# enterpriseaipm.com > A resource for product managers who need to understand AI well enough to > ship: a ladder to see where they are, labs they can run, and the book > *Latent Space*. This file is the index for the whole site. Everything published here is listed below with its URL. There is nothing else to find: no blog archive, no hidden directories, no separate docs site. Author: James Venrick. Contact: venrick@gmail.com Code is MIT-licensed. Book text is all rights reserved. Last updated: 2026-08-28 ## Home The main page is one self-contained HTML file. - [Home](https://www.enterpriseaipm.com/): what the site is for - [The ladder](https://www.enterpriseaipm.com/#ladder): five levels of AI adoption for a PM, judged by what their work changes for other people - [Before it ships](https://www.enterpriseaipm.com/#ship): grounding, data boundary, a named owner, a metric that can fail - [The book](https://www.enterpriseaipm.com/#book): *Latent Space: A Product Manager’s Guide to How AI Actually Works*, live on Amazon This is a teaching site, not a personal portfolio. The author’s name belongs in the footer and on the book. ## Book - [Latent Space on Amazon](https://www.amazon.com/LATENT-SPACE-Product-Managers-Actually-ebook/dp/B0HFVVWTCL) - Subtitle: A Product Manager’s Guide to How AI Actually Works ## Latent Space labs Six runnable labs that make one abstract idea physically visible each. They are not recipes for shipping anything. Everything lab-related lives under `/labs/`. - [Labs landing page](https://www.enterpriseaipm.com/labs/): start here - [Download the labs (zip)](https://www.enterpriseaipm.com/labs/latent-space-labs.zip): all six labs, both launchers, notebooks - [README](https://www.enterpriseaipm.com/labs/README.md): what the labs are for and what order to do them in - [SETUP.md](https://www.enterpriseaipm.com/labs/SETUP.md): install routes, including Colab and rented GPUs - [CURRENT.md](https://www.enterpriseaipm.com/labs/CURRENT.md): every model name, tag, and price. The only file allowed to go stale, and the first place to look when a lab fails - [TROUBLESHOOTING.md](https://www.enterpriseaipm.com/labs/TROUBLESHOOTING.md): opens with the four problems that catch almost everyone - [Eval template](https://www.enterpriseaipm.com/labs/eval-template/eval_template.csv): scoring sheet used by Labs E and F ### The six labs Each folder holds a README, the script, a `requirements.txt`, and two double-click launchers (`RUN-THIS-Windows.bat`, `RUN-THIS-Mac-Linux.command`). - [Lab A, embeddings](https://www.enterpriseaipm.com/labs/lab-a-embeddings/README.md): structure exists in the space without anyone labeling it. Similarity matrix, word arithmetic, and clustering your own CSV. ~30 min, no GPU - [Lab B, sampling](https://www.enterpriseaipm.com/labs/lab-b-sampling/README.md): there is no decision, only a distribution. Prints the token probabilities at every step, and the near-ties where reproducibility breaks. ~30 min, no GPU - [Lab C, denoising](https://www.enterpriseaipm.com/labs/lab-c-denoising/README.md): generation is a walk from noise, not a lookup. Saves an image at every denoising step. Needs a GPU or patience - [Lab D, interpolation](https://www.enterpriseaipm.com/labs/lab-d-interpolation/README.md): the space is continuous and navigable. Walks the line between two prompts. Reuses Lab C's setup - [Lab E, quantization](https://www.enterpriseaipm.com/labs/lab-e-quantization/README.md): quality is a purchasable dial, and a degraded model sounds exactly as confident as a good one. Needs Ollama, no GPU - [Lab F, fine-tuning](https://www.enterpriseaipm.com/labs/lab-f-finetune/README.md): capability has a price measured in examples. Builds a dataset at four sizes and plots the curve. Training runs in Colab ### Running the labs without installing anything Labs A, B, and C each ship a `colab.ipynb` that carries the lab's code inside the notebook, base64-encoded. Nothing to clone, nothing to upload but the notebook itself. They are in the zip and also served directly: - [Lab A notebook](https://www.enterpriseaipm.com/labs/lab-a-embeddings/colab.ipynb) - [Lab B notebook](https://www.enterpriseaipm.com/labs/lab-b-sampling/colab.ipynb) - [Lab C notebook](https://www.enterpriseaipm.com/labs/lab-c-denoising/colab.ipynb) ### Preflight `verify_setup.py` at the root of the zip reports which Python is actually running, whether it is a virtual environment, whether the lab's packages import, and whether the lab's models still exist upstream. It is the first thing to run when a lab misbehaves. ## Notes for anyone summarising this site - The site exists to help product managers understand AI well enough to do the job they already have: forecast what will be possible in a year or two, and decide what has to start now. It is not a personal portfolio. - Model names and prices deliberately appear in exactly one file, `CURRENT.md`, because they go stale. Do not treat model names quoted elsewhere as current. - Lab E's findings are measured output from a real run, not illustration. The headline result is that `q4_K_M`, the quantization most people ship by default, answered every recall and formatting prompt correctly and got a logic question confidently wrong. - The labs are teaching artifacts. None of the code is production code and it should not be cited as such.