An AI presenter, a voice that sets the clock, and a production line that checks its own work.

Watch the new ReelsSee the rebuild
Favorite
Reels · VideoFourteen phones, the talking shots of 15 Reels and a Studio preview, the text cards cut out. Silent; every frame is my own AI-made work.

Project facts

Role
Product owner, AI director
Timeline
May to Oct 2026
Built with
  • Claude Code
38.6Kpublic views on YouTube and TikTok (October 29, 2026)
9Reels approved in the four days after the rebuild
1command from a script to a finished, mastered Reel
494numbered laws the agents follow, most written after a real failure
~$2of AI generation per Reel now, down from $4 to $10

I designed a production line for 30-second explainer Reels presented by an AI host, Mira, and directed Claude agents to build it. In October I threw away its look and rebuilt it in five days: one command now turns a script into a finished, mastered Reel, Mira's approved takes are reused instead of regenerated, and every idea gets a drawn stage of its own. I am not a developer. My job is to decide what good looks like, and to say no.

01

One Idea in, One Finished Reel Out

Decyde is a decision-training product: courses teach what to think, Decyde trains how to decide. To market it I wanted a steady stream of short explainers with a consistent face, without filming anyone. So I designed an AI UGC factory and had Claude agents build it.

The Problem

One AI clip is easy. A series is hard. The presenter's face drifts from clip to clip, the mouth misses the words, graphics look random, and every Reel gets rebuilt by hand in an editor.

Working through AI agents adds a second problem: many short sessions with no memory have to behave like one careful engineer, and spend money only when I say so.

What I Built

A pipeline where one idea becomes a 30-second Reel: the point, a script in plain words, the voice, word-level timing, Mira's mouth synced onto an approved take, a drawn stage for the idea, sound, and since October one command that renders and masters the finished file.

Around it, the part I am proudest of: a written law of 494 numbered amendments, a gate whose checkers must prove they can fail, and one rule above all of them. Nothing is done until a human has watched it.

01.1

Under the Hood (October 29)

464commits in 39 active days
91Python tools: pipeline, checkers, emitters (37,287 lines)
146TypeScript files in the motion studio (25,674 lines)
31design-system primitives and 16 components on one contract
60self-tests that plant a bug and must go red
494numbered laws: A1 to A444, then A445 and 49 sub-amendments

My Role

Product owner and AI director. I set goals, made every taste and money call, recorded the voices until October and approved every paid take since, judged every output and said go before any spend. Claude agents wrote the code.

Timeline

May to October 2026. Month one: nine Reels posted from a cold start. October 24: the reset. October 25 to 28: nine Reels approved on the rebuilt system.

Team

Me, plus Claude agents in two seats: Cowork in a Linux sandbox, and Claude Code on my Mac.

Stack

Claude, Higgsfield (stills, talk takes, lipsync, pictures), ElevenLabs (the voice over the API since October), Remotion, Python; After Effects with ExtendScript until October, optional since.

02

The Reels, Newest First

The nine Reels approved after the rebuild, October 25 to 28. Each asks one everyday question and draws a stage of its own for the answer; in five of them you pick before Mira tells you. Below them, the Reels of May to September. Every video waits for your tap; nothing loads before it.

Showreel · VideoSilent showreel, 31 s, October 2026. Every frame is my own AI-made work.
Mira in a black blazer holding a small Decyde microphone, under the hook Two reps. Same client. $30K apart.Oct 25 · The rail

Two Reps. Same Client. $30K Apart.

The first Reel made end to end by our own renderer, ten hours after the reset: one price rail, Mira on an approved take, only her mouth bought (38.4 credits).

Mira in a black top in a room over the city at night, under the hook When a friend asks you for a favour, do you say yes right away?Oct 26 · The day

Do You Say Yes Right Away?

A Saturday as a ring that every yes eats. Three rounds of drawing on my notes, the first new Mira, and the closing that builds the Decyde logo.

Mira in a black top at night, under the hook Nobody decides to scroll an hour.Oct 26 · The night clock

Nobody Decides to Scroll an Hour.

The first Reel on the hook rules: the stake on screen in the first second. A study of about 2,000 people drawn as a clock that a limit stops at 38 minutes.

Mira in a white tank and an open blue shirt in a loft with books, under the hook Great advice, just not to yourself.Oct 26 · The maze

Great Advice, Just Not to Yourself.

Walls that rise round you, then flatten from outside. The first Reel in Sapphire, the voice I picked by ear for every Reel after it.

Mira in round gold glasses and a grey tank in a loft, under the hook Wait a month for $10?Oct 27 · Play along

Wait a Month for $10?

The first play-along Reel: you pick $100 today or $110 next month on a timer, then the same deal a year away. The road's perspective does the explaining.

Mira in a brown corset vest in a record listening room, under the hook Did they like you?Oct 27 · Play along

Did They Like You?

Rate out of ten how much they liked you. The liking gap is drawn at its true size, and a new Mira was made for it in one pass, 54 credits.

Mira in a black blazer in a bright kitchen, under the hook Bad hair day. Who noticed?Oct 27 · Play along

Bad Hair Day. Who Noticed?

Guess how many of eight people noticed. A 4-second approved take carries the whole Reel, the cheapest so far at 37.39 credits.

Mira in a white tank and an open blue shirt by a bookshelf, under the hook How long will it really take?Oct 28 · Play along

How Long Will It Really Take?

Pick your worst case on a calendar. Deadlines break, the calendar runs out of rows, and the paid voice was trimmed to 30 seconds for free.

Mira in round gold glasses and an oatmeal cardigan, under the hook Two trips. Same weekend.Oct 28 · Play along

Two Trips. Same Weekend.

Two ski passes, each a window onto its trip. Redrawn as a world the hour I called the first version too simple, for 0 credits: the art direction for every stage after it.

Finals are my own exports, re-encoded for the web. Four of the nine were public on October 29; the five play-along Reels are approved and waiting to be posted.

02.1

Earlier Reels, May to September

The first era's posted Reels, the two that were finished and never posted, the trend Reel, then the Studio studies of September and October. Tap any of them to play.

Mira, the AI presenter, holds a small black mic to her lips by a bright window; the caption reads win, you could.Reel 05

What If It Works?

Mira in a cream sweater speaks into a mic by tall city windows; the caption asks want to get rich in your 20s?Reel 10

Some Sports Are Access.

Mira on a rooftop at golden hour with city towers behind her, holding a mic; the caption begins be ridiculously.Reel 17

Be Ridiculously Optimistic About Your Life.

Mira in a black leather jacket in a cafe, speaking into a small mic; the caption says you're not tired.Reel 21

You're Not Tired. You're Carrying.

Mira in a cafe with shelves behind her, holding a small mic; the caption says restarting isn't free.Not posted

Restarting Isn't Free.

Mira on a sunny cafe terrace in a white shirt and denim shorts, speaking into a mic; the caption says you think they're just boring.Not posted

You Think They're Just Boring, Right?

Mira in a black blazer in a warm kitchen, holding the Decyde mic with its Y mark; the caption says so say you cut it.Preview

Good, Better, Best

Reel 07: Mira in a cream sweater speaks into a black mic; the caption says I'm tired.

I'm Choosing the Pressure Again.

Reel 14: Mira smiles behind a bunch of white flowers; the caption says the future isn't found, it's chosen.

