level-1-essentials/individual/02-accounts/README.md

02 — Accounts

Create accounts in this order. Each unlocks capabilities the next one builds on.

Files in this folder

1. accounts-order.md — the priority list with costs and rationale
2. api-keys-setup.md — how to provision keys for each provider
3. password-manager-template.md — how to store credentials

The principle

Consumer subscriptions first (they give you the widest exploration surface). API keys second (they give you programmatic power). Local tools third (they give you speed).

Time budget

• Consumer subscriptions: 30 minutes total
• API keys: 45 minutes total (including billing setup)
• Password manager entries: 15 minutes

Definition of done

• All Tier 1 consumer subscriptions active
• All Tier 1 API keys in your password manager
• Billing caps set on every API key
• One key stored in a .env.local file only (never .env — never commit)

level-1-essentials/individual/02-accounts/accounts-order.md

Accounts — Priority Order

Prices are USD, monthly, as of August 2026. Confirm current pricing before subscribing.

Tier 1 — The core four (get these all)

Tier 2 — Stack-dependent

Tier 3 — API accounts (get all six, cap billing)

Tier 4 — Nice to have

• Together AI, Fireworks, Replicate — additional inference options
• Hugging Face Pro — dataset + model hub
• Kaggle — free notebooks + datasets

Rule of thumb

If you're spending more than $150/mo on subscriptions, one of them is unused. Audit quarterly.

Enterprise upgrades

If you're using any of these for work with sensitive data, upgrade to the Enterprise / Team / Business tier for that provider. The delta buys you: - SSO / SCIM - No training on your data (with contract) - Admin console + audit logs - DPA / BAA availability - Increased context or rate limits

level-1-essentials/individual/02-accounts/api-keys-setup.md

API Keys Setup

Universal rules

1. One key per project. Never share a key across projects; you lose cost attribution.
2. Set a hard spending cap on every key at the provider dashboard.
3. Set a soft email alert at 50% and 80% of cap.
4. Store in a password manager, not in Notes.app or a text file.
5. Never commit a key to git. Use .env.local (git-ignored) or a secret manager.
6. Rotate every 90 days, or immediately after a suspected leak.
7. Prefer service-specific project keys over root keys wherever the provider supports them.

Per-provider setup

OpenAI

1. platform.openai.com → Settings → Billing → add card, set hard limit ($20 to start)
2. API Keys → Create new secret key → name it after the project
3. Enable usage tracking → set alert at 50% and 80%
4. Turn on "Usage limits" per project if you use multiple projects

Anthropic

1. console.anthropic.com → Plans & Billing → add card, set monthly budget
2. API Keys → Create key → tag it
3. Enable Workspace-scoped keys if you want isolation between projects
4. Note: Anthropic bills per model; Opus is expensive, Sonnet is the workhorse, Haiku is cheap

Google AI Studio

1. aistudio.google.com → Get API key → tied to a Google Cloud project
2. In Google Cloud Console → Billing → set budget alert
3. Enable Generative Language API on that project
4. Note: Gemini has a free tier — use it for experimentation

Perplexity

1. perplexity.ai/settings/api → Generate key
2. Purchase credits (pay-as-you-go)
3. Note: sonar / sonar-pro are search-grounded (unique to Perplexity)

Groq

1. console.groq.com → API Keys → Create
2. Free tier is generous; upgrade to pay-as-you-go for higher rate limits
3. Use for: fastest inference of open-weight models (LLaMA 3.x, Mixtral, Qwen)

OpenRouter

1. openrouter.ai → Keys → Create
2. Add credits (one-time)
3. This becomes your "try any model" playground — one key routes to 300+ models
4. Great for experimentation and fallback

.env.local template

# AI provider keys — never commit this file
OPENAI_API_KEY=sk-proj-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=AIza...
PERPLEXITY_API_KEY=pplx-...
GROQ_API_KEY=gsk_...
OPENROUTER_API_KEY=sk-or-v1-...

# Optional
COHERE_API_KEY=
TOGETHER_API_KEY=
FIREWORKS_API_KEY=
REPLICATE_API_TOKEN=
HUGGINGFACE_TOKEN=

Add .env.local to your global .gitignore:

echo ".env.local" >> ~/.gitignore_global
git config --global core.excludesfile ~/.gitignore_global

Spending sanity floor

Start with these caps and adjust up as you learn your consumption:

Total starter API budget: $90/mo. Most individuals never exceed this while learning.

level-1-essentials/individual/02-accounts/password-manager-template.md

Password Manager Template

Create one entry per provider. Recommended tool: 1Password, Bitwarden, or Dashlane.

Entry template

Title: [Provider] — [Purpose]  (e.g., "OpenAI — Personal Dev")

Username: [account email]
Password: [account password]

Custom fields:
  api_key:          sk-...
  api_key_created:  2026-08-19
  api_key_rotates:  2026-11-17
  spending_cap:     $20/mo
  billing_alert:    50% and 80%
  scope:            [personal | project-x | client-y]
  provider_url:     https://platform.openai.com
  key_management:   https://platform.openai.com/api-keys

Notes:
  - Data class allowed: [public / internal / confidential]
  - Zero-retention: [yes / no / see DPA]
  - Notes on quirks

Rotation calendar

Add these to your calendar as recurring events:

• Every 90 days: "Rotate API keys" — walk through password manager, regenerate each
• Every 30 days: "Review AI spending" — check each provider dashboard
• Every 7 days: "Check billing alerts" — during Friday weekly review

Sharing model

• Never share a personal API key
• For a company, use the LLM gateway pattern (see company/02-accounts-procurement/llm-gateway-selection.md) — no one gets raw keys
• For a family, each person gets their own key on the same billing account (where the provider supports this)