Offloading Execution to Protect Flow

My family had gone up to Northumberland early to see the in-laws and were staying on. I'd just gone for a long weekend, a lack of annual leave prevented a longer trip. It was a great weekend, playing with the kids, Barter Books, long walks, and I had a lot of audiobook time on the solo drives. The Critical Chain on the way up, The Culture Code on the way back, good books for long empty roads.

Forty-five minutes into the 3.5 hour drive home, I pulled over at a service station, not for petrol, to dump some thoughts into AI before they evaporated.

That became the pattern for the day. Get home, spend a few hours developing pathway ideas, go for a walk and let things settle. End up at my favourite pub with a chair by the fire, spinning up chat after chat as ideas kept arriving.

By evening I'd started maybe a dozen conversations, some will become posts; some already have, some will fizzle out completely.

What made it work wasn't that every thought turned into something valuable. It's that I had somewhere to put them that didn't require me to develop them fully in the moment.

---

I work on this blog mostly alone. There are people I bounce ideas off occasionally, a friend helps me edit; but this is fundamentally a solo project. No team to grab a coffee with and say "I'm thinking about this, what do you reckon?" No colleague to hand a rough idea to while I go and think about something else.

In my day job, that kind of thinking-out-loud is constant. You're in a meeting, something occurs to you, you scribble it on the whiteboard and someone picks it up. Or you send a message to a colleague and they respond three hours later with a thought that takes it somewhere you hadn't expected. The thinking is distributed. It happens in the gaps between people.

Working alone on a personal project, that doesn't exist. Usually a half-formed thought either gets developed immediately, which stops whatever else you were doing, or it evaporates.

That Tuesday, AI was filling that gap.

---

I'd have a thought on the walk, open a chat, throw it in and prompt AI to ask clarifying questions which I'd come back to later. Sometimes the questions revealed there was something worth pursuing, sometimes they revealed there wasn't. Either way, I found out quickly without having to commit serious thinking time upfront.

This isn't like jotting things down in a notebook; A notebook is passive storage, the thought sits there exactly as you left it. When AI asks clarifying questions, the answers draw out crucial details and content that weren't in my initial note, things I knew but didn't think to write down, assumptions I was making, connections I saw but didn't articulate. When I come back hours later, I'm not looking at a cryptic misspelt note trying to remember what I meant, I'm looking at my original thought plus the details which the questions extracted.

I had multiple conversations going and had to be cognisant of when to spin out a new chat versus continuing an existing thread. Sometimes an idea belonged with something I'd already started, sometimes it needed its own space.

The cost of exploring became low enough that I could afford to follow things that might not work out; ten ideas, most of which will fizzle. In the past, I'd have had to pre-filter, pick the "best" one, commit early and develop it fully. Now I can explore all ten and see which ones survive first contact with questions.

---

This shows up differently in my day job. When I'm writing presentations or communications for large audiences, the cognitive load that matters is the story. What's the narrative arc? What's the sequence of ideas? What do they need to understand first before the next bit lands?

Once I know the story, the actual writing, turning "here's the three-part arc I want to build" into slides or an email, that's execution labour. And it's cognitively expensive in a way that crowds out the story work if I'm doing both at once.

The old pattern was constant context-switching between thinking (what's the story?) and execution (how do I phrase this?) that switching kills my flow. I never got sustained time at the thinking level because execution keeps pulling me back down.

Now I stay at the story level continuously. Structure, sequence, what lands with this audience, AI handles the draft. Later, when the story work is done, I come back and validate. Edit what doesn't sound right and rework what missed the mark.

---

I've written before about working in two modes — exploratory and execution. That framing mostly holds. But what I noticed that Tuesday was something sitting underneath both of them.

The thinking had been happening over a few days. Reading time in Northumberland was inputs; but it was the drive home where things clarified. Pulling over forty-five minutes in to deposit thoughts meant they didn't evaporate. Hours at home developing pathways. A walk to let things settle. The pub was where multiple threads came together and I could spin up conversations at a higher level because the earlier deposits were still there, I wasn't trying to remember what I'd thought three hours earlier.

Then home to refine. Which is what's happening right now, actually. This post started as one of those pub conversations. Now it's the validation pass.

That arc wouldn't have worked the old way. Too many thoughts arriving at different times. Too much context to hold. Too easy for things to evaporate between the walk and the pub, or between the pub and getting home.

AI didn't make the thinking happen, the thinking happened because I'd had good inputs, time to process, and space to let ideas settle. What AI did was hold the execution layer so I could stay at the thinking layer across the entire arc.

---

I'm not sure what to call this. It's not a delegation. It's not a collaboration. It's something closer to cognitive infrastructure, the external scaffolding that lets the thinking happen without losing what arrives in the gaps.

In a team, that scaffolding is people, the colleague you message, the meeting where you think out loud, the whiteboard that stays up for a week. Working alone, you don't have that. But for a solo practitioner working on a personal project, AI fills enough of that gap to change what's possible.

Most of the conversations I started that day will go nowhere. A few might become posts. One already has, you're reading it.

The difference is I got to find out which was which without having to commit upfront to developing all of them fully. Not a perfect capture of every thought. Permission to explore without committing.

And occasionally, on a good day with the right inputs and enough space to think, that produces something worth keeping.