STEVEN HOLT · LEARNING IN PUBLIC

Learning to make videos.
With AI, in the open.

I’m documenting how I learn to create videos and build a personal social media presence using AI, including the real prompts, workflows, gear, and editing techniques I use and test.

Browse the Prompt LibraryRead why I’m starting
MY STORY

Why I’m starting.
From zero, on purpose.

I’ve never seriously invested in creating personal social media content, and I don’t particularly want to become an influencer.

I’m doing this because I have projects worth sharing, and I want to learn how AI can make content production easier for someone starting from scratch.

This site is my notebook. The useful part for you is the working material: the prompts, workflows, equipment, software, and editing techniques I actually use and test, including the ones that don’t work.

THE EXPERIMENT

Walk. Talk.
Record everything.

I’m experimenting with conversational walk-and-talk recordings, sometimes called yapping, filmed on my phone with an Insta360 Flow 2.

Ground rules from my plan

  • Film on my phone only for the first 30 days.
  • Reach 20 videos before I start refining the subject matter.
  • Record time, cost, and what went wrong alongside what worked.
EXPERIMENT 01Underway

Walk-and-talk content with the Lake Method

Format
Conversational walk-and-talk recordings (“yapping”)
Camera
My phone
Stabilizer
Insta360 Flow 2
Library
Obsidian vault: original media + transcripts
Method
The Lake Method, explained below
Running logPlaceholder
Recordings logged
—
Hours recorded
—
Videos published
—
Money spent
—

Not tracked yet. Real numbers will replace these dashes as recordings are logged.

WATERFALL VS. LAKE

Don’t cut one video down.
Fill a lake instead.

Most repurposing advice is a waterfall: finish one video, then turn it into many smaller pieces. I’m testing a different shape. Original media and transcripts go into an organized Obsidian library, and AI pulls material from many recordings to make something specific for each platform.

The usual approach

The Waterfall

Finish one video, then cut it into smaller and smaller pieces.

  1. One finished videoEverything starts here
  2. ClipsCut from that video
  3. ShortsCut from the clips
  4. Posts + quotesCut from what is left
One source. Each step down has less to work with.
What I’m testing

The Lake

Every recording flows into one organized library. AI draws from all of it to build each piece.

  • Walk-and-talk 01
  • Walk-and-talk 02
  • Screen recording
  • Voice note
The content libraryAn Obsidian vault
  • Original media
  • Transcripts
  • Topics + tags
  • Linked notes
AI retrieves across recordings
  • YouTubefrom Walks 01 + 02
  • Shortfrom Walk 02
  • LinkedIn postfrom Walk 01 + screen recording
  • Blog articlefrom Every transcript on a topic
Many sources. Each platform gets a piece built for it.
WaterfallLake
Starts withOne completed videoEvery original recording and its transcript
Material flowsOne way: down from a single sourceInto a shared library, then out on request
Each platform getsA smaller cut of the same pieceA piece built for that platform and audience
A weak recordingWeakens everything cut from itIs one source among many
Reuse over timeEnds when the video is used upGrows with every recording added

Hypothesis The diagrams above are illustrations, not real recordings or published pieces. Whether the Lake Method actually saves time is what the experiments will record. The library itself is planned as The Library project.

WHAT YOU’LL FIND HERE

The working material.
Shared as I use it.

Prompt Library

Searchable prompts for ChatGPT, Claude, research, scripting, Obsidian, video editing, and content production. Each with what it is for, what actually happened, its limits, and its revisions.

Being built

Video Workflow

My production process from filming and transcription through editing, publishing, and analysis.

Being built

My Setup

The equipment, software, settings, and costs behind the workflow, marked as owned, testing, or considering.

Being built

Experiments

Real-world tests, including failures, time invested, costs, and lessons learned.

Being built
HOW EACH PROMPT ENTRY WILL WORK

Copy it. Then read what actually happened.

Every prompt in the library will show its purpose, real results, limitations, and revision history, so you can judge it before you use it.

Open the Prompt Library
Example layout · not a tested prompt
ClaudeScriptingObsidian

[Placeholder] Outline a video from several walk-and-talk transcripts

Prompt text will appear here once it has been written and tested.
Purpose
What the prompt is for.
Actual results
Not tested yet.
Limitations
Where it falls short.
Revisions
v0 · draft

Follow the experiment.
Mistakes included.

Links to my channels will go here once the first video is published. Until then, the library and the experiment log are where the work shows up first.

Placeholder · channel links coming
Browse the Prompt LibrarySee the experiments