Use-case examples

The best place to explore what Memex can do is Memex Universe, a public directory where users share their best ideas and conversations.

Below we highlight some detailed walkthroughts to inspire for your agentic hacking. To share your own experiences, get help / ideas, or discuss industry trends, join our discord!

1. Creating a Generative AI data labeling app

Generative AI applications produce a lot of unstructured and multistructured data. As part of building Memex, we found ourselves spending a lot of time combing and labeling our conversations. There are a lot of SaaS tools available for data labeling, but none that met our requirements.

So we thought, what if we built our own with Memex! It ended up taking less than ten minutes.

Below is a demo of how we did it. We think its a good example of how you can use Memex to build and deploy apps for your different use-cases.

2. Web research data collection and analysis

Humans are curious beings. Memex can help you research new topics faster and more efficiently than ever, however you like to learn, whether via visualizations, summaries, or even by curating and accessing primary sources.

The demo below shows how we used Memex to autonomously compile and visualize data from the internet, in this case about the Space Economy.

We've used Memex to research energy generation, rare earth materials, monetary policy, Mesoamerican civilizations, among many others. What are you interested in researching?

3. Hardware design: creating 3D models

Digital 3D surface and solid modeling is usually thought to require advanced knowledge and access to expensive and difficult to use tools. There are, however, a growing number of powerful code libraries that can be used for this.

The demo below shows how Memex can leverage these libraries to design and create 3D objects without the need to learn how to use them, how to install them, or which dependencies it requires.

4. Accurate LLM research with inline citations

There are some use-case where access to trusted primary sources is an absolute must. The demo below shows how we used Memex to find and curate the best and most relevant sources to help deal with a medical situation.

5. Train a transformer in a jupyter notebook locally

In this example, Memex ...

• created a poetry environment

• installed jupyter

• created a hello, world notebook and tested it worked

• initialized git with a .gitignore and committed the work

• created a simple transformer training notebook, resolved the dependencies, then verified it worked properly

The above GIF was made after those steps were completed

6. New feature for a web project

A friend has an OSS project to track the lineage of foundation models: Unified Model Record. They wanted to add a dynamic visualization of the lineage of each foundation model, such as:

• What data was it trained on?

• What models synthesized data that it was trained on?

• What models was it distilled from?)

Memex was able to implement it (see screenshot).

It got stuck a few times and I had to help it out. But it was also able to implement the feature end-to-end without learning either of the frameworks the project uses (graphviz and pelican). And I did everything through Memex - no terminal / IDE / etc. the pull request here shows everything Memex did.

Caveats: Implementing this with Memex cost a whopping $65 with gpt-4o 🤯. And there were plenty of things it struggled with, including:

• Installing conda

• Editing files containing multi-line strings

We put our heads together and we think we have a path to >10x cost reduction that we’ve got in our queue.

7. Stock analysis app

One of the first test users created this stock analysis app in streamlit:

“Memex created this stock analysis app pretty quickly (just for fun...). Impressive! the only thing that didn't work was that i had to figure out that the streamlit set up was not complete (it was missing an email or skipping that step) so had to fix that on my own”