AGI is here. Are you there yet?
Superintelligence is already here on the model side. What stands between it and you is a prompt. AI Guru is a workshop, in person, for startups, operators and creatives: one guide, your own laptop, your own work. One day per level, from prompted to promptless. You leave with it running. One workshop, the certification, examines you on it.
Pick a workshop. Tell us when and where.
Each workshop runs inside a window, at a few tentative areas. You ask for the day, the slot and the area that suit you. When enough people ask for the same one, a batch forms and everyone on it gets the date. Some are free. Some seats are sponsored. Some run inside your office. One of them is the certification: two days, an exam, and a numbered AI Guru certificate.
Free seats, your name on them
A company, a college or a fund pays for a batch. The seats go free to creatives, students or founders you pick or we pick. Your name on the certificate line and on the room. Ask about sponsoring →
Your team, your space, your work
We bring the workshop to your office for a team of six to twenty. Any level, or the two-day ladder. Everyone who clears the level is certified. Ask about in-office →
Certified for what you can run, not what you sat through.
The certification is its own workshop: two days, eight seats. Day one you climb the whole ladder on your own laptop. Day two is the exam: your harness against plain Claude Code on the same brief, on the clock, in front of the room. Clear it and you get a numbered AI Guru certificate with what you built written on it. The other workshops get you ready for it. See the certification workshop →
Prompted
You can brief a model the way you’d brief a good freelancer, and you’ve done it on one real job.
EXAMINED ON: YOUR PROMPT SET, RUN LIVE
Primed
Your own AI, tools and context are connected once, and real apps run on them. You can set this up for someone else.
EXAMINED ON: YOUR SETUP, ONE PROJECT THROUGH THREE APPS
Promptless
You built the thing that does the job without being asked, and it passed the parity run against plain Claude Code.
EXAMINED ON: ONE HARNESS, SHIPPED, THE PARITY RUN
It writes the code. Drafts the contract. Reads the scan. Argues the case. On the benchmark, it is already superhuman.
The felt experience of using it has barely moved: you sit at a blank box and explain yourself, again, from scratch. The intelligence is extraordinary. The interface to it is a text field.
Between how capable the model is and how little of that capability reaches you on an ordinary day. It is not a modelling problem. It is a setup problem. You can close it in a day, with someone beside you.
It’s here for everyone. Are you using it like it is?
Launching something
Still hiring for what the model already does? A brand to build, ads to make, a site to fix, a deck to send. We coach you through all of it on your own AI: the brief in Brandbrain, the ads in AdForge, the site in Redline.
YOU WALK OUT WITH: THE BRAND BRIEF, THE FIRST ADS, THE SITE REVIEWED
Running the thing
Still the person the inbox lands on? Ops, chief of staff, the one holding it all in their head. We coach you into a workspace where the model already holds your context, and a routine that drafts tomorrow’s moves overnight and waits for your yes.
YOU WALK OUT WITH: A CONTEXT VAULT, ONE ROUTINE ON THE CLOCK
Making the thing
Still doing the busywork around the work? Designers, editors, writers: keep the taste, lose the rest. Logo directions in Crest, cuts in Cut, dictation in Flow on a local model that never leaves your Mac.
YOU WALK OUT WITH: YOUR TOOLS ON YOUR AI, YOUR TASTE FILE
Getting there is a day, not a degree.
You bring the job
A brand, a client, a weekly report, a support queue. The day is built on it. No sample projects.
We sit next to you
A small group, or just you. When yours doesn’t work, we fix it with you at your keyboard, not from the front.
You leave with it running
Every level ends with something working on your own laptop: a prompt set, a workspace, a harness. Not notes.
We stay reachable
A group with your coach for the weeks after. When it breaks on a Tuesday, you ask, and we answer.
“My feed is full of people building with AI. Most people I know just want an easier way to get something done.”
Between those two groups sits a setup problem: install this, connect that, bring an API key, explain your project again. Coaching is how we close it for one person at a time.
“Superintelligence is here. It just doesn’t feel like it yet.”
Superintelligence lands the moment it becomes intuitive to use, not the moment it clears a benchmark. Closing that gap is the whole job. For you, it’s a day.
Where are you? Start there.
Prompted, primed or promptless. Each level is one coached day. Pick the one that matches where you are and we tell you honestly if it’s the wrong one. The board shows how much of your setup you’ll be running by the end of it.
Something running. And someone to call.
Your prompts, written once
The five lines for your work, reviewed, kept where you work.
Switchboard on your Mac
Your own AI, your own context, your own data. The context resident. Real apps running on it.
One harness you built
Tools, instructions, context and a process for one real job. Running on your own setup.
Your coach, after
A group with your coach for the weeks after. Questions answered, breakages fixed, next level when you’re ready.
A room in Mumbai, or your studio.
A small group, one coach, one level
- 10:00 – 17:30, one room, lunch in. Founders, operators, creatives in the same room
- Level 1: any laptop. Levels 2–3: macOS 13+, Apple Silicon, your Claude Code or Codex sign-in
- Bring what you’re working on. The day runs on it
- Levels run in order; take one or all three
Your studio, your work, your coach for the day
- Just you, or you and your co-founders, or the whole early team
- Any level, or a two-day ladder on the one thing you’re launching
- Your real brand, real files, real weekly job
- Mumbai and Pune in person. Elsewhere, ask
Claude Code and CodexToday, that includes Claude Code and Codex. Levels 2 and 3 run on the access your team already has.Before you book.
Do we need API keys or a new subscription?
No. Level 1 uses whatever your team already uses. Levels 2 and 3 run on the Claude Code or Codex access your team already has. Your provider’s usage limits still apply. Local models cover voice and some text tasks with no account at all.
I’m not technical. Is this for me?
Yes. Most people in the room aren’t. Level 1 and Level 2 need no code at all. Level 3 is for anyone who wants to build the thing that runs their job; we write the seventy lines together.
Can we skip a level?
Yes. A team that prompts well every day starts at Level 2. A team with Claude Code already in the workflow starts at Level 3. Tell us where you are when you book.
Does our data leave our machines?
Switchboard runs locally. Context and data stay in folders you own. When an app calls a cloud model, that request goes to the provider you chose, and each app has to be granted access first. You watch the calls during the consent block.
Windows or Linux?
Not yet. Switchboard is a Mac app, macOS 13 or later on Apple Silicon. For on-site sessions we can pair a few Macs for a mixed team; tell us the mix when you book.
What should we bring?
A laptop, a sign-in for Claude Code or Codex for Levels 2 and 3, and one real project: a brand, a product, a client, a weekly job.
Is the AI Guru certificate real?
It is a numbered certificate from The Last Prompt, issued by your guide at the end of the certification workshop, with what you built written on it. It is not a university credit. It says one specific thing: this person can run this, and here is what they ran.
Who coaches?
Sameep, and the people building Switchboard at The Last Prompt. We run our own company on these tools every day. You get the same people beside you, not a trainer reading slides.
How many people in the room?
Small. An open day is a handful of founders, operators and creatives, one coach. 1:1 is you and the coach.
What happens after the day?
You keep a group with your coach for the weeks after. Ask when something breaks, send what you built, and tell us when you’re ready for the next level.
Not there yet? Say where you are.
What you’re building or running, how you use AI today, how many of you. We write back within a day with the level we’d start you on, dates, and the price.
Mail sameep@stayoften.com with your team size and city. Same reply, same day.
Download Switchboard from the latest release and sign in to Claude Code or Codex. Both take a few minutes. Level 1 needs nothing but a laptop and one weekly task you want done better.