AI Playbooks · The PMF Guy
Actionable guides on AI productivity, prompts, design, GTM, and growth. Clear processes, useful tools, and lessons from building and scaling products.
Browse by topic
Showing all 19 articles
Anthropic, OpenAI, HubSpot, IBM, GitHub. I checked every cost and enrolment rule against the issuers' own pages, and the most prestigious one on the list is gated behind partner-network access rather than price. Every link included.
ReadThirteen tools, eleven of them free, chained into one pipeline that can research, watch, hear, animate, edit, render, and publish. Every line priced, including the three places where the free tier quietly runs out.
ReadHow creators, content, SEO, founder-led sales, and product-led distribution help AI companies test positioning, solve the cold start, and validate product-market fit faster.
ReadVocals generates an original song sung in your own voice, or one from its library. References and Styles hold the sound together. But the copyright screen that runs before you generate anything is what decides whether you can actually ship the output.
Read2.8 trillion parameters, a 1 million token context window, and first place on a blind developer leaderboard above Claude Fable 5. The weights land July 27. Here is what is actually true, what is being oversold, and which jobs I would actually give it.
ReadSlash ghost. Slash god mode. Type them into Claude Code and nothing happens, because they were never commands. So I wrote all five as real skills: /ghost, /uda, /buddha, /l99, /godmode. Full SKILL.md files inside, and a one-minute install.
Read"Usage limit reached, back in 3 hours." He clocked out mid-task again. The 14 commands that keep him at his desk: /model opusplan, /compact, /effort, /context, /rewind, subagents on Haiku, and more. Every one checked against the docs.
ReadClaude Code starts every session cold. These are the five plugins I install to give it a memory, catch its mistakes, and cut its cost: Claude-Mem, Ponytail, Code Review, Obsidian, and Anthropic's official pack, with install links.
ReadDrop a PDF or screenshot into Claude and every page is billed as an image plus its text, so files cost several times more tokens than pasted text. The math from Anthropic's own docs, and the three-second habit that fixes it.
ReadBuild a reusable ChatGPT Work workspace with a folder structure, three context files, GPT-5.6 reasoning, and only the connectors your workflow needs.
ReadConfigure Claude memory, custom instructions, dictation, and Anthropic Academy in five minutes. Here is where each feature lives and what it changes.
ReadPromptive Sentry transfers a complete AI conversation between ChatGPT, Claude, and Gemini. Here is how it works, what it stores, and the privacy tradeoffs.
ReadUse Skill UI, Impeccable, UI/UX Pro Max, and Playwright CLI to give Claude Code a design system, stronger visual judgment, and a way to inspect its own output.
ReadTurn what Fable 5 learned about your workflow into an audit, reusable skill files, and CLAUDE.md instructions you can carry into cheaper models.
ReadOne instruction before bed. By morning, thepmfguy.com was live in production, the Framer subscription was cancelled, hosting had moved to Cloudflare's free tier, and the full SEO, GEO, and AEO layer was built: robots.txt welcoming fifteen AI crawlers, llms.txt, four schema blocks, and a nine-page sitemap. The build log, with receipts.
ReadEvery launch, every demo, every funding announcement is selling the same thing. A promise that your future will be different from your present. The honest read on why the AI industry runs on belief, why the foundation is still being laid in real time, and what it means if you are building or buying right now.
ReadOne prompt built a full SpaceX marketing site - live Three.js WebGL scene on a Martian horizon, six navigable routes with camera dolly transitions, 3,000 particle dust simulation, procedural Web Audio, and a custom cursor. All as a single self-contained HTML file. Here is the prompt and what made it work.
ReadThree distribution channels that work for AI products, how to sequence them by stage and category, and the trust problem that kills most AI GTM strategies before they generate any signal. Built on the playbook that took Emergent from $0 to $10M ARR in 60 days.
ReadThe exact prompt I used to get Claude to redesign Masonry's homepage from scratch - complete rebrand, WebGL animations, interactive cards, and an aspirational hero section. What design agencies charge $10K for, done in an afternoon with the right prompt.
Read