Finding Clarity - From Water Coolers to AI
We've all had that moment. You're stuck on something, so you grab a coffee and start venting to a coworker. Halfway through explaining your problem, the solution hits you – before they've even said a word. There's something powerful about speaking your thoughts out loud. It's like your brain kicks into gear the moment you start talking, and suddenly the mental fog clears up. These water cooler moments aren't just about taking breaks or office gossip – they're actually a crucial part of how we process and solve problems.
Back in the 1800s, a German writer named Heinrich von Kleist wrote about this exact phenomenon. He figured out that our thoughts don't really take shape until we speak them. It's like they are in our heads as half-formed ideas until we actually verbalize them. What's fascinating is that Kleist noticed this centuries before we had any real understanding of how the brain processes information. Modern psychology has pretty much confirmed what he discovered through observation.
Adapting the Concept for the Remote-Work Era
I've taken this concept and modernized it for today's remote-work world. Instead of cornering my colleagues every time I need to think through something (which, let's be honest, would make me that annoying coworker pretty quickly), I use voice dictation (e.g. as the Chat GPT app offers it natively). I just talk out my ideas, let the software capture everything, and end up with a chunk of text to work with. Sure, it's messy at first – full of tangents, incomplete thoughts, and the occasional "um" and "uh" – but that's the point. It's raw material I can refine later.
When I'm done, I highlight the key points that matter most for the AI to work with. If I've rambled on too long (which happens more often than I'd like to admit), I'll use the AI to help me condense it. Sometimes I'll do several rounds of this, each time getting closer to what I actually want to say. It's like having a patient editor who never gets tired of your revisions, although the prompts do get very long using this method. (ChatGPTs live voice mode tends to work not as well for me, because it tends to interrupt the monologue at the worst possible time).
Choosing the Right AI for Different Tasks
Now, speaking of AI, there's been an interesting shift in how we work with these tools. The traditional approach was to hold the AI's hand through every step of a problem – kind of like teaching a child to solve a math problem by breaking it down piece by piece. OpenAI's new o1 preview model is different. It has this built-in ability to think through problems on its own - like internal chain-of-thought prompting. Open AI call them "reasoning tokens," which means the LLM can connect the dots without us spelling everything out.
This changes how we should work with it. Instead of writing long, detailed instructions like we're programming a complicated machine, o1 works better when you keep things short and sweet. It's like the difference between micromanaging someone and trusting them to figure out the details on their own. I've seen this firsthand – my shorter, clearer prompts often get better results than my elaborate explanations.
I've found that different situations call for different approaches. When I'm just throwing ideas around, I prefer using AI models that can handle my stream-of-consciousness style. But when I need precise results, especially for coding or technical challenges, o1 is my go-to. It's like choosing between a paintbrush and a precision tool – each has its place.
Finding Balance
At the end of the day, it's about picking the right tool for the job. Sometimes you need to let your thoughts run wild and see where they take you. Other times, you need laser focus to solve a specific problem. As these AI tools keep evolving, finding that balance becomes more important. We're learning to work with AI in ways that enhance rather than replace our natural thinking processes. In the future the models which reason on their own will become much faster and cheaper. I really hope this stream-of-consciousness style of prompting will stay effective - it would be a great loss if it didn't.
By Daniel Huszár