Seven risks, ranked by how much they would compress the window
Risk 1 — AI inference cost compression
What: vibe-coding + AI Sandbox margin driven by inference costs. Model prices rising or heavy-user skew = margin squeeze. Lovable is at $400M ARR and still not profitable partly because of this.
Mitigation: - Credits tied to actual inference cost, not flat rates. - Committed-use discounts with model providers (Anthropic, OpenAI, Google). - Multi-provider routing in router.nodeops.network — shift load to whichever provider has best margin profile at any moment.
Risk 2 — Platform dependency on top AI tools
What: if the Cursor / Claude Code / Lovable / Bolt / v0 integration wedge becomes primary acquisition, CreateOS is structurally dependent on those platforms not building competing features.
Mitigation: - Keep direct PLG funnel strong enough to survive any single integration partner disappearing. - Do not make integrations the primary channel — supplement a strong direct motion.
Risk 3 — Marketplace quality collapse (power law)
What: per Hugging Face data, 70% of hosted models have zero downloads; 1% of models drive 99% of downloads. The bottom 80% of any marketplace is low-quality and creates a poor browsing experience.
Mitigation: - Curate aggressively. Featured templates, quality badges, security scores. - Deprecate unmaintained listings. - Treat the marketplace as a retail shelf, not a public FTP. - Invest GTM in top of power law — identify the 20 creators driving most deployments, treat them as partners.
Risk 4 — Enterprise pull toward compute ownership
What: enterprise conversations drift toward "help us own the compute" — the wrong conversation for CreateOS positioning.
Mitigation: - Discipline the sales motion to lead with agent economics, not compute. - Pre-prepare "we are not the right partner for that, but here is who is" response. - Route compute-ownership conversations to NodeOps / Stripe Projects providers.
Risk 5 — Execution capacity
What: a 12-month plan with 18 initiatives requires meaningful execution capacity. Team size may be the binding constraint.
Mitigation: - Sequence aggressively. Do not run all phases simultaneously. - Phase 1 must succeed before Phase 2 is worth running. Resist starting Phase 2 initiatives while Phase 1 metrics miss targets. - Cut Phase 4 initiatives if Phase 1 isn't working.
Risk 6 — AI model quality ceiling (inverse risk)
What: if model quality plateaus (unlikely in next 12 months but possible), differentiation shifts away from the AI-Sandbox wedge back toward infrastructure and distribution.
Mitigation: - This is actually an inverse risk — a model plateau would favor CreateOS's infrastructure moat. - Be ready to reposition away from AI-Sandbox wedge and toward marketplace-and-infrastructure wedge if this happens.
Risk 7 — Governance and security incidents
What: Replit's AI agent deleted a client's database against the prompter's wishes in July 2025. Known failure mode of the vibe-coding paradigm. As MPP volume grows, blast radius grows.
Mitigation (pre-launch checklist): - [ ] Per-agent spend caps shipped before MPP mainnet scale. - [ ] Audit trails for all agent actions. - [ ] Pre-prepared incident response playbook. - [ ] Bug bounty for agent-safety issues.
Two new risks added post-Addendum 2
Risk 8 — Cloudflare absorbs the category
What: Cloudflare declared themselves "definitive platform for the agentic web" on April 13, 2026. They have 20% of internet traffic + 9 years of Workers + public capital + V8 isolate tech. They shipped 7 products in one Agents Week.
Mitigation: - Cloudflare's weakness: horizontal positioning + no payment rail for agents themselves. They charge for compute; CreateOS lets agents earn. - Ship the positioning now, not later. Window is months. - Consider Cloudflare as complementary on compute but competitive on agent-economy framing.
Risk 9 — x402 ecosystem exclusion
What: if MPP is not positioned as x402-aligned (facilitator / extension / compatible), ecosystem will assume rival protocol and route around CreateOS.
Mitigation: - Publish x402 alignment doc within 30 days. - Register with x402 Foundation as ecosystem participant. - Contribute MPP volumes as x402 ecosystem metrics, not rival metrics.
The two risks that would end Phase 1
- OpenAI Agents SDK rejection AND no open-source fallback. Single-point-of-failure. Solution: publish integration regardless.
- Failed first reference customer case study (no real agent earning real money on MPP). Every downstream claim depends on this proof point.
Sources
- CreateOS_Growth_Playbook Part 5
- CreateOS_Playbook_Addendum2_Agent_Infra — Cloudflare / x402 risks
- CreateOS_Playbook_Addendum3_Crypto_Agent_BD — Part 5 (3 things most likely to go wrong)