I’m writing a new book, using a mix of my brain and AI (Claude, deepseek, ChatGPT, and Raycast’s Advanced AI), and I wanted to share a tiny workflow that I’ve found very helpful, when using chatGPT.
It turns out that it’s really good at doing dictation in the app, and then cleaning it up.
Let me show you:
First, I started by typing this in, so that it did what I want - listen then clean up, just a bit.
Clarke: Okay, so I want to tell you a story, and I want you to listen to each bit as I go, and then just tidy up my words a little bit and play them back to me, so that I can collect the story, and then copy and paste it into a Word document. I don’t want you to summarize it, I just really want you to tidy it up a little bit. Is that okay?
I then dictated the first chunk of what I wanted to say into the chatGPT app, which it then turns into text using it’s whisper model, and it does a pretty good job.
It’s the circled microphone in the picture:
Note:
- this is different from the mac’s built in dictation, and from chatGPT’s conversational / advanced mode.
- I spoke until (a) I got sick of talking and (b) I started to worry that it maybe wasn’t listening to me.
I pressed the microphone🎙️button, and spoke for several minutes, off the top of my head:
Okay, so in this story, which is a true story, and I’m going to include it in my new book, which I think is going to be called Cash Cows Make the Best Burgers, it is set in Sheffield in the UK, though that doesn’t matter. I was asked to help them manage their first ever big Agile project. This was a company that sold software to banks that they used to run their mortgages and their mortgage accounts on. I’m going to refer to the two parties as the vendor and the bank. The vendor called me in because within the Agile community at that stage there were very few people who understood Agile and also project management. I was the only one that was available, I guess, and they contacted me and I went to visit them in Sheffield and I sat down and I spoke with the Chief Operating Officer, Jim, and the Chief Technology Officer, Mike. Those weren’t their real names, I don’t won’t to include them in the published book. When we sat down to talk, they basically confessed that the only reason they were doing Agile is that their relationship with the bank had gotten so bad over the decade or so that they had been working together that they absolutely hated each other, but the bank needed them to do a big upgrade to the software and the vendor was contractually obliged to work with them. The vendor was happy to take the million or so pounds that the bank would pay them, but they were dreading the experience because they had had roughly a decade of doing projects together that were just horrible and unpleasant, like an office-based civil war where the staff on both sides hated each other and had horrible, unpleasant, brutal experiences trying to deliver the projects. And previously, all of their projects had delivered very, very late and they had been very, very buggy and the bank were emphatic that they hated working with the vendor and the vendor was emphatic that they hated working with the bank, but they had this project and the bank and the vendor had both had tiny pockets of Agile tried within their businesses and the team seemed to like it, so they both decided that they would try to do this project in an Agile way, even though they had no idea how to do it. And they didn’t really want to do it, but they figured it probably couldn’t be much worse than all of the other projects they had done beforehand.
And then I pressed the button to stop recording, waited for chatGPT to convert the file it had just recorded, then clicked send.
It replied with this:
Got it! Here’s your cleaned-up version:
This is a true story, and I’m including it in my new book, which I think will be called *Cash Cows Make the Best Burgers.*It takes place in Sheffield, UK.
I was brought in to help manage their first big Agile project. The company sold software to banks—software that banks used to run their mortgages and mortgage accounts. I’ll refer to the two parties as the vendor and the bank.
The vendor called me in because, at the time, there were very few people in the Agile community who understood both Agile and project management. I was the only one available, I guess, so they contacted me. I traveled to Sheffield and sat down with their Chief Operating Officer, Jim, and their Chief Technology Officer, Mike. (Those were their real names, but I won’t include them in the published book.)
When we sat down to talk, they admitted that the only reason they were doing Agile was that their relationship with the bank had become toxic. After working together for over a decade, they absolutely hated each other. But the bank needed a major software upgrade, and the vendor was contractually obligated to work with them.
The vendor was happy to take the million or so pounds the bank would pay for the project, but they were dreading the experience. Their past projects had been brutal—like an office-based civil war. Staff on both sides hated each other. Projects had been late, buggy, and filled with conflict. The bank was emphatic that they hated working with the vendor. The vendor was just as emphatic that they hated working with the bank.
But they had this project, and neither could avoid it. Both the bank and the vendor had experimented with small Agile initiatives in their companies, and the teams seemed to like it. So, they decided to run this project in an Agile way. They had no idea how, and they weren’t particularly excited about it, but they figured it couldn’t be much worse than the disasters they’d already experienced.
That’s the first bit, cleaned up but staying true to your style. Ready for the next part?
And, I repeated the process for the next bit of the story.
This is the most enjoyable, least stressful approach I’ve found to dictation. I feel comfortable doing short bursts, because I can see the output and check if I’m going in the right direction.
And, I can keep adding to this story, as I need, and I have a draft.
A rough, rough draft!
And then the fun begins. I’ll copy the draft into ulysses, edit it a bit, move things around. And then I’ll pop it into one of the AIs (probably claude, since it seems to write nicely, but I’ll use raycast AI, because it has far better usage limits), and I’ll start to snazzy it up a lot.
I’ve tried for decades to do dictation because lots of authors have succeeded, but this is the first time I’ve found it helpful.
Speaking of helpful … I hope that helps.