Today I learned about Chain-of-Thought Prompting which is a technique to get potentially better results from an LLM where you craft your prompt with an example of the chain of thought or steps in reasoning to solve a complex problem. This gives a good explanation: https://www.promptingguide.ai/techniques/cot
I’ve been discovering recently that ChatGPT is actually pretty good at helping write Architectural Decision Records (ADRs). Like most writing tasks, I’m finding ChatGPT is good at giving me a nice starting point instead of staring a blank template. Even with as simple a prompt as “write an architectural decision record on why we chose to go with react instead of angular”, it’ll give a decent list of pros and cons on the topic in a common ADR template....
Today I learned that chatGPT is pretty good at explaining cryptic bash scripts and commands. I was trying to understand exactly how a git prepare-commit-msg hook was working so I asked chatGPT to “explain the following git hook to me” and it did a pretty good job. For example, it returned tidbits like: [[ $BRANCH_NAME =~ $BRANCH_REGEX ]]: This conditional statement checks if the branch name matches the BRANCH_REGEX pattern. If the branch name starts with one or more digits, this condition will be true....