The most-shared business essay of the season says the human moves above the loop. It's right. On IBM i, the platform is built to make that altitude possible — the capability is already in its design. The work that decides the outcome is the part no platform ships: redesigning the org and reorienting the humans to use it.
Peter Diamandis's The Death of the Company (Aug 2026) wraps its five-layer agent stack in a governance band: every action logged, every agent evaluated, rollback ready, a human standing by. It is the entire difference between a system you can trust and one that is quietly lying to you — and the one layer he says a company must build.
On IBM i, it isn't a band you build from nothing. The capability is already in the platform's design — the enforcement primitives are native, and where an access or governance layer still needs to be stood up, the tooling to do it is here and accelerating. That doesn't put you ahead of the two-person startup; it puts you on even footing with them on the one axis the essay obsesses over, and it removes the technical excuse for not doing the part that actually decides the outcome: redesigning the org and reorienting the humans to take advantage of the capability.
Diamandis puts the human above the loop: setting the standard, hitting yes or no, while agents carry the flow. Right picture — but it only holds if the standard is enforced somewhere the human doesn't have to stand. Otherwise "above the loop" collapses back into a person wedged into every transaction, and you've rebuilt the bottleneck you were escaping.
Governance is the human above the loop: setting the standard, holding the exceptions. The gate enforces that standard on every action as it happens. The failure mode is gated: a human mistaken for the gate, stranded in the flow, approving items one at a time. Governance you keep; the gate you push down into the platform. A person doing the gate's job by hand isn't judgment — it's the constraint wearing judgment's clothes.
Get it right and the human governs from above precisely because the gate holds below, without them in it.
The agent-ready platform is a five-layer stack built in order around a core you never rewrite. Walk it as technology and something turns over at each layer: the technical control is an authority decision that was made long ago and encoded. The stack is org history, hardened into a platform.
Agents find you and learn what you permit. The reach is technical; "what you permit" is an authority statement before it is a setting.
The agent understands what you do — the pricing rule's one exception, the month-end sequence. That is encoded operating judgment, made legible without being moved.
The agent reaches the sovereign core — RPG and Db2, the business encoded — without changing it. You reach the core; you never rebuild it. The access path is real and open, not hypothetical — Mapepire is one live example of agents reaching Db2 for i under governed, identity-scoped access. No competitor can buy the core itself, because it was never for sale.
Real work, at speed, under governance. The overnight batch becomes an observable, auditable, agent-driven process, running against your own data, in your own environment.
Then the layer that is the whole argument.
Prove what agents did; stop what they shouldn't. And here is the part IBM i shops keep underselling: the platform already enforces in silicon what everyone else is scrambling to wrap around their agents in software. Object-level security is identity per action: someone was granted that authority. Journaling is a tamper-evident record: someone can be held accountable. Commitment control is integrity before the write lands: the permit boundary, enforced before, not after. An agent proposes; a human commits; the gate holds before the write. This is not a claim we are asking you to take on faith; it is the platform’s documented behavior, gathered on the evidence page.
That is Diamandis's governance band — logged, evaluated, reversible, accountable — except it isn't a band you bolt around the outside. It's the substrate. The reason it can be the substrate is that the authority decisions were made decades ago, by the people who ran the business, and written into the platform. The gate is org judgment, made durable.
But notice exactly what's native and what isn't. The enforcement capability is decades proven and built into the platform — some of it ready as-is, some needing configuration to point at agents rather than people, all of it accelerated by tooling that already exists. The governor — the human above the loop who now sets policy for agents, and the org structure that positions them to do it — is not native to anything. Those rules were written for people doing the work by hand, in an org shaped around that. The platform can be readied for agents far faster than most shops assume; the org that must now govern those agents, and the human in the governing seat, are the parts you build from where you are. The platform gives you the gate’s capability. It does not give you the governor, or the org chart that makes room for one.
Diamandis casts the established firm as the target and the two people with agents as the winners. Read one layer deeper and the casting doesn't so much invert as dissolve. The governance band he says you must build — accountability, reversibility, identity per action, integrity before the write — is capability the "legacy" platform was designed around from the start. The startup builds it from scratch, under deadline. The IBM i shop starts with it native, needing configuration rather than construction, with tooling to close the gap fast. On that axis, the two firms end up even.
Which means the platform was never the thing standing between an IBM i shop and an agent-native future — and once you see that, the excuse to wait is gone. The capability is present and accelerable. What isn't handed to you is the part it was never going to ship: redesigning the org so a human can govern from above the loop, and reorienting people into that role. That is the same restructure the startup faces — the difference is you no longer have a technical reason to postpone it. This is why the transformation here is woven, not bolted: you can't bolt agents onto a watch, and the phase discipline that carries the org and human shift is who governs when agents execute.
The whole argument comes down to a division of labor. On IBM i the gate isn't a roadmap fantasy or a wrapper you buy; the capability is in the platform's design, proven in the substrate, and the tooling to ready it for agents is here and accelerating. That is the platform's half — the tractable half. The firm's half — deciding who governs which loop, redesigning the org so a human can govern from above it, and reorienting people into that role — is the open one, and it's the half that decides whether the company crosses or stalls.
Two questions this piece leaves open, on purpose. Whose authority the gate enforces, and over which loop, is the organization's discipline, not the platform's — carried in the Collective field note this piece belongs to: Diamandis Drew the Map. Two Coordinates Came Back Blank. And what it costs to keep a human fit to govern rather than gated in the flow is the human dimension: Above the Loop Is Not the Same as Free of It. All three close where the canon does — a dispatch from 2029, from inside a company that made the crossing.