The Future Isn't Found. It's Chosen.

Reel 11: Mira in a black top speaks into a studio microphone; the caption says stay strong.

Some Days Are Heavy. That's Normal.

Reel 12: Mira in a black top holds a clip-on mic in a warm room; the caption says at work.

You Don't Lack Time.

Reel 13: Mira by a window with a fluffy mic; the caption says if you're.

Listen. Don't Defend.

Reel 15: Mira on a Paris street in a cream cardigan holding a mic; the caption says leaders do.

Great Leaders Don't Just Tell People What to Do.

Reel 16: Mira on a sunny terrace in Lisbon holding a mic; the caption says let me say.

Trust Isn't Built by Helping People.

Reel 18: Mira in a cream sweater speaks into a black mic in a dim room; the caption says everyone plans.

Everyone Plans How to Win.

Reel 19: Mira on a couch at night with fairy lights behind her; the caption says liking who you are.

What Success Actually Means

The trend Reel: Mira, the AI presenter, sits on an open window sill above a city street; the caption reads one more time.Trend

Alright, Let's Do This One More Time.

Studio preview: a hand-drawn magnet and the words open high on a dark frame.Preview

Open High: The 30-second Template

Studio preview: the words the Decyde on a dark frame.Preview

The Scene Kit

Studio preview: the words the plan nobody on a dark frame.Preview

Motion Study

Studio preview: a round hand-drawn character named Dot next to a ladder on a dark frame.Preview

The Cast: Dot and Stick

Studio preview: a grid of fifteen small hand-drawn animations, each labelled with its name.Preview

Fifteen Motion Primitives

Studio preview: three stick figures under price tags on warm paper with the words opens high.Preview

The World Grammar, Paper

Studio preview: the same scene on a dark ground with the words opens high.Preview

The World Grammar, Night

03

Four Apps, One Face

Every Reel is posted by hand to Instagram, TikTok, YouTube Shorts and Facebook as @decydehq. These are real public captures, taken logged out on October 29, 2026; the newest posts on top are the rebuilt Reels.

The Decyde channel on YouTube in a phone: the Y logo, 18 videos, and the newest Shorts on top, each opening on Mira with a serif hook line.
YouTube Shorts18 Shorts, about 15,500 views. The four newest are the rebuilt Reels, at 237 to 294 views in their first days.
The decydehq Instagram profile in a phone: the Y logo, the bio 'Courses teach what to think. Decyde trains how to decide.' and a grid of Reels with Mira in every frame, the newest with serif hooks.
Instagram Reels18 Reels. The four newest open on Mira with a serif hook line under her neck.
TikTok's profile card for decydehq: the Y logo, the bio and the four newest videos, a drawn maze and three Mira hooks, at 220 to 251 views each.
TikTokTikTok through its own profile embed: 14 videos, about 23,200 views. The four newest are the rebuilt Reels; the 16.7K video is a trend Reel.

Captured logged out on 2026-10-29. Counts move every day. The same Reels go to Facebook.

04

Meet Mira

Mira is an AI-generated presenter, not a real person. Every picture of her was generated, then checked against her approved face before it was used. These are her looks, her expressions and the sets she talks in.

She began on the evening of May 22 as candidate number 7 of a weekend of Soul portraits. That one picture is still the ground truth: every new image of her is measured against it.

In July she got a written canon: the face, black hair with a platinum underlayer, layered gold chains, the house mic with its Y mark. Then a wardrobe: fashion looks shot in four views each, 28 named expressions, idol poses, and a look system of 18 rooms and 12 lights. Tap any picture to see it large; fashion looks open with all four views.

Since October 25 she is also a gallery: seven approved talk takes, each checked for face match and framing, that any new script is lip-synced onto. A new Reel starts with one picture of the takes, and I pick her by eye.

4views of every fashion look: stride, front, back, close-up
28named expressions a prompt can ask for
21approved faces her identity is checked against
345Higgsfield generations of Mira, Luna and casting candidates
7approved talk takes in her gallery, carrying nine Reels
04.1

Before Mira: One Weekend of Casting

On May 22 and 23 the Soul model produced 47 portraits from a few written descriptions, and the face kept changing. Candidate number 7 became Mira. Here are sixteen of them.

04.2

How a New Mira Is Made

On October 25 Mira put down the mic and got new clothes. Outfits written in words came out plain, so my own outfit sheet went in as a picture beside her anchor face. Each fix below is now a rule.

A Picture, Not Words

Mira in a black-and-cream tweed jacket over a black top, seated in a dark studio.
Words
Mira in a black leather jacket slipped down to her elbows over a black top, seated in a dark room with city lights.
My picture

Written outfits came out plain: a checked print, covered arms. My sheet's outfit, given as a picture, was copied on the first try.

Cut the Outfit at the Neck

Mira in a black top with a leather jacket on her shoulders, a dark room at night.
Face 0.446
Mira in a black top with a leather jacket draped on her shoulders, a dark room at night.
Face 0.574

A face inside the outfit picture leaked into hers. Cut below the chin, her face held.

Cream Needs a Black Layer

Mira in a black sleeveless top, seated in a dark studio with plants and brass lamps.
Asked for cream
Mira in a camel vest over a black top, seated in a dark studio.
Held

With her black-clad anchor as the only picture, a cream vest came back black. A black top under a coloured vest held.

Never Outpaint Her

Mira in round gold glasses and a cream cardigan over a grey tank, seated in a loft by a tall window.
Face 0.503
The same photo of Mira in glasses, wider, with more of the room around her.
0.439, not used

Widening a photo redrew all of it and smoothed her face. The fix: ask for the framing in the photo itself.

04.3

Framing Is Measured, Not Guessed

The video model pushes in on its own. Look K2's first take grew her face from a quarter of the frame to a third. The photo was remade with her head at a fifth of the frame and calm moods, and the framing held for all 11 seconds: that take is m05.

First take Pushed in by 37 %. Not used.
m05 The framing held. Used in two Reels.
04.4

The Score Is a Check, My Eye Is the Verdict

Every photo and take is scored against her anchor face, and a take needs 0.50 to enter the gallery. Glasses cost the score, not the person: the best of these, 0.518, lost to an earlier photo at 0.496 because I liked her hair tied up, and m07's take entered on my eye with frames as low as 0.44.

The glasses look, October 26: her anchor face, then four photos and their scores.

04.5

How a Reel Picks Its Mira

One picture of the gallery takes with the size of her face in each frame. I choose by eye; the face check and the framing rule were already passed when the take entered the gallery.

The choice for worst-case, October 28: m06 won.

04.6

The Gallery

05

Eleven Stages, One Clock

A Reel moves through eleven stages. By October 28 most of them were code, three were my hands, and every stage had been rebuilt more than once. Pick a stage to see its tools, its numbers, the laws it obeys and how it changed.

Stage 01 of 11

Research and the Point

Since October Claude researches and pitches three everyday topics, each with a peer-reviewed finding behind it, and I pick one. The hook rules put the stake on screen in the first second.

Every Reel argues one point from one Decyde course module, always in the same shape: it was never the obvious thing, it's the real thing. One drawn object, the carrier, threads the whole Reel.

module_script.py pulls a module's own facts and never invents: a gap prints as [NEEDS WRITING] instead of a plausible number. story_check.py drafts the beat plan and refuses a plan that argues nothing.

