Dumping your brain using LLMs
Dump of a random thought. AI tools are your friend in verbalizing your knowledge and thoughts.
Disclaimer: While the title might suggest otherwise, the first draft of this text was written by a human without using LLMs. Although I use LLMs for proofreading, the text still remains largely my own words.
In the previous blog post, I described how context matters to get quality output from LLMs and AI tools. As a next step, I’ve lately been writing down my domain knowledge about the systems and software I’m working on. It’s required to delegate more tasks to AI tools.
When we work on software development tasks — though I suspect this isn’t unique to software development, I’ll keep the scope narrow here — we leverage a lot of tacit knowledge. While it is sometimes articulated for sharing, most of it stays trapped inside individual minds, and people often aren’t aware it exists. However, AI tools need it to do a better job.
It’s a daunting task to verbalize tacit knowledge as a logical document. Even if we understand the concepts and nuances in our brain, we easily run into trouble explaining them in natural language. It’s difficult even in my mother tongue, Japanese. Needless to say, it’s even harder in English. This is one reason tacit knowledge largely never leaves individual minds. However, AI tools also make the hurdle of verbalization much lower.
Tacit knowledge is more like intuition rather than structured memory. Unless we are intentional, it stays unorganized. To verbalize it, we need to find words that match the nuances of the knowledge. This process takes a long time, but this is the point we can accelerate with AI tools. They are very fast to expand short nuanced concepts into logical documents. The output may not always be perfect, but it’s still better than letting the knowledge stay only in individual minds.
Verbalizing knowledge and thoughts is good for organizing memory, and it also sometimes leads to new findings. To be honest, I was surprised by the amount of domain knowledge I silently assumed while working on tasks, until I wrote it down.
Are you concerned that your job security is threatened if you expose knowledge only you know? I’m optimistic about that point. If your knowledge is valuable, you should be valued for the ability to produce that knowledge. Also, knowledge is just information. It creates value when it’s utilized to solve real problems. Your skill in seamlessly applying domain expertise to complete tasks should also be rewarded. I feel rather relaxed after writing it down because I’m no longer the single person who knows it. I can now let other members take on tasks that need this knowledge.
Dumping tacit knowledge is valuable for both you and others. AI tools support the heavy lifting of verbalizing it. Spending a little time on verbalization improves your overall productivity for sure.