Timing isn't an accident: Anthropic started metering programmatic Claude usage on June 15. So the first thing I built wasn't "buy more credits" — it was the honest map of what's actually free or claimable after the change: the real free tiers, the few signup credits anyone can get, and why the "$1,000 free" deals on perks portals are mostly gated ceilings (the ungated OpenRouter credit is $1, not $1,000; Anthropic's is ~$5). No affiliate links, eligibility stated plainly, every number checked against the provider's own page → https://autoresearchplaybook.com... The punchline is boring and true: for anything you run on a loop — prompt optimization, SEO, copy, debugging — a local model (Ollama/MLX) is $0. No signup, no expiry. (ML *training* still needs a GPU — I'm explicit about that.) The Playbook is just the 12 program.md templates that make that loop run safely, with a hard max_cost_usd stop so it can never bill you by accident. Start free with the 12-point assessment: https://autoresearchplaybook.com... PH launch bonus (this week only): take the free assessment and I'll auto-send you 2 ready-to-run program.md templates — the Zero-API-Cost + Hybrid prompt-optimizers, both $0 to run locally on Ollama — straight to your inbox. No purchase. Full Playbook is $97, 14-day refund if it doesn't fit: https://justmattg.gumroad.com/l/... Not affiliated with Karpathy — this builds on his open-source method. Happy to get into the cost math.

Autoresearch Playbook

Let AI test your marketing and keep only the winners

Let AI test your marketing and keep only the winners