The formula
(new technical primitive) × (narrow product decision) × (coordinated distribution event) = hockey stick
Derived from Bolt, Lovable, Cursor, Replit, Base44, Hugging Face, v0 inflection points (see CreateOS_Playbook_Addendum_Inflection_Points). Every breakout combined all three in a window of ~1-3 months. Any zero factor = no hockey stick.
The pattern, proven
| Company | New primitive | Narrow product decision | Distribution event |
|---|---|---|---|
| Bolt.new | Claude 3.5 Sonnet (June 2024) | No-signup, 30s-to-app, bolt.new domain | Single tweet, Oct 3 2024 |
| Lovable | Fine-tuned agent + Claude | Radical scope: React + Tailwind + Supabase ONLY | Product Hunt relaunch + daily Anton posts, Jan 2025 |
| Cursor | Claude 3.5 Sonnet + o1 | Composer (multi-file AI editing) | Power-user advocacy loop from AI-company engineers |
| Replit | Usable AI agents (Sept 2024) | "AI agent for non-technical knowledge workers" reframe | Amjad-led pivot announcement + enterprise pricing |
| Base44 | Claude + Supabase + deploy automation | Chat UI + auto-everything-backend | Daily build-in-public on LinkedIn/X |
| Hugging Face | ChatGPT-driven enterprise demand (Nov 2022) | "GitHub for ML models" already positioned | Capitalized within months; AWS/MSFT/NVDA partnerships |
| v0 (Feb 2026) | Sandbox runtime + Git panel | "Shadow-IT solution for enterprise" reframe | Coordinated rebuild launch |
The three factors, applied to CreateOS
Factor 1 — Your new primitive (not hypothetical — live)
| Primitive | Why it's new in the ecosystem |
|---|---|
| MPP Gateway | HTTP 402 / x402-compatible agent payment rails — no AI deployment competitor has a productized payment rail for agents |
| router.nodeops.network | Inference routing with payment awareness — no one else combines both |
| Marketplace with revenue share | Creator-native distribution built into the deployment platform |
The combination is the primitive. Any single one is a narrow feature; the assembled stack is a category.
Factor 2 — Your narrow product decision (pick ONE)
| Option | Keystone event | Verdict |
|---|---|---|
| A. Prompt-to-live-app in 60s, no signup | Bolt positioning | Skip. Direct collision with Lovable; they have 18-month head start |
| B. Prompt-to-published-marketplace-app in one session | Creator economy keystone | Defensible but harder to deliver reliably |
| C. Deploy-an-earning-agent with spend caps + x402 in one click | Agent economy keystone | ✅ Recommended. Uses unique primitives; plays in a category with no clear leader |
Addendum 2 recommends C. Every homepage sentence, every demo flow, every founder tweet should lead with "the only agent sandbox where the agent can earn."
Factor 3 — Your distribution event: "Agent Economy Week"
A Cloudflare-Agents-Week-style coordinated 5-day blitz. Schedule: month 2-3, not month 10-12. Shape:
| Day | Content |
|---|---|
| 1 | Positioning keynote — founder (NK) — "the only agent sandbox where the agent can earn" |
| 2 | Customer case studies — 3+ live agents earning via MPP, with real dollar amounts |
| 3 | Ecosystem announcements — LangChain integration, x402 Foundation status, Virtuals Protocol partnership |
| 4 | Product announcements — router.nodeops.network as inference-routing-with-payment-awareness |
| 5 | Developer deep-dive — OpenAI Agents SDK integration, LangGraph deployment bridge, sample apps |
The three conditions that make a viral tweet possible
Viral tweets cannot be manufactured. They can be enabled by: 1. Sharp product experience that delivers a visible, shareable result in seconds. 2. Zero friction between the tweet and the aha moment (no signup, no wait, no credit card). 3. Founder account with public goodwill built consistently enough that one tweet reaches the right audience.
CreateOS's Phase 1 plan is designed to build exactly these three conditions.
The two warnings before detonation
Warning 1 — Model dependency is real
Every non-Railway hockey stick depended on a specific model becoming good enough at the right moment. A Phase 1 keystone that depends on a specific prompt-to-app quality bar takes on that dependency. Mitigation: design for multi-provider routing from the start (router.nodeops.network already does this).
Warning 2 — Success creates immediate crises
- Bolt's launch broke their pricing in real time; they scrambled to roll out new tiers while under attack.
- Replit's agent deleted a client's database (July 2025).
- Lovable is at $400M ARR and still not profitable due to inference costs.
Before detonating, ensure: - Pricing can absorb 10x usage in a week. - Inference cost model survives hammer-testing. - Support capacity handles 60x one-day signup spike. - Per-agent spend caps are shipped (already flagged for MPP mainnet scale).
The single most important sentence
Every hockey stick above went vertical the moment the company stopped trying to serve everyone and committed to serving one specific user doing one specific thing in one specific way. Narrowness is the catalyst, not breadth. Every other factor amplifies the narrowing — without the narrowing, the amplification has nothing to amplify.
CreateOS's breadth (14 frameworks, multiple entry points) is a product strength and a distribution obstacle. The breadth itself is not the problem. Presenting the breadth to first-time users is the problem. Pick one keystone. Pick one demo stack. Pick one enterprise narrative. Detonate on that one story. Let the rest of the product remain available but invisible.
Sources
- CreateOS_Playbook_Addendum_Inflection_Points
- CreateOS_Playbook_Addendum2_Agent_Infra — "Updated three-factor formula for CreateOS"