Day 1: The Magic Prompt Formula, a Toolkit Tour, Thinker-and-Doer Data Analysis, the Blind-Men-and-the-Elephant Research Method, and a Living Pitch Deck

Vaibhav's one technique 'applicable 99% of the time': state a role, the task, instructions and context (audience, tone), and the data to use (facts, quotes). Extended live in Claude with extended thinking and web search: a cold email to Uber's CEO that first found his LinkedIn posts and recent positive news - the model plans, calls tools, then writes. The formula recurs in every later demo: CRO prompts, resume rewrites, mock interviews, the data-analysis consultant prompt.

Vaibhav's argument from a Microsoft jobs list and a WEF future-of-work report, with Altman's one-person billion-dollar company as the endpoint: 10-20 years of experience - become the AI strategist who runs generalists; 3-9 - be the hands-on generalist; 1-3 - early adoption is the unfair advantage. The curriculum behind it: Level 1 Foundations, 2 Workflows, 3 Creator's Playground (image, video, audio), 4 Agent Builder, 5 Vibe Coding. The same roadmap sells the paid accelerator later in the day.

Productivity: Wispr Flow dictation with snippets; Gmail's built-in Gemini for thread summaries and replies. Learning: the Emily Chrome extension to chat with a YouTube video or article; Fireflies to attend meetings for you. OCR: screenshot anything - a dashboard, a list of links in Zoom chat - and ask ChatGPT to explain or extract it. Distribution: Supergrow turns videos into LinkedIn posts, auto-comments on target profiles, scores and schedules. Research: Perplexity, and Perplexity Labs building a live Kanban of AI news; Happenstance searching your Gmail/LinkedIn/Twitter contacts in plain English. Learning again: NotebookLM's audio overview you can join live, mind maps, flashcards, video overview in twenty languages. Sheets: Numerous.ai generates formulas and runs =NAI('classify sentiment', A2) per row. Suno writes a song from one line.

PROCEDURE: save the LinkedIn profile as PDF; have the model summarise it for context; paste the job description and rewrite the resume for it; switch to voice mode and demand a hard mock interview in a named persona, with follow-ups until you fail; run the interviewer's LinkedIn through the Crystal Knows extension for a personality profile and communication dos and don'ts. Generalised on air to VC pitches and sales calls.

PROCEDURE on Outskill's own landing page (6% -> 14% target): (1) GoFullPage screenshot, because models cannot browse a link but can read an image; (2) ChatGPT in thinking mode as a 20-year CRO expert, section-by-section critique and rewritten copy; (3) strip the chat and paste only the analysis into Emergent with a visual-designer brief (black and neon green, glassmorphism); (4) answer Emergent's questions (CTA behaviour, mock data) and get a deployable coded page. PK notes Emergent runs a design / dev / test multi-agent team on Claude underneath.

Dileep's Swiss-army-knife vs chef's-knife rule: horizontal tools (ChatGPT, Claude, Gemini) frame and synthesise; vertical tools (Julius, Chronicle, Emergent) execute. PROCEDURE: (1) upload the dataset with a long senior-strategy-consultant prompt to a reasoning model and ask for one governing question plus 3-4 supporting questions (growth, efficiency, risk, advantage), each tied to an action - the McKinsey pyramid principle; (2) take each question to Julius AI (~15 free messages) for analysis, insight, a reallocation plan and an interactive dashboard. Repeated on a second unrelated dataset to show it is data-agnostic.

PROCEDURE: send an identical deep-research prompt to ChatGPT, Perplexity (70+ sources), Gemini (editable research plan), Claude, Genspark and Manus (384 sources), then paste each output in turn into one ChatGPT thread asking it to incorporate the new insights into an ever-improving master report. Demoed as a LatAm EdTech market report and an eight-slide investor deck in 20-30 minutes; a Vinod Khosla quote pulled from a video via Emily; the deck rendered in Chronicle. Then the living pitch deck: import to Google Vids, regenerate images, a Veo-generated intro scene, transitions, ChatGPT speaker notes, an avatar narrator, render.

Identify a specific friction point; find a vertical tool that removes exactly that; chain tools so one's output is the next's input; validate on real use before trusting it; start with one workflow, master it, and add a tool only when it demonstrably improves the result. Illustrated by the Grab It + WebSync pair that bulk-imports YouTube links into NotebookLM, and by the Genspark AI Sheets / AI Docs job-search pipeline. Toolkit taxonomy: thinkers, doers, integrators, enhancers.