  • Claude
  • module_script.py
  • story_check.py
  • story.ts (roles and shapes)
  • Decyde course library
  • hook and topic rules (H1 to H8, T1 to T5)
  • a fact-checked topic bank
Source library
267 modules, 9 domains, 71 courses
Argument shapes
4: denial, mismeasure, price, twins
Story checks
10 checks, proven on 15 seeded failures
Drawable carriers
26 figures
The hook
first word at frame 0, a title of six words at most, the stake by 1.5 s

How It Evolved

  • May Ideas came from short videos I liked: the insight reworked, a script written around it.
  • Aug 22 module_script.py: the Decyde course library becomes the content engine.
  • Aug 31 A320: only Decyde material from now on.
  • Sep 1 A321: the story grammar becomes a gate (roles, one carrier, rhythm).
  • Oct 26 Hook and topic rules from a study of 54 creator Reels: everyday topics, the stake in the first second.
  • Oct 29 Topics beyond decisions: fun facts and myths, each fact-checked with a safe line.

Laws It Obeys

  • A297 The point comes first, before any draft. The draft is the cheap part.
  • A258 The content engine extracts. It never invents.
  • A320 Decyde course material only, so every Reel promotes the product instead of generic motivation.
05.1

A Reel, Start to Finish (Since the Rebuild)

This is the path since October 25, for a Reel on a gallery Mira. My hands are on six steps: pick, say go three times, approve, post. Everything between them is code that checks itself.

System work and every preview cost 0 credits. A Reel costs its mouth, a few pictures and its voice: about 41 Higgsfield credits and 900 ElevenLabs characters on the play-along Reels, roughly $2.

  1. I pick the topic from three researched pitches, and Mira from one picture of the gallery takes.
  2. Claude writes the script in plain words; voice_check refuses the paid voice until it passes.
  3. Free preview reel2.py build with a stand-in voice and the gallery take: the whole Reel for 0 credits. I watch and steer.
  4. Voice, after my go tts2.py makes five takes on one budget; voice_pick.py picks one; its word timings become the clock.
  5. Pictures, after my go two to four cut-in pictures at 0.25 credits each, graded to our paper.
  6. Mouth, after my go line 1 is lip-synced onto the approved take first; I look; then lines 2 and 3.
  7. One command reel2.py renders with Remotion and masters to -14 LUFS; the layout probe and the goldens check it.
  8. I watch and approve with sound, on my phone. My words go into the spec as the approval.
  9. reel2.py post the video, caption, title and cover land in one folder, and I post to four apps by hand.
06

Eight Eras in Five Months

It began as a chat that made short clips of an AI presenter and a fresh After Effects script for every Reel. Every failure became a rule, the rules became gates, the gates became code, and the code became a studio. Then, on October 24, I reset the look, and in four days the rebuilt system made nine Reels.

01May 2026v1 to v3.25 · A1-A66

The Launch: A Face, a Mic, Black Text Cards

Mira, a photoreal AI presenter, talks into a small foam mic that hides her mouth, in 4-second clips cut against black kinetic text cards. I recorded every voice, Claude wrote a fresh After Effects script for each Reel, I ran it, asked for fixes and posted by hand.

  • Reel 1 failed five ways at once: identity in about 75 % of clips, a script that only did typography, one sound effect, music at -32 dB, dense scripts.
  • v3 locked the format the same day: one anchor photo, short seedance clips, text bridges, about 42 credits per Reel.
  • Nine Reels posted in month one: 17,042 views from zero followers.
Turning point. Reel 1 failed and the plan that replaced it died the same day; v3 made the text cards a first-class part of the format. we still play with solid layer and text animation remember. (Rido)
What if it works?I'm choosing the pressure again.Some sports are access.The future isn't found. It's chosen.
02June to Julyv3.26 to v3.49 · A67-A175

The Canon and the Walls

The same look, now built from a written Mira canon: one identity reference in every prompt, outfit cards, 28 named expressions. Claude started writing the scripts from my briefs. Then Reel 17 hit the wall that an earlier Reel had already explained.

  • July 17: a stranger's face four times out of four, 104 credits lost, a lesson Reel 2 had learned for 0.48.
  • Within three days: hard blocks, a character library and a spending protocol. No table in chat, no spend.
  • July 30: the repo moved into git; the first rules became code the next day.
Turning point. The identity collapse turned a style guide into hard blocks. why the face, fashion, and everything become so different? (Rido)
Be ridiculously optimistic about your life.Everyone plans how to win.What success actually meansYou're not tired. You're carrying.
03Augustv3.50 to v3.98 · A176-A319

Checkers and the Emitter

122 commits in 11 days. Whisper became the clock, one command emitted a whole After Effects project from a voice take, a synthesized sound kit replaced the library, and the gate grew to 25 stages with checkers proven able to fail. August 2 was the last post.

  • 15 forked editor scripts, only 19 % alike, became one golden master plus emitters.
  • Measured on August 16: 31 system versions, 2 Reels shipped. The pipeline contract rationed system work.
  • The first real After Effects run of the shared engine exported 14 of 30 seconds as black screen.
Turning point. The system session wrote down its own stopping condition, after I said this. i'm tired of this, i don't even know what we're doing, the loops never end. (Rido)
Restarting isn't free.You think they're just boring, right?A dark vertical frame titled two kinds of decisions, with a yellow U-turn arrow labelled reversible and a straight arrow running into a wall labelled one-way, and the line know which one you're in.A dark vertical frame with small capitals reading the real cost, a large white number 87%, and small text reading never get a second look.
04Aug 31 to Sep 6A320-A407 · 162 commits in 7 days

The Studio: Drawn Explainers, Luna, a Mouth That Matches

When I asked where the motion graphics were, the answer was a studio: a Remotion design system with a hand-drawn pen, a compiler that turns words into seconds, a story grammar, a second presenter, and on September 2 a mouth that finally says the script.

  • 78 commits in one day (September 1): the pen, the grammar, the compiler, face_check and Luna.
  • Lipsync stopped being a prompt problem: probe p 0.37 to 0.00, receipts or no render.
  • The first viewers found the minute-long cut boring: back to 30 seconds.
Turning point. My own four-clip test showed what makes a mouth accurate, and sync_so repainted it. the most important and essential aspect is to make sure the footage talk video is match with the VO. (Rido)
Good, better, best (Studio preview)Good, better, best with Luna (Studio preview)A dark vertical frame with a white hand-drawn box reading best 180 on the left, a yellow box reading better 85 on the right, a small stick figure between them and a yellow arrow pointing right.A dark vertical frame with a large yellow hand-drawn arch magnet and yellow hand-lettered words below reading it pulls them up.
05Sep 7 to Sep 24v3.99 to v4.13 · A408-A440

The Bridge Into After Effects

The agent could drive After Effects but not see it. So it built a panel inside AE that answers back. On September 24 an emitted Reel built itself inside my After Effects for the first time, and exposed a crash every gate had missed.

