A Readable Guide to Decisions, Planning, Memory, Tools, Learning, and Cooperation
Companion material for The Mathematics of Artificial Intelligence Agents by Jason Karpeles. Each chapter has a Jupyter notebook that reproduces its calculations and figures, and a skill that an AI assistant can use to teach the chapter.
Interactive version: https://karpeles.com/companions/math-ai-agents/ About the book: https://karpeles.com/publishing/math-ai-agents
notebooks/: one notebook per chapter, already run, so the results show on GitHub; orientation and a document-release capstone are included.skills/: a master skill for the book and one skill per chapter.maa- folders hold the same canonical skills used by the installer.src/, tools/, data/, and tests/: shared calculations, teaching inputs, and checks.content/, assets/, and solutions/: explanations, equations, figures and separate worked answers.Examples identify constructed teaching inputs and source-reported measurements. Their assumptions and limits are stated with each calculation.
You need Python 3.11 or later.
pip install -r requirements.txt
jupyter lab
Open any notebook in notebooks/ and choose Run All. Keep the companion folders together.
For terminal commands, install the local package from the repository root:
pip install -e .
python -m math_ai_agents list
python -m math_ai_agents run --chapter 6 --case example --text
python -m math_ai_agents run --chapter 6 --input my-inputs.json --output report.json
Copy data/examples/ch06.json, or the file for your chapter, to get the input shape, then replace the values. The matching web demonstrations also run from the terminal:
python -m math_ai_agents demo-list --chapter 6
python -m math_ai_agents demo --chapter 6 --id C06-D01 --text
Run the checks with python -m unittest discover -s tests.
Start with skills/math-ai-agents/, the master skill. It routes a question to a chapter method or demonstration. Each chapter skill names its equations, notebook and input contract.
Keep the repository together and use the same Python environment as the notebooks. Run the supplied installer from the repository root, for example:
python tools/install_skills.py --target ~/.claude/skills
For Codex, use --target ~/.codex/skills. The installer selects the 28 canonical skills without duplicating numeric compatibility folders and records their runtime location. Keep the repository at that location. Existing skill folders are preserved. Run python tools/install_skills.py --help for its destination options.
Copyright Jason Karpeles. The code in this repository is released under the MIT License (see LICENSE).