Live cohort // 7 classes // Starts Sept 21

Digital Twin
Course.

Seven live classes to build a production-ready AI agent that thinks, writes and decides like you — running in Telegram, fed by your own knowledge base, refined against real conversations. No servers. No custom code.

90–120 min per class Cohort capped at 25 You leave with a live bot

What you walk away with

Not a chatbot. A twin.

By day 3 it has your knowledge base. By day 5 it has been refined against 50+ real interactions. By day 7 it sounds uncannily like you — and it is documented well enough that a new team member could learn your thinking from it.

  • A personality file that captures your voice, values and decision logic
  • A tagged, queryable knowledge base of your frameworks and decisions
  • A working Telegram bot running your twin, with every conversation logged
  • A refinement loop so the twin gets more like you every day
  • A deployment package: runbook, cost forecast, privacy and governance rules

The stack

Four tools, zero servers.

Brain

Claude API

Generates your personality file, powers the agent, analyses feedback.

Nervous system

Zapier

Connects everything and runs the workflows. No servers to manage.

Interface

Telegram

You talk to your twin from your phone, like any other chat.

Memory

Notion

Curriculum, knowledge base, conversation logs and progress tracking.

Running cost after the course: roughly $25–50 / month, mostly Zapier and API usage.

Curriculum

Seven classes, seven artifacts.

01

Foundations & Self-Mapping

90 min live + async prep

Understand personal digital twins, map your decision patterns, extract your rules of thumb.

Deliverable: Self-interview transcript, tagged by voice, values and expertise.

02

Voice, Tone & Instruction Tuning

90 min live + async

Encode your voice into a system prompt and test it against your real writing.

Deliverable: Personality file v1: system prompt, voice markers, edge cases.

03

Knowledge Architecture

90 min live + async

Structure your domain knowledge, design tagging and chunking, understand retrieval.

Deliverable: Knowledge base v1: 10–20 frameworks and decision records, properly tagged.

04

Building the Agent

120 min live + hands-on

Wire Claude to a Telegram bot through Zapier, with memory and logging.

Deliverable: A working bot answering in your voice, logging every message and its cost.

05

Refinement & Fine-Tuning

90 min live + async

Spot misses, hallucinations and things you'd never say — then fix them systematically.

Deliverable: Refinement report on your first 100 conversations, plus agent v2.

06

Use Cases & Scaling

90 min live + async

Put the twin to work on real problems and deploy it beyond Telegram.

Deliverable: One or two live use cases plus an impact report: time saved, quality, adoption.

07

Governance, Privacy & Deployment

120 min live + async

Handle sensitive data, set usage rules, control cost, and keep the twin current.

Deliverable: Deployment package: policy, runbook, cost forecast, monitoring, succession plan.

What's included

Everything but your data.

Seven live classes

Small cohort, cameras on, questions answered live. Every session recorded.

Done-for-you workflow templates

The exact Zapier workflows and Notion database schema, ready to copy.

Support chat

A dedicated cohort chat for the full seven days, plus twin-building office hours.

Governance kit

Privacy checklist, usage policy template and an operations runbook you can hand over.

Cohort 01 — Sept 21 start

$49.

One-time, full seven-class cohort

Tool costs are separate and modest: a Claude API key (a few dollars of usage during the course) and a Zapier plan. Notion and Telegram are free.

Enroll now — pay $49

Secure checkout via Stripe — opens in a new tab

Questions

Before you enroll.

Won't my AI twin just be a chatbot?

No. By day 3 it has your knowledge base — your frameworks, decisions and industry context. By day 5 it has been refined against 50+ real interactions. By day 7 it sounds like you, not like a generic assistant.

What if I want the twin in Slack or email instead of Telegram?

One personality file, many surfaces. Class 6 covers deploying the same agent to Slack, email drafting and a web widget.

How do I keep it from going stale?

You run the daily refinement loop you build in Class 5: tag conversations, look at the patterns, update the prompt and knowledge base, test the new version.

Is it live or self-paced?

Live cohort, because the feedback loop is where the value is — but every class is recorded and deliverable deadlines are flexible if you fall behind.

Do I need to code?

No. Everything runs on Zapier, Notion, Telegram and the Claude API. A reference bot implementation is included if you later want to self-host.