GAMS waits, deliberately. Allocation over a living topology only matters once there is a fleet to allocate: agents, sessions, budgets, capacity. The factory builds runtime, ledger, foreman, relevance, and intent first; GAMS arrives when contention does. See the blueprint →
GAMS is the organ designed to hold the operational topology of the organization: who can do what, who substitutes for whom, what's available, what's committed, what's blocked. It is the part of the stack that will turn "I want this to happen" into a feasible plan, or surface, fast, that it isn't one.
Most organizations carry this topology in the heads of a few senior people. GAMS will externalize it. Capacity, dependencies, substitutions, budgets, timelines: all become first-class objects the rest of the stack can query.
When a new project hits, the question is never just "can we do this?" It's: who, by when, instead of what, depending on whom, at what cost, and what breaks if one of those people gets sick? That tangle lives almost entirely in the heads of a few senior operators.
They are the bottleneck — and worse, the topology they carry decays the moment they're offline. New hires don't know it. Cross-team handoffs collapse on it. The fact that the operational topology isn't a first-class object is the single biggest hidden tax in most organizations.
Not just headcount. The full shape: who can do what, what they're already on, how loaded they are, who their viable substitutes are, what budgets are committed against them, and what timelines depend on their throughput.
GAMS is not a project tracker. The design: a constraint solver running over a living topology. When the topology changes — someone leaves, capacity opens, a dependency slips — the solver re-runs and the rest of the stack finds out.
The GAMS view is not a sheet you update. It will be a continuously recomputed picture of the organization's operational shape, refreshed as the underlying ledger changes.
A new project proposal hits the factory. GAMS will return, within seconds: "This is feasible if we pull A off the migration project (4-week slip on that, acceptable per recent priorities), pair B with C on the technical lead (B is at 85% — would put them at 100%), and contract for design overflow for 6 weeks (estimated cost: $42K). Without the design contract, this delays Q3 launch by 5 weeks."
It is not estimating. It is reading. The topology already exists in the ledger. GAMS will just make it legible, answering in seconds the questions that used to take senior operators days. The senior operators are then free for the questions GAMS can't answer.
And the first real GAMS topology is likely the agent fleet itself: which worker bodies exist, what they've claimed, what tokens and compute they consume. GAMS will learn to schedule agents long before it ever schedules humans.
GAMS is designed as the feasibility membrane. Every other organ that proposes work will pass through it. GOMS asks "can this decomposition actually happen?" GEDS asks "is this reallocation realistic?" GRAMS asks "do we have the capacity to take this contract?" GAMS answers — or rejects — fast.