  • Mira's identity re-anchored: her own genesis prompt returned a stranger at 0.168; her raw anchor picture won.
  • A probe run in my After Effects showed the emitter could not set a single font.
  • Reference Reels measured frame by frame, behind guards, numbers only.
Turning point. Driving AE blind was worth almost nothing; a read-back channel changed that. you do end to end, perhaps from the background, while i do anything else. (Rido)
Mira on a city street at golden hour, traffic and buildings blurred behind her, in a cream cardigan over a white tee, looking up past the lens with sunlight in her hair.Three-quarter portrait of Mira against a grey wall in soft overcast light, in a grey ribbed knit, her hair in a low ponytail with platinum strands, a neutral expression.Open high: the 30-second template (Studio preview)
06Sep 25 to Oct 24A440c-A444

The Motion Seat and the World Grammar

A second seat opened: Claude Code on my Mac, for motion. It built 15 motion primitives that convert to native After Effects shapes, fixed time that was rounded in four places, ran the whole gate on the Mac, and rebuilt a measured reference grammar in our own ink.

  • 15 fx primitives, 11 of them native AE shape layers, proven against a strict offline mock.
  • The AE project now follows each board's own clock, cues and music fade.
  • A month of credits went unused: the output had not yet earned my trust.
Turning point. The seat opened with a short brief, and two weeks later produced a whole new grammar. be more creative - add animations, cool graphics that you create with svgs or whatever (Rido)
Fifteen motion primitives (Studio preview)The world grammar, paper (Studio preview)The world grammar, night (Studio preview)
07Oct 24A445 · one evening

The Reset

I watched the newest previews and rejected the look: unrelated metaphors, words that repeat the voice, colours off the Decyde brand, a hard cut out of Mira. That evening three art directions were built on the brand book and judged by blind reviewers. I picked A, the Schematic, with C's compare card as the proof.

  • The deliverable became the finished mp4 from our own renderer; After Effects turned optional.
  • The voice moved to the ElevenLabs API, behind my go.
  • By 22:32 one command built a Reel from a spec. By 05:52 the next morning the first Reel of the new lane was approved.
Turning point. After five months of building, I stopped before executing and asked for research and polish. we dont need after effect anymore if you can build your own version (Rido)
Direction A, Schematic (the one picked)Direction B, On the tableDirection C, Decision replay
08Oct 25 to Oct 29A445-1 to A445-50

The Rebuild: Nine Reels in Four Days

A gallery of approved Mira takes, a mouth bought instead of a clip, a design system built in one evening, a voice the AI picks, and a drawn stage for every idea. Nine Reels were approved in four days; in five of them the viewer plays along.

  • Cost per Reel: about 98 credits for new clips became 37 to 54 for the mouth alone.
  • Viewers' notes were measured before they were fixed: a blurry Mira, quiet music, pictures too small to notice.
  • October 29: a 50 to 60 second cinematic format prepared, and hand-coloured etchings approved as the picture style.
Turning point. I approved the redrawn two-trips and asked for the same art in every stage after it. be much more creative like how you make art (Rido)
Two reps. Same client. $30K apart.When a friend asks you for a favour, do you say yes right away?Wait a month for $10?Two trips. Same weekend.
06.1

Who Did Each Stage, Era by Era

Media stages that spend money stayed with me or behind my go the whole time. Assembly went from my hands to code in August and September. After the October rebuild the voice moved from my microphone to the API behind my go, and assembly became one command. Publishing never automated.

Stage01 May02 Jun-Jul03 Aug04 Studio05 Bridge06 Motion seat07-08 Rebuild
ResearchAI + me, behind my go: adapted from short videosAI + me, behind my go: a brief, then ClaudeAI + me, behind my go: course library draftsAI + me, behind my go: library + story gateNo change: no new topicNo change: no new topicAI + me, behind my go: Claude pitches three, I pick
ScriptBy hand: I write or trimAI + me, behind my go: Claude writes, I editAI + me, behind my go: anti-AI-tone rulesAI + me, behind my go: budget at her paceNo change: reusedNo change: reusedAI + me, behind my go: plain-language gate
VoiceBy hand: ElevenLabs appBy hand: ElevenLabs appBy hand: API closedBy hand: Laura and ClaireNo change: reused takesNo change: scratch voiceAI + me, behind my go: API, five takes, the AI picks
TimingAI + me, behind my go: Whisper + hand chunksAI + me, behind my go: Whisper + hand fixesAutomated: the voice is the clockAutomated: the compilerAutomated: karaoke on stampsAutomated: board clock into AEAutomated: ElevenLabs word timings
Presenter imagesAI + me, behind my go: one locked anchorAI + me, behind my go: canon + my approvalAI + me, behind my go: library + prompt gateAI + me, behind my go: looks + face_checkAI + me, behind my go: re-anchoredNo change: noneAI + me, behind my go: a gallery pick, or a new Mira by recipe
Presenter videoAI + me, behind my go: 4 s seedance clipsAI + me, behind my go: 4 to 8 s, fittedAI + me, behind my go: four-clip budgetAI + me, behind my go: from the voiceNo change: reusedNo change: noneAI + me, behind my go: seven reused takes
LipsyncNo change: mic over the mouthNo change: mic foamNo change: "unfixable"AI + me, behind my go: sync_so + receiptsNo change: reused clipsNo change: noneAI + me, behind my go: mouth only, behind my go
MotionAI + me, behind my go: type in each scriptAI + me, behind my go: phrase burstsAutomated: motion labAutomated: Remotion pen + castAI + me, behind my go: bespoke scenesAutomated: fx toolkit + worldAI + me, behind my go: a stage per Reel on one contract
After EffectsAI + me, behind my go: a script per Reel, I run itAI + me, behind my go: copy the last oneAutomated: emitterAutomated: board to AEAutomated: built via the bridgeAutomated: native shapes (offline)No change: optional: one command renders
Sound effectsAI + me, behind my go: library, per momentAI + me, behind my go: library, per momentAutomated: synthesized kitAutomated: bound to entrancesAutomated: measuredAutomated: the board's own cuesAutomated: registry, set by peak
MusicAI + me, behind my go: library at -7 dBAI + me, behind my go: levels measuredAutomated: resolved by nameAutomated: level from the boardAutomated: unchangedAutomated: fades before the closerAutomated: a CC0 bed shaped for phones
QAAI + me, behind my go: self-check in the scriptAI + me, behind my go: preflight echoAutomated: gate: 25 stagesAutomated: gate: 45 + receiptsAutomated: AE probe, guardsAutomated: 45 of 45 on the MacAutomated: goldens, layout probe, my watch
PublishingBy hand: I post on three platformsBy hand: I postBy hand: Aug 2: the last postNo change: noneNo change: noneNo change: noneBy hand: four apps by hand
By handAI + me, behind my goAutomated
07

Rebuilt in Five Days

On October 24 I rejected the look the system had grown into. Ten hours later the first Reel of a rebuilt system was approved, and by October 28 there were nine. This is what changed, and what it cost.

07.1

Five Days, Step by Step

Every step below is a commit in the repo, with my words recorded in its amendment. 104 commits from the reset to October 29.

