📌 What it is
SameI is Sam's own AI assistant: customized to his workflows, knowledge, and decision-making patterns. Named after Same Solutions, with the same brand DNA: different problems, consistent quality.
Not a replacement for Claude — built ON TOP of Claude (via API), with Sam's context, tools, and integrations baked in.
The CC + planning-chat + substrate workflow Sam is doing right now is the prototype. SameI is what happens when Sam stops being the manual message broker.
🏗️ 3-Tier Architecture
Tier 1 — Foundation $20–50 / mo
Build now, mostly free. Anthropic API + cron + scripts. No fancy infra.
Stack:
- Anthropic API key from console.anthropic.com (~$3–15 per million tokens, $5 starting credit)
- Node.js or Python scripts that call Claude API
- Cron jobs running on Sam's laptop (or $5/mo VPS — Hetzner / DigitalOcean)
- Standardized prompts simulating "Sam's AI"
- Reads/writes to substrate (the file system Sam already has)
Tier 1 capabilities:
- Daily briefing generator — cron each morning, summary to phone
- Bet analysis batch processor — cron each night, generates next-day analyses
- Inbox processor — scans substrate
[INBOX] blocks, processes notes
- Quote generator — Sam provides specs, gets formatted quote draft
Triggers to start Tier 1: Sam hits consumer chat limits regularly · time saved would exceed $50/mo · Sam wants automation that requires programmatic access.
Tier 2 — Real Product $50–200 / mo
Now SameI "knows" Sam's context. Auto-triggers, web UI, semantic search.
Stack additions:
- Vector database (Pinecone, Weaviate, or pgvector on Postgres)
- All Sam's docs/notes indexed semantically
- File watcher → automated processing on changes
- Simple web UI (Next.js, Astro, etc.)
- Webhooks from Cloudflare/GitHub trigger SameI actions
Tier 2 capabilities:
- Ask SameI anything about Sam's work, pulls relevant context automatically
- Auto-generates analyses when new data arrives (e.g., new game scores → re-run backtest)
- Coordinates multiple tools (CC + Bambu export + customer email + manage app)
- AI vs AI: validates Claude's output against other models for honesty checks
Triggers Tier 1 → Tier 2: 3+ Tier 1 tools running successfully for 1+ month · Sam can describe the architecture without me · real product opportunity emerging (friends, customers).
Tier 3 — Full Product $500+ / mo
Multi-model, multi-agent, possibly commercial.
Stack additions:
- Multi-model architecture (Claude + local Llama for sensitive data)
- Custom UI matching Same Solutions branding
- Multiple agents coordinated (analyst, scheduler, writer, etc.)
- Eventually: SameI-as-a-service for other small businesses
Tier 3 capabilities:
- Full agent orchestration — handles complex multi-step workflows
- Privacy-tier routing — sensitive data stays local, general work via API
- Branded UX that Sam (or customers) experiences as a product
- Revenue stream if commercialized
Triggers Tier 2 → Tier 3: Tier 2 demonstrates real value · Sam wants to commercialize · business case justifies $500+/mo.
📅 Honest Timeline
This is a 6–24 month build. Not a weekend project. Each tier is sequenced.
Tier 1: 1–3 months part-time. ~5–10 hrs/week. First tool ships in week 1–2.
Tier 2: 3–6 months part-time. ~10–15 hrs/week. Requires Tier 1 stable.
Tier 3: 12+ months sustained effort. Possibly with co-builder. Requires Tier 2 demonstrating value.
Important: The CC + planning chat work Sam is doing NOW is the prototype for SameI. The substrate is already the file system. The [INBOX] blocks are already the prototype for inbox processing. SameI Tier 1 is just removing Sam from the loop on routine tasks.