§ 00 ── AI-NATIVE STARTUP OS · SEED – SERIES A
What a Seed-Series A
engagement actually looks like.
For YC-stage AI-native startups. 5–30 employees. $0–3M ARR. Your methodology is forming RIGHT NOW. Per Jack Dorsey's framing of the AI-founder at Block: still builds, still coaches, leads by example — AI strategy stays with the founder, not delegated. Per Diana Hu's YC framing of AI-native operating companies: AI as the operating system the company runs on, not a tool layered on top. The substrate captures founder methodology before fragmentation begins and folklore replaces architecture.
§ 01 ── YOUR THREE PAINS
You know these. We do too.
PAIN 01
“My methodology lives in two heads — mine and my co-founder's. The team grows, and the methodology fragments.”
Every decision you make right now compounds. You'll hire 20 people in the next 18 months. They'll learn “how we decide” through osmosis. OSMOSIS IS LOSSY. By Series B, “how we do things” will be fragmented across teams. The substrate captures your operating principles, decision heuristics, and strategic frames — explicitly, with your confirmation, into a versioned methodology overlay that's yours (per MSA § 8.3 customer IP retention).
PAIN 02
“Customer signal isn't flowing. I learn about a customer concern weeks after the AE first heard it.”
Your 14 design partners are talking. Each conversation lives in someone's inbox, someone's call recording, someone's CS notes. Synthesis happens in your head when you finally have time. The substrate synthesizes continuously across every customer touchpoint — sales calls, support tickets, success notes, product feedback — into per-customer health summaries and segment-level sentiment trends that surface signals in HOURS, not WEEKS.
PAIN 03
“I've made strategic decisions I still debate. No replayable record of why.”
You've pivoted twice. You remember the reasoning for both pivots, but the reasoning isn't replayable. New hires ask “why did we decide X?” and the answer takes 30 minutes of context-loading. The substrate captures every substantive decision with full audit chain context — the evidence available at decision time, the diverge-and-reconcile traces, the actual commitment made. Three years from now, replay is structurally identical (LLM-stochastic at narrative level, structurally deterministic).
§ 02 ── WE BUILT FOR THESE THREE PAINS
Three pains. Three canonical agent capabilities. Each load-bearing.
FIG.01 — Three pains map to three load-bearing primitives
FIG.01 — Three pains map to three load-bearing primitives
01 · KNOWLEDGE CURATOR · CLASS A · T2 · METHODOLOGY CAPTURE
Knowledge Curator
Consent-bounded. Per-meeting opt-in. Detects methodological signals in founder conversations — operating principles, decision heuristics, strategic frames. Surfaces candidates for your confirmation BEFORE encoding. Builds your L2 methodology overlay continuously. The AI-founder posture stays the founder's — the substrate just makes it operational, queryable, replayable. By Phase 1 close, 80+ captures. By Phase 2 month 6, your reasoning is QUERYABLE. At Seed-A, Knowledge Curator runs in extended methodology-capture mode across both founders. Brand DNA Step 0 research (8 themes · LIGHT depth · 25–35 sources) seeds the capture corpus.
02 · CUSTOMER OPERATIONS LEAD · CLASS A · T2 · F2 CONSENT-BOUNDED
Customer Operations Lead
Per-design-partner health summaries running continuously. Sales call synthesis. Support ticket aggregation. Customer success note integration. Escalation flag detection. Drafts customer-facing communications. F2 architectural cap on customer data: per-category opt-in only, no inherited consent, no default-ON. YOU send customer-facing comms, never the substrate — Customer Operations Lead drafts, a human approves and sends. Voice calibration Cohen's weighted κ ≥ 0.85 against your captured voice samples.