  1. Oct 24 · 19:16 The reset (A445) I rejected the look and the hard cut. The deliverable became our renderer's mp4; the voice moved to the API behind my go.
  2. Oct 24 · 20:48 Three art directions Built on the brand book and scored by blind reviewers. I picked the Schematic.
  3. Oct 24 · 22:32 One command reel2.py built a whole Reel from a spec, free, on a stand-in voice.
  4. Oct 25 · 05:52 The first Reel approved Made end to end by our own renderer: the real voice, the mouth only, no After Effects.
  5. Oct 25 A gallery, a new Mira, a design system Approved takes to reuse; a new Mira from my outfit sheet; the design system's phases 0 to 4 in one evening.
  6. Oct 26 Three more Reels and a new voice Hook and topic rules from research; Sapphire becomes Mira's voice; viewers' notes measured and fixed.
  7. Oct 27 Play-along I asked for Reels the viewer plays along with. Three were approved that day.
  8. Oct 28 The money measured, the art redrawn Two more Reels; the real ElevenLabs bill audited; two-trips redrawn as a world.
  9. Oct 29 The next format A 50 to 60 second cinematic format prepared, and hand-coloured etchings approved.
07.2

One Reel, Three Art Directions

The same story told three ways on the Decyde brand book, built and judged in one evening. Two blind reviewers who never saw the script read all three correctly. C stated the lesson most clearly, A looked the most premium. I built A and borrowed C's compare card.

A Schematic, B On the table, C Decision replay. Tap to see it large.

07.3

Play Each Direction

The three tests, 24 seconds each, on a stand-in voice and an old take: 0 credits.

A, Schematic (picked)Preview

A, Schematic (Picked)

B, On the tablePreview

B, On the Table

C, Decision replayPreview

C, Decision Replay

07.4

One Lesson, Three Systems

The same negotiation lesson, the first number sets the deal, as each system told it: September's hand-drawn dry run on a stand-in, October 10's world grammar that I rejected, and October 25's final on the rebuilt system.

September: the hand-drawn dry runPreview

September: The Hand-Drawn Dry Run

October 10: the world grammarPreview

October 10: The World Grammar

October 25: the rebuildPosted

October 25: The Rebuild

One command

From a Spec to a Finished Reel

A Reel is a spec that names words, never seconds. reel2.py builds it: the voice, the word timings, the timeline, the render in Remotion, and a master to streaming loudness. The same command makes the free preview on a stand-in voice, so I see the whole Reel before anything is bought.

After Effects went from the deliverable to optional in one evening. The old code-built lane still passes its 45-stage gate; none of the nine new Reels needed it.

10h 36mfrom the reset to the first approved Reel
-14 LUFSevery master, true peak under -1.5 dBTP
0credits for a full preview
python3 jsx-rules/reel2.py check  wait-a-month   # names every fault before anything is bought
python3 jsx-rules/reel2.py build  wait-a-month   # voice, clock, render, master: one mp4
python3 jsx-rules/reel2.py post   wait-a-month   # only after my recorded approval

The three commands a Reel needs once its script and spec exist.

Free preview

Watch It Whole, Then Pay

Before any paid step, the same command builds the whole Reel on a stand-in voice and a gallery take, for 0 credits. Her mouth does not match yet and the voice is a placeholder, but the stage, the timing, the captions and the sound are all there.

I watch it and give notes, and only then say go to the three paid steps: the voice, the pictures and the mouth. Every credit is spent on a Reel I have already seen.

3paid steps, each after my go: voice, pictures, mouth
~41credits on a play-along Reel, all after the preview
The free preview, on a stand-in voice0 credits

The Free Preview, on a Stand-In Voice

The Reel as postedPosted

The Reel as Posted

every-yes: the preview of October 25, then the Reel I approved on October 26.

The gallery

Reuse the Take, Buy Only the Mouth

Making new talk clips for every Reel was the cost and the risk: the face could drift in any of them. Now an approved take is lip-synced onto any new script, priced per second of her talking, line 1 first so I can stop before the rest.

A take enters the gallery only with my words, a face match over 0.50 and steady framing. Seven takes carried nine Reels; a new Mira is made only when I ask, and its recipe fell from 141 to 54 credits in three days.

~98 to ~41credits per Reel: new clips, then the mouth only
3.6credits per second of lip-sync, quoted to the hundredth
7 to 9approved takes to Reels
Before: the approved takePlate

Before: The Approved Take

After: synced and sharpenedSynced

After: Synced and Sharpened

One line of two-trips: the same 3.7 seconds of take, before and after its mouth was synced to Sapphire's voice.

The voice

Five Takes, and the AI Picks

The voice used to be me in the ElevenLabs app. Now it is the API, behind one go a Reel: five takes on one budget, a word-exact transcript check, and a score calibrated on the only two picks I ever made. It reproduces both.

Then I audited the bill. The account's own history showed ElevenLabs charges 0.40 of the characters our scripts count, the same ratio four times out of four. The voice budget turned out to last far longer than we planned for.

5takes per go; Claude picks one
0.40billed per character counted, measured four times
1voice for every Reel, Sapphire, picked by my ear
The first Reel in SapphireOct 26

The First Reel in Sapphire

own-advice was voiced twice: in Laura, then in Sapphire after I called the first take flat.

The design system

Built in One Evening, Without Moving a Pixel

I asked for a design system rich enough that every next Reel is easy, consistent and on brand. It came in five phases: tokens as roles, a motion language, 31 primitives in nine families, 16 components and one stage contract.

Each phase had to leave the approved Reel identical: 29 golden stills compared after every step, and timelines rebuilt byte for byte. Text is measured from the font files themselves, within 0.5 % of the browser, and a probe checks every line against Instagram's own buttons.

31primitives, each with a meaning and four test moments
13maths proofs in the law checker
29golden stills guarding the approved Reel

The primitives reel: fourteen pieces at work, each named. 23 seconds, rendered in about 25.

Checks you can see

Every Check Leaves a Picture

I judge by eye, so every check leaves something to look at. Golden stills guard the approved Reel: change the renderer, move one pixel of these moments, and the check fails. The mouth sheet shows her lips every eighth of a second with the word under each frame; when the lip-sync checker cannot judge, the sheet decides.

The numbers run first: a law checker with 13 maths proofs, the plain-voice and word-exact voice gates, a layout probe against Instagram's own buttons, and blind reviewers on the frames. Nothing reaches me before they pass.

29golden stills of the approved first Reel
1/8 sbetween frames on a mouth sheet
Golden stills The approved first Reel, moment by moment.
Mouth sheet two-trips, line 1: you booked two trips for the same weekend.
Feedback, measured

Viewers Said It. I Measured It, Then Fixed It

People called Mira blurry: the renderer added nothing, the 720p clip did, so a free Apple super-resolution pass runs after the sync instead of paying 2.7 times for 1080p. People could not hear the music: through a phone it sat 16.5 LU under her voice, now 8.9. People missed the pictures: 280 pixels in a corner for 0.8 seconds became 940 by 671 for 1.8.

The first Reel's audience found it too niche, so the scripts went plain. Openings felt flat, so four new openings wait for my pick.

16.5 to 8.9LU between voice and music on a phone
280 to 940pixels wide for a cut-in picture
0credits for the sharpening fix
Before: pictures in a corner for 0.8 sPreview

Before: Pictures in a Corner for 0.8 s

After: big, under her captions, for 1.8 sPosted

After: Big, Under Her Captions, for 1.8 s

scroll-hour: the free preview under the first picture rule, then the Reel as posted after the fixes.

The closing

Every Reel Ends on the Logo Drawing Itself

I asked for an animated closing for Decyde, and to be creative. Now the blue Decision Marker that has moved through every stage of the Reel draws the Y of the logo stroke by stroke, the name writes itself, and the line Practice business decisions before they matter appears. 2.4 seconds, in every Reel.

every-yes, the last 2.5 seconds.

