Build your own superpower
The ability to consume and synthesize information at scale into a personal information graph is a superpower. ChatGPT can help. Read: "Everything Starts Out Looking Like a Toy" #177
Hi, I’m Greg 👋! I write weekly product essays, including system “handshakes”, the expectations for workflow, and the jobs to be done for data. What is Data Operations? was the first post in the series.
This week’s toy: an auto-summary of any YouTube URL. This one’s cool because it builds chapter summaries, auto-codes a link and a thumbnail, and is otherwise awesome. Edition 177 of this newsletter is here - it’s December 26, 2023.
If you have a comment or are interested in sponsoring, hit reply.
Data Operations (“Everything Starts Out Looking Like a Toy”) is a reader-supported publication. Please consider subscribing.
The Big Idea
A short long-form essay about data things
⚙️ Build-your-own superpower
Think of the most effective people you know. One of their characteristics is that for a given set of tasks, they are much better than anyone else at following a procedure, reviewing information, and methodically reaching conclusions. Most people do one of these things well but not all of them.
Hidden in the news about GPTs and chatbots becoming more effective is that they are now enabling modular tasks that weren’t possible before. You don’t need GPTs to do an entire task to get value from compressing a workflow from many steps into fewer steps. That means GPTs can help anyone become more productive quickly.
ChatGPT lets you develop individual workflows to have more effective outcomes like those high performers. You don’t need (or want) ChatGPT to think for you. But you do want your workflow to follow a structured process every time you start a complicated task and be additive to your knowledge.
One of these optimizations is the ability to read a link, summarize it, and produce a visual diagram of the ideas in that link. Let’s take a look at a way that ChatGPT can help you get a quick summary from any URL.
“Auto-Summary” as a Service
There are a lot of articles that come across your path every day. How do you read and synthesize the information? More importantly, how do you assess which articles are worth reading?
If you increase your reading velocity and add some amount of comprehension, you might learn more. One way to do that is to build an article summary where ChatGPT “reads” the article and produces a summary in diagram form.
I’ve written about building diagrams with ChatGPT…
and about writing diagrams with code …
But those diagrams are the outcome of creating a process and diagramming it so other people can read it.
Reversing this idea, you can use ChatGPT to read a process and diagram what it means. Kyle Williams and I have been noodling on a ChatGPT to visualize articles. (Disclosure: he’s done all of the hard technical work. I’ve been heckling, qa testing, and trying to break the thing.)
Ask that bot to summarize an article like the one above on documenting code and it produces a concept diagram:
Is this summary perfect? Nope. At the moment it’s a topic or concept map helping you to know the major points of an article in the way you would write index cards when studying for an exam. But it did this for you automatically.
Where would we invest to make this better?
The goal is to build a superpower to summarize and learn about any topic, taking into account what you already know.
If the current iteration of this feature is v0, what would you do to improve it?
Identify Ideas: you’d want to not only summarize but also build a topic idea or idea map. What’s the idea? What does it advance?
Aggregate Knowledge: if this article identified unique knowledge, you’d want to group this into a topic map linking similar articles, especially one stored in your local LLM
Classify Knowledge: decide whether to add this new knowledge to the existing topic map. Is it a good argument and does it reinforce or contradict the existing map?
Answer Questions: it would be neat if you could then ask the topic map to answer questions and have it produce annotated answers based on the references stored in the topic map
If you remember the Card Catalogs that used to exist in libraries, that’s the basic functional idea. Identify a concept, find a related reference, and then assemble a topic map of related ideas.
What’s the takeaway? ChatGPT adds value by increasing the velocity to take in information. But more information without a filter is not particularly useful. Building a bot to help you identify, incorporate, and aggregate new information is an interesting feature.
Links for Reading and Sharing
These are links that caught my 👀
1/ Dashboard stylin’ - Here’s how to make your Google Sheets look more like Dashboards. TL;dr: add whitespace, tint the background to create a border, use consistent styling, and select different fonts than the default.
2/ Procrastination as time travel - When we procrastinate, we delay the work we need to do and often end up in a different place than we intended. Why does this happen?
3/ Make your PRs better - Excellent summary here of a pull request for a dbt project that helps the reviewer understand the context, what’s changing, and the risk/reward of the change.
What to do next
Hit reply if you’ve got links to share, data stories, or want to say hello.
The next big thing always starts out being dismissed as a “toy.” - Chris Dixon