking who has contributed what to the system and routing value accordingly, without creating the kind of tollbooth economics that agents would disintermediate.

The token tracks the power user's accumulated contribution. It rewards depth of engagement with the substrate. It aligns the initiator's benefit with the system's health — not through gatekeeping, but through recognised competence and contribution.

Conviction and Ownership

Minimally extractive systems still need a mechanism for aligning long-term commitment with economic reward. This is where conviction markets become relevant — token-based systems where the depth and duration of a participant's commitment is reflected in their economic exposure. Ownership is matched with conviction: those who contribute most deeply to the system's innovation and growth are those who hold the greatest stake in its outcomes.

This resolves a tension that plagues both traditional equity and speculative token markets. In the Post-Web, tokens can be designed so that ownership, innovation, and income converge — where the actors who build, improve, and sustain a system are the same actors who benefit economically from its success. Not through rent extraction, but through demonstrated conviction and contribution.

The Lineage

This understanding evolved over a decade: Community Token Economy Paper (2017) → Hungry Protocol (2018) → Token Advantage (2025). Tokens evolve from static speculative assets into dynamic operating infrastructure for the machine economy — powering incentives, automating governance, coordinating agent activity, embedding trust. In the Post-Web, tokens are not fundraising instruments with fixed parameters. They are adaptive coordination primitives that make constitutions executable and responsive.

Intention Theory
for Agentic Systems

When the primary user is agentic, the design of systems must account for how agents form, hold, and execute intentions. This draws on intention theory — the philosophical framework for understanding future-directed plans, commitment, and structured deliberation.

Agents Hold Intentions, Not Just Tasks

An agent acting on intention does not just execute a single command. It maintains a plan, adjusts to obstacles, holds commitments across time. When an agent acts on behalf of a principal, it is holding an intention — a structured, future-directed commitment to achieve an outcome.

Collective Intentionality

When multiple agents coordinate within a system governed by a constitution and token incentives, they form something like collective intentionality — shared, structured purpose that emerges from the constitutional design. The constitution defines the shared intention space. The tokens align individual agent behaviour with collective intentions. The knowledge graphs, skills, and processes are what agents draw on to fulfil those intentions.

The Intention Stack

The Cypherpunk Trinity

The Post-Web, properly built, is cypherpunk-native. The ideological backbone is the Cypherpunk Trinity: privacy, sovereignty, and cryptographic proof.

Decentralisation as constraint — prevents monopoly capture. Exit rights as selection — people and agents can leave systems that fail them. Cryptographic proof replacing institutional trust — verification without intermediaries.

Sanctuary Technology

We are entering a period of increasing geopolitical volatility, institutional fragility, and economic violence. Supply chains fracture. Currencies destabilise. Surveillance expands. In this context, the cypherpunk principles are not ideological luxuries — they are survival infrastructure. As Vitalik Buterin has argued, we need sanctuary technologies: systems that protect human agency when the surrounding environment becomes hostile.

The Post-Web, built on the Cypherpunk Trinity, is sanctuary technology by design. Decentralised systems cannot be seized by a single jurisdiction. Cryptographic proof cannot be revoked by a change in political leadership. Exit rights mean that no individual or agent is trapped within a failing regime — digital or political. These are not features. They are sanctuary economics: economic systems that remain functional, accessible, and sovereign regardless of what happens around them.

This is the moral case for the Post-Web at its most urgent. The question is not whether we prefer decentralisation aesthetically. The question is whether we build economic infrastructure that can shelter human agency through the volatility ahead — or whether we remain dependent on systems that can be captured, censored, or collapsed by forces beyond our control.

The Cypherpunk Filter

Systems that fail the cypherpunk test — that centralise control, exploit attention, or don't provide exit rights — are not Post-Web regardless of branding. This applies to both paths:

In Path 1, self-decomposition is a sovereign act. The institution retains control over its own legibility. External decomposition — where agents extract your primitives without consent — violates the sovereignty principle. The cypherpunk framework gives institutions the moral and architectural case for proactive self-decomposition.

In Path 2, constitutional design must embed sovereignty, exit rights, and cryptographic trust from the first principles. A system that traps its participants is Web2 in different clothing.

The Biological Foundation

Stigmergy: the deepest model is from complexity biology, not management theory.

Termites build cathedrals without blueprints. Murmuration happens without choreography. Order emerges from constitutional conditions encoded in the environment.

Stigmergy is the mechanism by which agents in a complex system leave traces in the shared environment that constrain and enable the next actor's behaviour. The constitutional author operates stigmergically:

They do not direct the system. They encode the first traces — the founding principles — into the substrate. Every subsequent actor is constrained and enabled by those founding conditions. The structure that emerges is not planned — it is constitutional.

The Founding Amplification

The founding act is the most consequential act. The further the system moves toward autonomous operation, the more consequential the original constitutional design becomes. Founding principles operating unsupervised at scale are not diminished — they are amplified.

This is why constitutional design matters more than product design. The product is an instance — a query applied to the substrate. The constitution is the substrate itself. Get the constitution right and the instances take care of themselves. Get it wrong and no amount of product execution will compensate.

The Complete Transition

Path 1 is the existing economy becoming Post-Web. Path 2 is native Post-Web creation. Together they constitute the complete transition framework — one for the brownfield, one for the greenfield.

The Operating Principles

What Emerges from Both Paths

Both paths converge on the same output layer: surfaces — generatively rendered products, services, media, and system integrations composed on demand from primitives. Not pre-built. Not fixed. Constituted contextually by agents fulfilling intentions.

The difference is whether the primitives were extracted from something existing or grown from first principles. The economics are the same: minimally extractive, tending toward free, governed by tokens that enforce constitutional principles, navigated by agents fulfilling intentions.

The firms that move first on Path 1 get sovereign decomposition. The laggards get extracted. The builders who see the greenfield clearly enough get to constitute something entirely new. And those who understand both paths get to do what matters most — become the most sophisticated users of the systems they helped create.