What I use myself when building AI tools by hand: checklists, prompts, templates. Leave your email - unlocks access to all materials, and I'll send new ones as they come out.
Move from "chat in the browser" mode to "AI as a working tool" mode in one evening. Three mental shifts, a 15-minute install, five moves that are enough for a beginner, and where to start your first project. No code, no jargon.
The same path from "chat" to "working tool," for people who already have ChatGPT. A 15-minute Codex setup, agent permissions, AGENTS.md instead of custom instructions, five beginner moves, and an honest answer to "Codex or Claude Code?" No code, no jargon.
A coverage matrix: around 50 business-analysis techniques from Wiegers and BABOK v3, each one scored honestly on how much an AI assistant covers and what's left to the human. Five levels from "Full" to "None," with a clear scale. Bottom line: roughly 58% of a BA's formal tasks are already meaningfully automated, and only about 10% stay purely human.
All the theory from AI Sandbox's first session in one place: chat vs. working tool, context engineering, tokens and the context window, skills and agents, evals. Plain language, with diagrams. The fastest way to understand how LLMs and agents actually work.
A ready-made prompt: runs any live page - landing page, pricing, a form - past 5 fictional but realistic users from your audience. Each one reads top to bottom and says what caught their eye, what was confusing, and where they'd close the tab. Output: a verdict by section and a top-5 list of copy fixes. Calibration before you show it to real people.
Build yourself a private agent in Claude Code: matches you against job postings, sharpens your resume, decodes what recruiters are actually saying, and drills you hard for interviews. One prompt, plus a step-by-step guide for non-coders. Everything stays local, nothing leaves your laptop.
A ready-made prompt for Claude: feed in job postings, get back a structured breakdown of requirements, hard and soft skills, AI-depth scoring (0-4), and market gaps. The exact one I run myself.
A step-by-step guide: how to validate demand before writing a line of code. Hypothesis, the riskiest assumption (RAT), problem interviews, a real demand test, and a pre-launch checklist.
My product-discovery engine, packaged as a Claude Code plugin. 19 skills, one orchestrator - and an idea rides through 14 stages: problem, research, segments, AJTBD, synthetic interviews, PRD, red-team, UX prototype, pricing, GTM, messaging, growth. Installs with one command, drives the pipeline itself, and learns from run to run.
Turn a meeting transcript or vague request into a testable spec, surface hidden decisions before development, and validate the requirements with a clickable prototype. Includes the end-to-end workflow, three ready-to-paste prompts, and a Human Check.
Four ready-to-paste prompts from the AI for Managers evening at Imaguru: a meeting transcript to decisions and follow-ups, team updates to a stakeholder status report, and a rough idea to a spec and a clickable prototype. Each comes with a human-check list.
Five ready-to-paste prompts from the Imaguru AI Product Sprint workshop: evidence-based discovery, testing your riskiest assumption, a PRD-lite brief, building the core value loop, and publishing a demo. Copy, paste, fill in the [BRACKETS].
Prefer Telegram? I post this there too - @productiz.