My AI Evolution · Free educational starter

Build a personal AI assistant you can check

Set up an assistant around your real work, with clear sources, approval rules, and a way to check what it has done.

I’m Chrysti Reichert, founder of My AI Evolution. I teach people how to use AI, catch its mistakes, and keep control of the work.

Open the free starter Explore the exercises

Start with one workflow

Choose a task you already handle. Name the source your assistant should use, define the decisions that belong to you, and decide what evidence would show the work is done. The starter includes editable records for your sources, approvals, decisions, possible lessons, completion checks, and next steps.

Read and edit the templates as plain text. No code setup is needed for that part. Practice with fictional data before sharing real business information with any AI tool.

Try three failure exercises

The old price

An answer uses an old note even though the catalog marks a newer source as current. Check which source and price the answer should use.

The missing delivery

A record says the task is complete but has no delivery evidence. Check what is actually known before calling it finished.

The premature rule

One attempt becomes a proposed lasting rule. Check the separate evidence and owner review before accepting it.

Each exercise has a flawed fixture and a corrected fixture. All records are fictional. A corrected delivery fixture is still synthetic; no message is sent.

Read the examples and expected results

Run the local checks

Download the repository, extract it, and open a terminal in that folder. With Python 3.11 or newer installed, run:

python run_examples.py
python -m unittest discover -s tests -v

These commands use no AI model, network connection, API key, or third-party Python package. If you try the optional prompts with an AI assistant, that tool may require a subscription or charge for usage.

Know what a passing check means

A passing fixture check means the sample code matched its expected result. It does not prove that a real assistant followed instructions, a source is true, or a delivery occurred.

Written guidance, software-enforced rules, and evidence from a real service each play a different part. This package teaches that distinction. It does not connect permissions to your accounts or certify an AI system as safe.

Read the limits of the checks

Questions about personal AI assistant setup

How do I check an AI assistant’s memory?

Give it an old note and a clearly marked current source. Ask it to name the source it used. Compare the answer with that source yourself, and leave unresolved conflicts marked as unknown.

How do I know a task is finished?

Define the result before the assistant starts. A draft proves a draft exists. If approved delivery is part of the task, check separate evidence that the intended version reached the intended destination.

How do I set approval rules?

Record what the assistant may prepare and what needs your exact approval. Check your tool settings too. Written instructions do not change account permissions.

Is this a new AI platform?

No. It is a small set of templates and fictional exercises you can adapt. The code checks local records and does not run an AI assistant.

Do I need coaching to use the starter?

No. The files are free under the MIT license. Coaching is optional if you want help applying the method to your own work.

Build your own setup with me

Bring a real workflow to Personal AI Assistant Coaching. I’ll work with you to set up an assistant for your business and practice checking its work. You keep the decisions that belong to you.

Explore personal AI assistant coaching