//Insights

Dispatches from the crossing.

Short theses on the agentic economy — written from inside the arena, not from the sidelines. No predictions we aren't building against.

Economics

The seat is dying. We never sold one.

For thirty years, software priced humans: per seat, per license, per year. The model made sense when humans did the work and software helped. That premise is gone.

When agents do the work, there are no seats to count. 83% of AI-native companies already price by usage; the analysts now project that by 2030, some 40% of enterprise software spend shifts to usage, agent, or outcome pricing. The budget line itself is migrating — from renting tools to buying results.

The deeper shift is accountability. A seat license transfers risk to the buyer: you pay whether it works or not. Outcome pricing keeps the risk with the builder — which is exactly why most vendors resist it, and exactly why we don't. Emelum's platforms are priced on what they deliver. When the work is done, we get paid. Not before.

Discipline

Validation precedes visibility.

The agentic gold rush has an announcement problem: platforms revealed at the idea stage, benchmarked in demos, priced on promises. We run the opposite discipline.

Nothing leaves the Foundry with our name on it until it has survived contact with the market — live environments, real workloads, paying customers, measured outcomes. What doesn't validate is retired without sentiment. What does, earns its announcement.

This is slower publicity and faster truth. Customers of agentic platforms carry real operational risk; they deserve operators who filtered the failures out privately, at their own cost. Announcements follow proof. Always in that order.

Structure

What an agentic P&L looks like.

An agentic platform doesn't have SaaS economics. Revenue scales with work performed, not humans licensed. Cost scales with compute, not headcount. Margin improves as the platform learns.

This changes what the asset is worth holding. A SaaS product decays: competitors copy features, churn erodes seats. A validated agentic platform compounds: every unit of work sharpens it, every outcome priced is a proof reinvested. The rational owner of such an asset doesn't flip it at the first mark-up — they hold it while it compounds, and exit only when a strategic acquirer values it beyond its compounding curve.

That is why Emelum is built as a foundry with an asset manager above it — not a studio with a for-sale sign.

The Era

This era rewards the owner-operator.

A VC funds the transition from outside. A SaaS vendor rents tools to it. An agency bills hours against it. None of them own the outcome.

The agentic economy pays whoever builds the platform, runs it, prices it on results, and holds it long enough to compound. That demands a rare combination: foundry speed and family-office patience in the same structure. Speed without patience sells too early. Patience without speed arrives too late.

We've been building ML and AI companies since 2013 — health, energy, defense, cyber, mobility — through hype, winter, and now the crossing. The lesson from every cycle is the same: the durable returns went to whoever operated what they owned.

The full thesis, with the evidence behind it.

The Thesis