07.6

The Design System, One Sheet

Eight families of primitives on one contract. Each shows four moments and the one that lands on the word, in blue. Scroll sideways.

A wide catalogue sheet of eight panels, each a family of small animated diagram pieces shown at four moments, with the landing moment framed in blue.
WideThe primitives sheet, scroll sideways
07.7

The Catalogue, Page by Page

Every page is drawn at phone size from the tokens themselves, so what you see here is what a Reel gets. I told Claude to follow its own recommendations for these pages; the camera and the sound they encode are the ones I approved by watching.

07.8

The Motion Pages, Moving

Easing by purpose, a camera that glides from rest and never overshoots, and two timing tiers with one pen speed.

DS 06 The camera A glide from rest, beside a default spring that overshoots.
DS 05 Easing Every move eased by what it is for.
DS 07 Timing Two tiers of timing, one pen speed.
07.9

Every Reel Gets Its Own Stage

Nine stages behind one contract: a new stage is a folder, one line in a registry and its hooks. In the play-along ones a timer asks you to pick before Mira answers. Two frames from each final.

07.10

Motion With a Cause

Nothing moves for decoration. The walls rise because it is your own problem, the trip you would love more comes alive while you pick, each yes takes a slice of the day, and the calendar has to unroll a week nobody planned for. Cut from the posted Reels, silent.

own-advice Inside your own problem the walls rise; from outside they are lines again.
two-trips The $50 trip comes alive while you pick.
every-yes Each yes takes a slice of Dina's Saturday.
worst-case The guesses break and an eighth week unrolls.
07.11

How a Stage Is Found

Stages are not drawn once. every-yes went through three rounds on my notes: a price rail that could not tell a story about time, a time bar, then a Saturday as a ring.

Round one The rail, with every fault written on the frame.
Round three The day as a ring: each yes takes a slice.
07.12

Words Off Her Face

I asked twice: the captions, then the title, below her face. Since own-advice the title starts under her neck, and the layout probe checks it on every take.

Mira in a black top in a night room, the title When a friend asks you for a favour, do you say yes right away? across the top over her head.
Over her head every-yes, October 26.
Mira in a white top and a pale blue shirt, the title Great advice, just not to yourself across her chest below her face.
Under her neck own-advice, the same day.
07.13

Two Picture Fixes, Before and After

Small things a viewer feels before they can name them.

Bright Enough to See

A dark painting of a smiling scholar pointing at a page, towers of papers and a candle around him.
35 of 255
The same painting of the scholar, brighter, the papers and the candle easier to see.
About 65

Dark pictures vanished on her black top. Every picture now reads at a mean of 50 or more.

One World, Not Two

A painting in an oil style of a laughing woman in a pink dress riding a wooden sled down a sunny snowy hill.
Oil painting
A hand-coloured etching of a woman in a pink dress on a sled in the snow, fine engraved lines and soft colour, mountains behind.
Etching

Oil paintings sat outside the drawn world. Hand-coloured etchings, approved October 29, belong to it.

07.14

The Pictures Inside the Reels

Between her lines, a picture lands under the caption: my own AI-generated scenes in a classical oil style with a modern twist, 0.25 credits each. The first three show one in place. Since October 29 every new picture is a hand-coloured etching instead, so the drawing and the picture are one world.

07.15

What Each Reel Cost

Measured from the account's own records. Higgsfield credits are the mouth, the pictures and, twice, a new Mira. ElevenLabs characters are what was billed where it was measured.

ReelApprovedHiggsfield creditsMade ofVoice
first-numberOct 2538.4the mouth on a reused takeLaura
every-yesOct 2654.28the mouthLaura
scroll-hourOct 2645.5five pictures, the mouthLaura
own-adviceOct 2691.16the mouth twice: Laura, then SapphireSapphire
wait-a-monthOct 2747.74pictures, the mouthSapphire
liked-you-moreOct 2795.44a new Mira (54), pictures, the mouthSapphire
nobody-noticedOct 2737.39pictures, the mouthSapphire (865 characters billed)
worst-caseOct 2839.79pictures, the mouthSapphire (895 billed)
two-tripsOct 2839.5pictures, the mouth; the redraw cost 0Sapphire (795 billed)
07.16

The Cost, in One Line Each

~41Higgsfield credits on a play-along Reel (37.39 to 47.74)
~$2of AI generation per Reel, from about $4 to $10 before
54credits for a new Mira by October 27, from 141 on October 25
~24Reels a month of credits now covers, from about 5 to 11
07.17

What I Still Do by Hand

PICK

The Topic and the Mira

From three researched pitches, and from one picture of the gallery takes.

GO

Three Paid Steps

The voice, the pictures, the mouth: each prints its exact text and price and waits for my go.

WATCH

Every Reel, With Sound

A Reel is done only when I have watched it and said it is good. My words go into its spec.

POST

Four Apps

From one folder, by hand. I was offered automatic posting and said no.

08

The Hard Problems

Generating a clip was never the hard part. Making a series hold together was: the same face, a mouth on the right word, motion that means something, an editor that builds itself, and AI sessions with no memory that behave like one careful engineer.

01 · Identity engineering

Words Fix the Type. Only a Picture Fixes the Person.

An AI presenter is only useful if she is the same person in every clip. Reel 1 held Mira's identity in about 75 % of clips, and in only about 42 % across three.

So her face got a number. face_check.py embeds the face with ArcFace and compares it with 21 faces I approved: 0.50 and up is her, 0.40 to 0.50 goes to my eye, below 0.40 is a stranger. The thresholds were calibrated on my own verdicts, not on the tool.

The measurements settled the method. On September 8 her original prompt, run again word for word on the same model, brought back every described detail on a different woman: 0.168. Her original picture with a short prompt brought her back: 0.627. The seed of that first portrait is gone; only the picture carries her face.

0.168her description alone
0.627a short prompt + her picture
21approved faces in her gallery
Labelled failure: a different woman with Mira's described hair, cardigan and gold chains, generated from Mira's original prompt with no picture.
Mira, straight-on with a small closed smile, a brass desk lamp glowing behind her, generated from a short prompt plus her original picture.
Words only: 0.168Her picture: 0.627

Drag the handle. Left: her original prompt, words only, on the same model. Right: a short prompt plus her original picture. Both are AI-generated; the left one is not Mira.

02 · Never deliver a stranger

104 Credits for a Lesson That Cost 0.48

July 17, Reel 17. A prompt enhancer quietly turned a New York brief into seven selfies in a Seoul apartment. A reference-only model then returned a stranger four times out of four, and 84 credits of video animated her.

Reel 2 had hit the same wall two months earlier and solved it for 0.48 credits. The finding never left its project folder. The postmortem's motto became the house rule for money: halting costs 0, improvising cost 104.

Two laws came out of it: findings are promoted to the rulebook the same day (A87), and every identity image must carry a trained Soul or her picture as a reference, never neither (A352), checked before a credit is spent.

4 / 4strangers from one model
217xthe cost of forgetting
0credits to halt and ask
Labelled failure. A laughing woman with long dark hair spinning on a gravel rooftop at sunset with her arms out, in a cream cardigan, black skirt and sneakers; the face does not match Mira.
What the model delivered The right scene, the wrong woman: a reference-only model missed her four times out of four.
Mira laughing mid-spin on a gravel rooftop between tall brick buildings at sunset, hair flying, in a black ruched cold-shoulder crop top and black cargo trousers with gold chains; a phone stands on the ledge recording her.
After the canon The same rooftop spin, rebuilt in an approved outfit and checked by eye before any video credit.