03 · DECISION BRIEF DRAFTER · CLASS A · T2 · F3 ADVISORY CEILING
Decision Brief Drafter
Detects substantive decisions across founder conversations, leadership reviews, strategic offsites. Drafts the decision brief per the 5-lens scorer (Risk · Customer-Impact · Regulatory · Reversibility · Cost). Triggers diverge-and-reconcile for substantive class. Captures expected outcomes, tracks actual outcomes, surfaces divergence. F3 advisory ceiling: the autonomous-decision code path does not exist. Decision Brief Drafter drafts, humans decide. Every decision audit-emitted via Audit Curator (Class B · F4 sole-emit · L0 doctrine · NEVER-deletable on critical). Three years from now, REPLAYABLE.
All Class A agents at T2 DRAFTER ceiling. All consent-bounded. All architecturally tier-capped. Your methodology stays yours. The founder leads the AI-native posture; the substrate captures and compounds it. The other 4 Class A canonical agents (Risk & Compliance Lead · Finance Operations Lead · Strategy Lead · Engineering Steward) activate per Customization Decision Tree applied to your Seed-A pattern. Plus 5 Class B always-active (Audit Curator · L8-Enforcer · Registry Maintainer · Eval Suite Runner · Version Controller · all T4 Steward).
§ 03 ── WHY NOW IS CHEAPER THAN LATER
Substrate-grade discipline at Seed-A is cheaper to install than to retrofit at Series C.
SERIES C · RETROFIT
A Series C company installing substrate-grade discipline has to do it while managing 100+ employees, multiple functional silos, established (and partially-fragmented) methodology, and external stakeholder pressure. The retrofit is EXPENSIVE in calendar time and harder in operational reality.
SEED-A · GREENFIELD
A Seed-A startup installing substrate-grade discipline does it while methodology is still in your head, the team is still small, decisions are still founder-led, and stakeholder pressure is still future-tense. The discipline COMPOUNDS from day one. Every replay improves calibration. Every captured decision deepens the audit chain. The startup advantage Diana Hu names at YC: you don't have legacy systems, org charts of routing-humans, or thousands to retrain. You can design your systems, workflows, and culture around AI from day one.
FIG.02 — Install cost vs stage. Slope steepens at every transition.
FIG.02 — Install cost vs stage. Slope steepens at every transition.
Time-to-substrate, like time-to-product-market-fit, has a cost curve. The slope STEEPENS at every stage transition.
§ ── METHODOLOGY DECAY · ESTIMATE
What's leaking right now.
Three inputs. A doctrinal estimate of un-captured substantive decisions per quarter. Conservative on purpose.
UN-CAPTURED · ESTIMATE
624
SUBSTANTIVE DECISIONS · HORIZON WINDOW
RISK SCORE
12/100
DOCTRINAL · CONSERVATIVE
TIER
LOW
READ THE SUBSTRATE FIRST
§ 04 ── THE ENGAGEMENT · COMPRESSED
Compressed for Seed-A scale. No premature heaviness.
The substrate scales to whole-company governance at Series D+, but the engagement model at Seed-A is calibrated to your scale. Compressed Phase 0 (1 week). Compressed Phase 1 (6–8 weeks · KPI gate at Week 5 per canonical Stage Sub-Framework). Lower operational footprint in Phase 2.
FIG.03 — Compressed engagement. Phase 0 is free.
FIG.03 — Compressed engagement. Phase 0 is free.
PHASE 0
Discovery interviews (D1–2) · Brand DNA Step 0 research (D2–3 · LIGHT depth · 25–35 sources · 8 themes) · Methodology Council Workshop (D3–5 · 5-seat Council formed at Seed-A compression · charter ratified · κ baseline lock) · synthetic-data PoC (D5–7 · working prototype you query live, not pitch deck) · GO/NO-GO at end of Week 1. If NO-GO, you keep all artifacts unconditionally (methodology v0.1 overlay · business case · Brand DNA research · PoC code · Council Charter draft · workshop notes).
