
AI-Powered Facilitation Tools
May 20, 2026
Content Extraction and Production
May 21, 2026To Claude chat:
Here’s a new exercise for you: This is a transcript of a video I made. The probem is, it’s 10 minutes long. I would like to create a short i.e. <2-min and and a 5-min video from this content.
The 2-minute short
Cold audience, written by Claude, no edits.
Approx. 165 words / ~1:50 at natural pace
Actual length of cloned version 1:25
Do you ever notice that the voice in your head doesn’t say I?
It says you. You should have handled that better. You should get a grip. You should be further along by now.
That’s strange when you think about it — because it’s your own voice. But it’s not talking to you as yourself. It’s talking to you as if you were someone else looking at you from the outside.
And that’s the actual problem. Not the content of what the voice is saying — but the position it’s speaking from. You’ve become an object to yourself. Something that needs to be assessed, improved, criticized.
The thing is, you’re not an object. You’re a process. Constantly changing, constantly in motion. And there’s a much more natural place to operate from — which is simply: I. Me, looking out at the world. Not me, imagining how the world is looking at me.
When you’re in that position — inside your own experience rather than outside it — difficult situations become easier to handle. Not because the situation changed. Because you stopped adding a layer of self-surveillance on top of it.
That shift — from you to I, from object to subject — is at the centre of the work I do. And it’s available to you in any moment.
The 5-minute explainer
Warm audience, written by Claude, no edits.
Approx. 450 words / ~4:45–5:00 at natural pace.
Actual length of cloned version 3:14
I want to talk about something I use a lot in the work — this distinction between experiencing yourself as a you or as an I.
Notice how the critical voice in your head talks to you. It doesn’t say I. It says you. You should never have done that. You should have handled it better. You should, you should, you should. It’s almost never a compliment.
And there’s something worth paying attention to in that grammar. Because the voice isn’t speaking as you — it’s speaking at you. It’s treating you as an object. Something external that needs to be evaluated, managed, improved.
Which means, without realising it, you’ve stepped outside yourself. You’re no longer occupying your own experience — you’re watching yourself from a distance, usually through imagined eyes that are considerably harsher than anything you’d apply to someone else.
I notice this in myself. When I’m in a stressful situation and it’s feeling worse than it probably needs to, it’s almost always because I’ve shifted into that outside view. I’m not asking: what do I actually need here, what’s actually going on? I’m asking: how does this look? What would someone else make of this? And the answer my mind generates is rarely kind.
People aren’t objects. They’re subjects. You are a process — constantly in motion, constantly changing, thoughts shifting, states shifting — with some thread of continuity running through the middle. When you relate to yourself as a process you’re inside of rather than an object you’re observing, something changes. The critical voice doesn’t necessarily disappear, but you’re no longer giving it the chair at the head of the table.
There’s also something worth naming here culturally. We’re living in a moment that is actively training us to see ourselves from the outside. Social media, the selfie, the curated profile — these are all outside-in orientations. You’re constructing a version of yourself for external consumption, which means you’re relating to yourself as an object almost by default. That’s not inherently wrong — there are situations where you need to consider how you’re coming across. But as a chronic operating mode, it’s exhausting, and it tends to make hard situations harder.
What we’re doing in the releasing work is shifting that orientation. Not through analysis of where the critical voice came from or whose voice it originally was — but by coming back to I. To the subject position. To me, here, looking out at the world — rather than standing outside myself watching.
And from that position, even when things are genuinely difficult, there’s more steadiness available. Not because the difficulty has gone — but because you’re inside your own experience of it rather than performing it for an invisible audience.
That’s the shift. And it’s what I mean when I write about the you versus the I.
Next steps
The cloned version of my voice is quite accurate, but the tone is not inviting. Claude did a great job of ordering a meandering script. I will try recording these scripts in my own voice.
But by doing this exercise I discovered a new application that would massively improve productivity when editing videos.
Video Editing Workflow with Claude Chat
Request to Claude
What I really want this to do the following:
- I give you a transcript of a video with timestamps.
- You edit it to make more sense and flow clearly, but with the a very strict rule that you cannot change any of the words (because the video is already shot)
- You then produce a more linear, clearer version of the script with time stamps so that I can then edit the video to match your script.
- Can you do that?
Claude’s response
Yes, straightforwardly. The constraint is clean: reorder and cut, never rewrite.
To confirm the exact workflow:
- You provide a timestamped transcript
- I identify the best segments for each version (2-min and 5-min), cutting repetition and false starts, reordering where the logic improves
- I output a edit decision list — the selected segments in the new order, each with its original timecodes, so you can execute the cuts in Filmora
The one thing worth flagging in advance: jump cuts. Where I reorder segments that weren’t adjacent in the original, you’ll have a visible cut. For a “thinking aloud” style video that’s probably fine — but worth knowing before you’re in the edit.
What I need from you:
The transcript with timestamps. The denser the timestamps the better — ideally every sentence or every 5–10 seconds rather than just at paragraph breaks. Most transcription tools (Whisper, Descript, Otter, YouTube auto-captions export) will give you per-sentence or even per-word timestamps. If you’re not already using one of those, Descript or running Whisper locally would give you exactly what this workflow needs.
Ready when you have it.