Reel 17, July 17. Both images are AI-generated. The one on the left is not Mira, which is the point.

03 · The fixes, measured

Every Fix Measured, Every Verdict Mine

Identity work became a loop: I judge a frame, the session measures why it failed, and the fix becomes a rule that runs before the next credit is spent.

Three of those fixes. Her makeup had never been written into a prompt, so the model invented a harsher face until the anchor's blush, lashes, lips and mole were transcribed word for word. The word mid-sentence in a still prompt rendered a grimace, so a still is now a composed portrait a breath before speaking. And a copy of a copy loses 0.02 to 0.03 of identity per generation, so every frame starts from the best original.

When the checker and I disagreed, I won: it called one frame a stranger at 0.435, I said it was her, and the threshold moved to 0.40.

0.122text only: a stranger
0.518one hop from the best original
0.40the threshold my eye set

The Makeup Clause

Labelled failure. A woman meant to be Mira, with a harder jaw, thinner lips and a toothy open mouth, holding a peace sign by her face in front of a rainy city window, black hair in a low ponytail with platinum strands.
Not Mira, my verdict
Close crop of Mira's face by a rainy window: rosy blush, glossy closed coral lips, the small mole beside her nose, gold ear cuffs and hoops, black hair with platinum strands pulled back into a low ponytail.
Mira, the same day

Her makeup was never in the prompt, so the model invented a face.

A Still Is a Portrait

Labelled failure. Mira in a cream blazer worn as a cape over a cream corset, hair in a braid, caught with her mouth open wide and teeth showing in a strained talking expression, holding a small clip-on mic.
Mid-sentence in the prompt
Mira in a cream blazer draped as a cape over a cream corset, a bubble braid tied with gold cords and platinum woven through, glossy closed coral lips, a rosy blush and the small mole beside her nose, holding a tiny black lavalier mic, sheer curtains behind.
Composed, lips closed

One word made a grimace. The video stage opens the mouth, not the still.

Never Copy a Copy

Labelled failure. A different woman with heavier brows and a fuller face, black hair with pale ends, in a black blazer over a black top, holding a large black foam microphone, in a warm room with a rust armchair and bookshelf.
Text only: 0.122
Mira in a warm loft with sheer curtains, a bookshelf, trailing plants and a rust armchair, a black blazer slipping off one shoulder over a ruched black top, fine gold chains and pearl drop earrings, holding a small cordless black lavalier at her chin, lips closed.
From the best original: 0.518

Identity decays with every copy, so frames start from the best original.

Every image here is AI-generated. In each column the top image is a labelled failure.

04 · The voice is the clock

Nobody Types a Second

Reel 21 was built on one voice take and shipped with another: picture and voice drifted up to 5.1 seconds apart. The fix was to stop timing anything by hand.

Now a spec names words. faster-whisper stamps every word of the take, and the compiler places each beat on its phrase: @two fifty cuts on the exact frame she says it. It refuses a phrase spoken twice (a coin flip), a phrase never spoken, and cues that run backwards. Re-record the voice, recompile, and every cut, caption and sound effect moves with it.

The output is deterministic: the same spec and the same words give a byte-identical board, which is what makes regression testing possible.

5.1 sdrift before the compiler
0seconds typed by hand
0.07 ssound leads the picture
// studio/specs/the-anchor-dryrun.json, trimmed
// a beat names WORDS from the take, never seconds
"point": {
  "shape":   "denial",
  "obvious": "open high, so there is room to come down",
  "real":    "open FIRST with the strongest number"
},
"carrier": "magnet",
"beats": [
  { "role": "hook",    "cue": 0,
    "say": "you know that first number in a negotiation?" },
  { "role": "obvious", "cue": "@everyone says",
    "scene": "illustration" },
  { "role": "denial+real", "cue": "@two fifty",
    "say": "two fifty doesn't buy room, it blows the deal up." },
  { "role": "proof",   "cue": "@floor seventy -0.1",
    "scene": "compare" },
  { "role": "verdict", "cue": "@it was never",
    "say": "it was never about going high, it's about going first." }
]

A real spec, trimmed. The compiler turns it into a board with real seconds, captions, sound cues and a camera move for every frame.

05 · Lipsync without reshoots

The Mouth That Was Unfixable

In May the microphone covered her mouth to hide bad sync. By August the mouth was written off as unfixable, backed by a correlation that looked like signal until a permutation test showed 27 of 35 random offsets beat the true one.

Then one controlled test changed everything: four clips from the same frame, differing in one input each. Words only: accurate. Audio only: following but not accurate. Both: the best. An old ban on putting the script in the prompt had caused the fault, and a blind instrument had certified it as unfixable.

Today the talk clip is generated from the voice, and a second pass repaints only the mouth. A receipt proves every clip: the sync probe went from p = 0.37 to p = 0.00, and 4 of 5 lip closures land within 40 ms.

p 0.00sync probe after the mouth pass
4 / 5p, b, m closures within 40 ms
~87credits per Reel, mouth only
The first Reel with the mouth repainted to the voice (Studio preview, sound on)Preview

The First Reel With the Mouth Repainted to the Voice (Studio Preview, Sound On)

06 · A hand-drawn motion system

Movement Without a Reason Is a Fault

The first Reels used black kinetic text cards. I wanted the warmth of hand drawing without drawing by hand, so the motion became a system in code: a doodle library drawn stroke by stroke, a two-puppet cast (Dot believes the obvious, Stick is sceptical), 15 motion primitives, and one yellow accent per scene.

When I complained about shaking, the complaint became measurable laws: one source of idle life (a 3 fps boil of 0.6 px), a shake ceiling, no frozen holds, and a beat never opens on an empty frame. A drawing checker found that the bridge read as a hammock and an arrowhead had floated 100 px off its arc for two months.

Same Reel, two systems: on the left, Reel 31 as it shipped with plain kinetic text; on the right, the same voice remastered with the motion kit. On October 24 I rejected this hand-drawn look as off brand. The laws and the checkers carry over to whatever replaces it.

15motion primitives, 11 native to AE
78doodle drawings, each machine-judged
1yellow accent per scene
Reel 31 as shippedBefore

Reel 31 as Shipped

Remastered with the motion kitAfter

Remastered With the Motion Kit

Both play with sound. Nothing loads until you press play.

07 · After Effects, written by code

An Editor That Builds Itself

By August there were 15 hand-forked After Effects scripts, 24,142 lines, only 19 % alike from one Reel to the next. Now one golden master (the last script that rendered) gets this Reel's data swapped in by an emitter: blocks, word stamps, clip slots, camera, the drawn layers, editable text, sound and the end card.

The agent could drive After Effects but could not see inside it. So Claude built a bridge: a tiny panel inside AE that runs a local web server. The agent posts a script, AE builds the Reel, and the result comes back as the answer.

Its first run exposed a bug every gate had missed: every emitted board since early September died at line 2332, because the gates parsed the script and none executed it. Two static guards now refuse to emit that class of bug.

