Welcome
In this issue, I bring you into my personal graph and share how I am using AI as a thought partner.
Problem
The good news about Roam is that it meats you at the level of thought.
It doesn't feign organization or overpromise. Crucially, it distinguishes thought and afterthought.
The bad news is that for an ADHD brain like mine, this free range leads to a sprawling, nested, half-formed stream of consciousness, that...
It's a mess, but it's a context rich mess. With a little disambiguation, it's good fodder for an LLM's response.
Modern LLMs are getting increasingly good at discerning human intent. And having an LLM at your fingertips, the disambiguation doesn't have to be your problem, it can be the AI's problem. The LLM can take care of the 'afterthought'
Solution
The LLM at my fingertips is | Fabrice Gallet's | Live AI | extension.
Fabrice's implementation of his extension can solve key questions for me.
'How soon can I summon an LLM to interact with my messy thoughts?'
'Can my messy thoughts serve as both prompt and context for the LLM?'
So, 'How soon can I summon an LLM to interact with my messy thoughts?'
well, how soon can you hit Cmd + Ctrl + A?
This command brings up the LiveAI interface and by default the 'focused blocks as prompt is selected'
hitting enter, the LLM takes the block on which your cursor is currently on as the prompt
And, 'Can my messy thoughts serve as both prompt and context for the LLM?'
To answer this question, it’s important to note that where your cursor is currently positioned determines what the LLM considers as its context and prompt.
In my case I will move the cursor up the level of blocks that I want it to consider as context and then run the 'focused block as prompt’ command.
*Note that this only works if you have checked the 'Include children in prompt' option in the native LiveAI settings
Quick Tip! I have found that including the last child block with the specific question you want the LLM to respond to works best.
i.e. The LLM will take the preceding blocks as context but specifically respond to the last block as prompt
Thoughts
The core insight here is about separating the burden of coherence from the act of thinking itself.
You know that frustrating moment when the perfect words or ideas come after the conversation ends.
Our initial thoughts rarely emerge in their clearest, most coherent form and unfortunately, traditional tools force us to organize and clarify before we can effectively work with our ideas.
This workflow anticipates the afterthought process that happens in our minds and reduces the latency between thought and afterthought.
The AI can interpret your raw, unformed thoughts and help articulate your true intention. This happens ‘in the moment’, not as a delayed reflection.
Roam meets you at the level of thought, the AI meets you at the level of afterthought.
Have an interesting workflow in Roam and wish to be featured? Please reach out below.
To support the creators and the tools mentioned
Fabrice Gallet (Live AI) - buy me a coffee