One of my current topics is “Output as easy as possible”. While I’m still working towards a full-circle content creation process, I want to share a good example from real life.
It is a common theme for this Substack to provide a sort of backstage access to what I do, and how I do it. Hence I’m providing the messages (prompts) and all that below.
The gist of it all is:
I listened to an article on my way to lunch
I had some ideas, thoughts, and connections forming
I used AI to capture, refine and translate that into output (on the go)
Yesterday, on my way to lunch, I listened to the audio version of AI Can (Mostly) Outperform Human CEOs, which had surfaced via a newsletter (information input).
Listening is a convenient, on-the-go-friendly form of consumption.
Some parts of the article caught my interest, resonated with my views and experience. I also made a connection, wanting to create a short commentary on the piece and share this with a group of executives for whom I recently gave a presentation on a related topic.
After having finished lunch and the article (audio ran about 13 minutes), I was heading back to my desk and office. I realized right then, in the spirit of output as easy as possible that it was now that I had the thoughts and the ideas.
I felt the urge to act on it. Get it out.
At our fingertips
So I used my smartphone, and talked to the AI. Here are my instructions:
Tools
Quick aside in case you’re wondering. I’m interacting with Claude 3.5 Sonnet in this case, using an app called TypingMind, which allows the models to use plugins.
This is why Claude can ‘grab the article’. It has access to a web reader plugin. Super useful, and unfortunately not available via the Claude app, which I would likely have used otherwise. This is what the plugins look like:
Keeping it relevant
Now, remember, I did not want to share a generic summary of the article, but highlight the points that I felt were relevant. So I had to provide some more context. Here‘s my next message to Claude:
Okay let's refine the key takeaways a little bit. I think, and you have to remind me how that was phrased in the article, but I think one of the key points that I want to talk about is that AI is particularly useful in strategic decision making.
Basically high-level type of work, knowledge work is where the technology will be particularly effective as long as there are, as long as there is data and there are metrics and the relationship between how these work and what the sort of like causal relationships are can be well understood.
AI can pick up on these things. It has already outperformed humans in strategic games such as Alpha, as we've seen with AlphaGo or Star, what was it called? Is it called Starlink? What's the game called? You probably can fit that in. And then the other aspect is, and this is where you need to dig up a little bit more from the article, is this hybrid approach where, because that's something, the collaboration with AI is something that I'm particularly focused on.
I love this part, where I could not remember that is was StarCraft, but based on experience, I was pretty sure Claude would get it. I said:
AlphaGo or Star, what was it called? Is it called Starlink? What's the game called? You probably can fit that in.
It did.
Then I also wanted to, excuse me, brag, or be transparent about, how I created the email:
add a postscript to this e-mail. Just to note, I listened to the article in its audio version while I was eating lunch and I dictated the entire idea about summarizing the article and drafting an e-mail to you, to my AI assistant, while I was walking back to my office.
Almost done
Back at the office, I copied and pasted the email draft into my mail client, made some minor adjustments, and recorded a couple of minutes of video commentary. The AI assistant had already prepared the place in the email draft with “[Insert link or attachment to your commentary here]”.
Done.
I love being able to act on inspiration immediately.
The tools are already available.
Now that I have this example use case, I’ll monitor if this comes up again. If so I could improve, simplify, and automate some of the steps. We’ll see.
This was a cool one!