JedAI-Nexus

A. Sequenced plan (0–8 months) from prototype to v1, aligned with continuity-first storytelling, NSFW gating, and creator publishing. Dates are placeholders; adjust once team/velocity are set. Phase 0 — Prototype (Weeks 0–6) Scope Single universe, 2–3 characters, short arc. Core memory loop (immutable canon + meters + recap/RAG + contradiction repair). On-demand scene image (one style). Local export only. Milestones M0.1 Turn loop working end-to-end. M0.2 Continuity Guard + targeted repair loop demo. M0.3 Image → cover export. Exit criteria (gates) Continuity score ≥ 80 on scripted probes. p95 latency ≤ 4s (text), ≤ 20s (image). Security & privacy basics (PII isolation, logging) in place. Phase 1 — MVP (Weeks 6–18) Scope Multi-universe registry (house IP), age/consent gates. Publish stories to public shelves with tags/warnings. Basic search & discovery; user library. Pre-publish automated checks + report flow. Milestones M1.1 Universe Builder (lite) + Character Editor. M1.2 NSFW gating live; explicit images require age+consent. M1.3 Publishing wizard + hash/provenance vault for public images. Exit criteria 500 closed-beta users. Story completion rate ≥ 35% for featured arcs. Moderation SLA: ≥ 95% severe reports triaged ≤ 4h. Phase 2 — Open Beta (Months 4–6) Scope Founder pricing tiers (Free / Plus / Creator). Creator world-building (paid tier) + publish universes. Comments + light analytics; referral links. Cost guardrails (queues, caps, caching). Milestones M2.1 Creator Console + world publish. M2.2 Reader comments + report tooling v2. M2.3 Cost/MAU and queue p95 dashboards live. Exit criteria 10k MAU, Free→Paid ≥ 3.5%, cost/MAU ≤ $0.50. Phase 3 — Monetization On (Months 6–9) Scope Creator Pool payouts + direct creator subscriptions (80/20). First licensed author universe (Tier B or flagship pilot). Continuity score used as a discovery gate. Milestones M3.1 Usage accounting + payouts dashboard. M3.2 Author lane ingestion (style/voice locks) + co-branded launch. M3.3 Continuity gates wired to shelves/search. Exit criteria ≥ 500 monetizing creators; payout/creator p50 ≥ $40/mo. Contradiction incident rate ↓ 30% vs. MVP. Phase 4 — v1 Launch (Months 9–12) Scope 3–5 licensed universes; seasonal event. Mobile “reader-lite” (PG-13) with deep links to web. Creator subs/tips live in core EN markets (US/CA/UK/AU/NZ). Milestones M4.1 App-store-safe reader shipped. M4.2 Two additional licensed drops. M4.3 Transparency report v1 + public enforcement stats. Exit criteria 50k MAU; ARPPU ≥ $12; creator RPM (per 1k read-min) hitting target. Phase 5 — v1.5 Scale (Months 12–24) Scope Internationalization (ES/FR/DE) + region gates. Forking with configurable rev-share; co-authoring. Advanced creator analytics. Milestones M5.1 Localization + regional compliance toggles. M5.2 Forking + rev-share contracts in-product. M5.3 Author event tools (AMAs, drops). Exit criteria 250k MAU; ≥ 3k creators earning > $100/mo; cost/MAU ≤ $0.25. Cross-cutting Staffing (minimum viable) Founding BE/Orchestrator, LLM Engineer, Full-stack/Web, Designer/UX, Infra/Ops (part-time), Trust & Safety Lead (part-time), Creator/Author Ops (1), Fractional Legal. Dependencies & risk controls Payments: multi-processor + wallet credits fallback. Costs: small-context prompts, caching, own image inference, burst throttles. Safety: disallowed taxonomy, age/consent gates, automated checks + human review. Continuity: immutable canon, structured facts + RAG, repair loop, continuity gates. KPI gates (summary) P0→P1: continuity ≥ 80; p95 text ≤ 4s; image ≤ 20s. P1→P2: 500 beta; completion ≥ 35%; severe triage ≤ 4h. P2→P3: 10k MAU; conv ≥ 3.5%; cost/MAU ≤ $0.50. P3→P4: ≥500 monetizing creators; p50 payout ≥ $40; contradictions ↓ 30%. P4→P5: 50k MAU; ARPPU ≥ $12; RPM target met. Reporting cadence Weekly: build velocity, incident log, cost/MAU, queue p95. Bi-weekly: continuity score trend, contradiction incidents, moderation SLA. Monthly: cohort retention, creator payouts, MAU/ARPPU, PR/partnership progress. B. Build a highly secure, smart CCTV solution based on current devices from scratch in EIGHT months, leveraging the benefits of a highly capable but extremely lite Tiny LM, addressing a specific security demand (which can be modularly swapped for another on IoT-Output-Consumer) and operates on exception-based action/alert. Phase 0 — Prototype (Weeks 0–6) Scope Develop a functional proof-of-concept demonstrating the feasibility of the Fractal LM architecture Implement a basic local training and inference loop to validate speed, efficiency, and overfitting resistance. Milestones Establish baseline architecture and environment (data pipeline, model scaffolding). Prototype model trained on 1–2 small datasets demonstrating faster convergence. Prototype benchmark results (vs. baseline models) showing training time and performance improvements. Success Criteria Show consistent model performance without overfitting. Demonstrate faster training than a comparable baseline LM on small data. Achieve internal validation that the architecture can scale toward MVP. Phase 1 — MVP (Weeks 6–18) Scope Develop a deployable MVP integrating the Fractal LM into a controlled IoT-like environment. Implement security and data flow controls aligning with the air-gapped and unified security goals. Support fine-tuning workflows, localized model deployment, and performance monitoring on simple initial hardware profile Milestones Integrate LM prototype into modular inference framework with data isolation and logging Conduct end-to-end MVP tests in simulated IoT deployment environments. Deliver MVP with demonstrable advantages in cost, speed, and control. Success Criteria Demonstrate secure, auditable data flow with no external dependencies. Achieve training and inference cost/time reduction compared to conventional LMs. Phase 2 — Beta / Pilot (Weeks 18–onward) Scope Deploy the LM solution to real-world partner environments or pilot customers. Collect operational metrics and user feedback for continuous improvement. Begin developing commercial integration pathways with paired IoT systems and SDK/tooling for developers. Milestones Live deployments in testbeds or client environments. Continuous optimization of training/inference throughput and compatibility. Expansion of ecosystem partnerships and demonstration use cases. Success Criteria Achieve stable performance in a real environment with measurable cost and latency benefits. Secure at least one commercial integration or pilot commitment. Establish scaling roadmap for production and enterprise-grade security compliance (SOC 2, ISO 27001 alignment) Staffing Plan (minimal viable) 1-2 LLM Eng, Backend Software Dev, Infra/Ops Lead Dependencies & Risks Experimental fractal SLMs could show little or no improvement in target areas during the benchmarking phase. Obtaining quality training data KPIs by Phase (roll‑up) Product: Industry feedback thru Partner network: KeyChain, Sonidel, Fusion, others Growth: Assess partner and Pilot Client feedback for specific applications ie unique IoT-Output-Consumer developed for a specific solution Cost/Risk: Measure efficiency, costs to deliver IoT-Output-Consumer Machine Learning programs, costs to run and maintain

Founded 2024