Agent economics: a lens, not a forecast¶
Sources: Huiwen Han, The Inference Rent (SSRN, 2026), doi:10.2139/ssrn.6582419; Competing in the Post-Code Era (2026), doi:10.5281/zenodo.21550394; The Enclosure of Episteme (2026), doi:10.5281/zenodo.21551616.
When execution becomes probabilistic and replication becomes cheap, the economics of software change: what is scarce moves, and so does the basis of durable advantage.
This page is a short orientation to a research program on the political economy of cognitive infrastructure. It is not investment, procurement, or legal advice. The papers are working papers and preprints; check their status before relying on specific claims.
Three ideas worth knowing¶
Inference rent¶
If model inference is a metered input to production, then platform owners can capture rent at the inference layer. The Inference Rent paper studies manufactured scarcity, physical constraints, and the durability of platform rent in the foundation-model era.
For adopters, the practical questions are:
- Which costs scale with usage, and which are fixed?
- What is the exit cost if a provider changes price or policy?
- Can routing, caching, or local models reduce dependence?
AxisVela relevance: NOETIVELA makes model and endpoint routing explicit, which is a prerequisite for cost and sovereignty control.
Exposure and replication¶
The Exposure–Replication model argues that assets which are highly observable and fully articulable become replicable once model capability crosses an asset-specific threshold. Durable advantage then migrates toward assets that are tacit, hidden, irreversibly experiential, or networked.
For agent builders, this suggests:
- code logic alone is a weak moat;
- proprietary context, relationships, and operational evidence matter more;
- validation and trust can become a premium.
AxisVela relevance: GNOSIVELA, ONTOVELA, and TEKMOVELA invest in context, evidence, and verification rather than only in generated code.
Validation premium¶
As generated outputs proliferate, the scarce good becomes trustworthy validation: knowing which output is correct, safe, and fit for a decision. This is one reason the ecosystem treats verification and evidence as products, not as test scripts.
Build vs buy, through this lens¶
| Question | Why it matters |
|---|---|
| Is the capability a commodity or a differentiator? | Commodity capabilities are candidates for routing or buying. |
| Where does the context live? | Context and relationships are harder to replicate than code. |
| Who owns the evidence? | Evidence is the basis for audit, learning, and trust. |
| What is the exit path? | Contracts and open surfaces reduce lock-in. |
| What is the cost curve? | Inference-heavy designs have different economics from deterministic ones. |
A caution¶
Economics arguments are easy to over-extend. The cited works make specific claims under specific models; this page only extracts design questions. Use the original papers for the actual arguments and evidence.