PHASE 1
Substrate deployment to your dedicated instance. 12-agent canonical fabric activated (5–7 Class A per Customization Decision Tree compressed for Seed-A + 5 Class B always-active). 6 canonical L8 floors (F1–F6) + 0–3 optional (F7 if unregulated · F8 if multi-tenant SaaS · F9 if EU employees) enforced architecturally — code-path absence, not policy. Internal owner transitioned by Week 5. 8 universal KPI gate at Week 5 (+ 0–2 customer-specific K9/K10 if defined at Phase 0). 6-of-8 universal PASS → Phase 2 GO.
PHASE 2
Monthly retainer at lower hourly footprint (matches Seed-A operational reality). Your internal owner runs day-to-day. 5-seat Methodology Council meets monthly + ad-hoc. Quarterly D7 §12 verification (canonical 5 steps: Inspect · Replay · Verify · Cross-check · Escalate). 12-month minimum from Phase 2 start. Monthly cancellation thereafter.
PHASE 3
Often deferred until Series A → B/C funding event triggers Stage Transition. Available when needed for evolution work, stage migration, or functional area expansion. Loop accretion toward Year-2 ~25 agents across 6 loops happens here.
Complexity-priced economics: discussed at Phase 0. No public pricing on this page.
§ 05 ── A WORKED EXAMPLE
What it looks like at a Series A SaaS company.
Ledgerline is a worked example we publish to make the engagement concrete. 22 employees · $1.6M ARR · 14 design partners · founder methodology in two heads. The /worked-example page walks through what Phase 0 produces, what Phase 1 deploys, and what Ledgerline looks like at Week 8 of Phase 1.
WHAT GETS CAPTURED
methodology captures across 5 Seed-A categories by Week 8. Founder voice corpus of 145 samples. CTO methodology in extended Knowledge Curator capture mode. Brand DNA research synthesis κ-graded.
WHAT GETS DECIDED
substantive decisions audit-emitted via Audit Curator (F4 sole-emit) in Phase 1. 12 triggered diverge-and-reconcile. 2 surfaced disagreements to founders for resolution.
WHAT GETS CLOSED
design partner health summaries continuous via Customer Operations Lead. 38 sales calls synthesized. 3 sprints integrated with sprint loop drift detection by Strategy Lead.
§ 06 ── UNDER THE HOOD · NDA-TIER
Phase 0 conversation depth available pre-engagement.
What's public on this page: Seed-A engagement model · 3 canonical primitive capabilities · 4-phase structure · KPI gate at Week 5 · 6-of-8 universal threshold · methodology decay estimate. A 20-min Phase 0 alignment call (no NDA required initially) covers:
- 01
Customization Decision Tree applied to YOUR Seed-A pattern · which 5–7 Class A canonical agents activate · which L8 floors apply (F1–F6 + optional F7–F9)
- 02
Pricing range for Phase 1 fixed-fee + Phase 2 monthly retainer at Seed-A scale (no public pricing — calibrated to your scope at Phase 0)
- 03
Brand DNA research depth recommendation for your stage (LIGHT 25–35 sources typical Seed-A)
- 04
First-loop selection · what loop your engagement starts with (Customer Operations · Strategy · Engineering · Product · per sector + stage adaptation)
- 05
Methodology Council seat composition recommendations (typical Seed-A: founder × 2 + 1–2 advisors + customer advocate = 5-seat compressed Council)
- 06
Worked examples beyond Ledgerline · additional Seed-A engagements across sectors (DevTools · B2B SaaS · etc.) · available at Phase 0
- 07
Replay-test mechanics · model swap survival · structural fingerprint discipline
§ ── NOT YOUR STAGE
The substrate is one architecture. The conversation calibrates per stage.
§ END ── PHASE 0 IS FREE
If you operate a Seed-Series A AI-native startup — Phase 0 is FREE. One week. You keep all artifacts either way (per MSA § 8.3 customer IP retention · enforced architecturally via F-FW1 at operator-tier).
At a different stage?