Writing has never been my strong suit. While math was the easy thing for me growing up, trying to write anything was hard to say the least. My Dutch lessons where always a chore and this is also why I always struggled with making blog posts or writing newsletters.
However, my workflow changed significantly with the advent of Chat GPT and AI. Now, what I do is first collect my notes in Logseq, organize them a bit using block references and give it a careful read. Sometimes I draw it on my whiteboard before going off on a rant.
Voice-to-Text is a Game Changer
This is where otter.ai comes in, I use it to turn my voice into text, and once all these thoughts are converted into text, I’ll give it to Chat GPT with a couple of prompts. That I store under prompts/ in Logseq so I can tweak them over time. One prompt “Prompts/Act like me” explains to ChatGPT how to act like me. And another one, in “Prompts/Write WordPress Article” tells it to read my ramble and produce something with headings.
And now instead of sitting behind my screen and having to type things and being in a passive mode, I can grab my phone and walk around to get a little bit of exercise. My brain gets a little bit more oxygen, and I can keep rambling at it until it turns into a blog post that I can set up. This also avoids me trying to self edit, no self edit in voice, you just have to accept what you get.
Good AI-generated Content still needs work

The end result is pretty good, but it might miss things or misinterpret things. It often spells “Logseq” wrong, so I have to redact it at the end and go through the text, make sure that I add any missing points and clarify any points that I think are not captured well. But this is much easier than writing things from scratch.
In part I’m writing this blog post so people understand that I use AI, but not to generate content. I use AI to turn my thoughts into a coherent set of texts that I can then edit and turn into actual content that people want to read.
Misconceptions about AI-generated Content
I think there’s a misconception when people talk about AI-generated content. If something is not unique and just AI-generated without anybody checking it afterward, you can spot it so easily these days. I just read a post from another content creator, I can see that they clearly just said, “Look at the text from this video. Give me a summary”, and then put it on the website. No editing was done at all. You can look at it and go like, “This isn’t tweaked to be personal, or tweaked towards the target audience. This is just a really bad summary.”
In that regards it feels a bit like filters in Instagram, sure they help make a picture, but you still need to bring a good source to the table.
My WorkFlow
To recap, here’s a simple flow that I follow:
- Collect notes/thoughts in Logseq
- Gather notes and do a little prep work like place on whiteboard or outline topics/key points
- Use Otter AI to go from voice to text.
- Send the text to Chat GPT using prompts from my Logseq Prompt Namespace (prompts/)
- Post the text into WordPress.
- Clean it up before posting and add a couple of pictures to make easier to read.
In conclusion, using AI for writing has been a game-changer for me. While I still struggle with writing, this simple workflow has made it much easier to create blog posts from my chaotic thoughts.
This was made together with our future AI overlords
I’d love to try this as I was thinking of recording my gpt4 prompts in Logseq so I save my gpt4 convos on Logseq for future reference and more importantly completing the lineage of information so it’s all clear what were the inputs and outputs in each case.
About the audio issue you mentioned, I wonder whether fixed words like Logseq can be defined or if we can call it, my personal knowledge graph instead. Or whatever and then use replace for the final result
I usually save the results on prompts when talking to ChatGPT, while I have the GPT plugin configured I really love the easy chat that allows me to fine tune till I get the result I need. This might change as I gather optimized prompts.
And I got pointed out that Otter.ai has a personal dictionary so going to fill that with applications I talk about a lot.