context-first · open-source · MIT
An AI-native operating system for go-to-market. Any team size. Context is the load-bearing layer — not the headcount.
Positioning that cites itself. Sales narrative that holds up under audit. Pages, sequences, playbooks, and AEO content all reading from the same canonical layer.
What a six-person GTM team used to ship — one operator end-to-end, or a team running the same shared canonical layer.
Every draft runs through a context check before it ships. The context is canonical — ICP, positioning, voice rules, competitive frame — not a prompt rewritten each time.
Missing context? Refused. Voice drift? Refused. Unverifiable claim? Refused.
pre-publish-check draft.md
REFUSED — 2 blocks
84 patterns, each sourced from 3+ operators. Every skill traces to a named pattern. Every pattern names the operators and research behind it. Nothing invented; everything cited.
Read the patterns: github.com/k3sava/substrate
The traditional GTM org has a dedicated head for every function because each function used to need one. Context infrastructure changes that math.
One operator with substrate covers positioning, content, AEO, performance, sales enablement, and retention — what used to need six dedicated seats. A team of five on a shared substrate layer covers what used to need twenty. It's not augmentation. It's compression.
The roles don't disappear. The dedicated headcount does.
Read the operator corpus: codex.iamkesava.com
Substrate handles the floor: citation, voice, freshness, drift detection, the gate check before every asset ships.
What's worth saying. What's worth shipping. What real value means in the room you're in. That part is still yours.
Solo operator, shared team substrate, or a private fork per client. The framework is the same. The context layer is yours.
One person covering positioning, content, AEO, performance, sales enablement, and retention. What a six-person team used to need dedicated headcount for each function.
Two to five operators on one canonical layer. Every skill reads the same ICP, voice rules, and positioning. No context drift between operators. No repeated setup per run.
Fork per client. Client context stays private under clients/. The public substrate is the base. Patterns that generalise get contributed back.
The whole picture: brand, retention, expansion. What each person in the loop actually moves the business by.
Not a proxy. The actual number. Substrate optimises for it by making each operator more effective — not by making the team bigger.
$ git clone https://github.com/k3sava/substrate
$ substrate --list
$ substrate pre-publish-check <draft.md>
MIT licensed. Read PRINCIPLES.md when you're ready.
MIT licensed. Clone the public repo, add your context under clients/, run any skill against your stack. Private-fork for each client — their context stays private; the framework compounds publicly.
Patterns that generalise get contributed back. That's how the OS gets sharper without leaking anyone's context.
AI-native GTM doesn't make marketing faster. It makes the traditional marketing headcount optional.