2.7 sto build a Reel headless
37layers, 0 missing files
9editable text layers
# the agent asks After Effects to build the Reel (localhost only)
$ curl -d "$PWD/projects/<reel>/generate-after-effects.jsx" \
       http://127.0.0.1:8723/runfile

# what came back on the first run, September 24, 2026
comp        27.47 s, 720x1280, 30 fps
layers      37   (9 text layers in Superbusy Activity)
footage     0 missing
project     saved by the script itself
render      H.264 + AAC, 15.2 Mbps

The bridge's request and the first real result, from the project log. The deliverable is the After Effects project, not an mp4.

08 · The gate

A Checker Earns Trust by Proving It Can Fail

gate.py runs 45 rule stages plus up to 12 per project, in spend order. Every checker that matters carries a self-test that plants the breakage and must go red. A result that only ever comes out positive is not a measurement; it is formatting.

The gate still lied once. Four motion laws written for my shaking complaint were never invoked: the gate called the checker without its frames flag and printed all clear. Since then a law that never runs counts as a failure, and the summary says NOT MEASURED when a stamp is stale.

And the final check is human. Claude must open the picture before describing it, and done means I watched the render.

45 + 12gate stages
47self-tests that must go red
224 / 372laws a checker enforces
# read-only runs, October 24, 2026 (summarised)
$ python3 jsx-rules/prompt_check_selftest.py
  53 seeded faults fire, 12 controls stay green
$ python3 jsx-rules/story_check.py --selftest
  known-good spine passes, 15 seeded faults red
$ python3 jsx-rules/svg_ae_check.py --selftest
  real recipes pass, 13 seeded faults red
$ python3 jsx-rules/rule_coverage.py
  224 of 372 rules enforced, 148 prose only, 18 retired

Self-tests plant the bug on purpose. A checker that cannot fail is not trusted.

09 · Working with AI agents

Every Failure Becomes a Law

I am not a developer. The system code, 56,573 lines by October 24 and about 77,000 after the rebuild, were written by Claude agents in two seats: Cowork in a Linux sandbox, and Claude Code on my Mac. My job was the part agents cannot do: decide what good looks like, say no, spend money, and watch.

Short sessions forget. So the memory lives in the repo: a numbered rulebook (A1 to A445), each amendment starting from my own words, with a trigger, a measurement, the rule and what is not proved. Laws that can recur become checkers. Permissions are code: secrets and destructive git commands are denied, and a hook blocks em dashes, my standing style rule.

Agents propose, I decide, and the decision is written down. When the system work outran the Reels (31 versions, 2 Reels), the fix was a contract: at most one system change per Reel shipped.

445numbered laws
363commits in 34 active days
735Kwords of docs the agents keep
the loop

  I reject something (in my words, verbatim)
    -> the session measures the failure, never argues it
    -> a numbered amendment: trigger, measurement, rule, what is NOT proved
    -> if it can happen again: a checker, wired into the gate
    -> the checker's self-test plants the breakage and must go red
    -> the gate runs it on every build
    -> I watch the result; my verdict outranks every score

How a complaint becomes a guarantee. On September 6, five complaints in one message became five measured laws in one day, at zero credits.

09

What the Numbers Said

A cold start: no followers, no ads, no friends reposting. The feed found the Reels, the hook decided who stayed, and every lesson went back into the law. After the rebuild the views dropped, and I treated that as data too.

38.6Kpublic views on YouTube Shorts and TikTok (October 29, 2026)
17,042views in month one, from zero followers
90.2 %best retention, "I'm choosing the pressure again."
96.5 %of month-one YouTube views came from the Shorts feed
237 to 294YouTube views of each rebuilt Reel in its first days

YouTube Shorts, Views per Short

Public counts on October 29, 2026. The four marked new are the rebuilt Reels, one to four days old.

Trend Reel: "let's do this one more time"2.4K
Some days are heavy. That's normal.2K
everyone plans how to win.1.3K
Trust isn't built by helping people.1.3K
Great leaders don't just tell people what to do.1.3K
What if it works?1.2K
be ridiculously optimistic about your life1.1K
I'm choosing the pressure again.1.1K
Listen. Don't defend.901
You don't lack time.653
You're not tired. You're carrying.431
The future isn't found. It's chosen.354
New: why the first number in a negotiation usually wins.294
what success actually means260
New: why your advice works for friends but not for you252
New: why saying yes to everything costs you249
New: why five more minutes turns into an hour237
Some sports are access.125

Month One, Retention by Hook

How much of each Short viewers watched, YouTube Studio, May 15 to June 13, 2026.

I'm choosing the pressure again.90.2 %
Some days are heavy. That's normal.68.0 %
Some sports are access.65.6 %
What if it works?57.3 %
You don't lack time.55.0 %
Trust isn't built by helping people.47.2 %
Listen. Don't defend.46.5 %
Great leaders don't just tell people what to do.35.2 %

Month One by Platform

Nine Reels cross-posted to all three, May 15 to June 13, 2026, from each platform's own analytics.

YouTube Shorts8,178
TikTokabout 7,623
Instagram1,241
09.1

Honest Learnings

HOOKS

The Hook Decides

Recognition hooks won on YouTube ("Some days are heavy. That's normal." kept 68 %). A corporate hook kept 35.2 %, the worst of the month. Both became law: three hook patterns (A78) and a ban on corporate words (A79).

PLATFORMS

Same Video, 18x Apart

One Reel drew 2,561 views on TikTok and 144 on Instagram. Personal declarations travel on TikTok; recognition travels on YouTube.

FOCUS

The System Outran the Output

By August 16 the repo had 31 system versions and 2 shipped Reels. The fix was a rule, not a tool: at most one system change per Reel shipped, and a written definition of done.

COST

Cheaper Every Month

A Reel's AI generation fell from about $10 when it made its own clips, to about $4 when it reused a look, to about $2 since the rebuild: only Mira's mouth is bought, on an approved take.

QUALITY

My Eye Beat the Gate

Every regression I hit (a font patch, music 8 dB off, a starved caption) was caught by watching while the gate read green. So the system got a memory of good, and a law: a green gate is not a verdict.

AUDIENCE

Views, Not yet Followers

The feed brought the views; almost nobody found Decyde by search, and few viewers followed. Turning reach into an audience is the open problem, and why the Reels moved to Decyde's own course material.

THE DROP

A New Look, Fewer Views

The four rebuilt Reels drew 237 to 294 views each in their first days, fewer than the old ones. The likely causes are named, not yet measured: several posts close together, a new look, slow openings. So the next Reels are the test, and each one records its skip rate and watch time.

10

Luna, and the Long Format

On October 29 I asked for Reels of 50 to 60 seconds that feel cinematic. The first pieces are built and free: a camera that dollies through an engraved world, pictures as hand-coloured etchings, four ways to open. Luna, the second presenter, is ready when a Reel needs her.

10.1

Meet Luna

Luna is the second AI presenter, made with the same identity method as Mira: an approved anchor, a face gallery, her own voice (Claire, measured at 2.26 words per second) and her own wardrobe.

Her first look was built once, three stills and three 8-second plates for 152.5 credits, and every Luna Reel reuses it. Her first Reel bought only its mouth: about 32 credits.

0.60face score of her best still
152.5credits for a look every Reel reuses
~32credits for her first Reel
2.26words per second, measured
Luna's first Reel, mouth synced to her voice (Studio preview, sound on)Preview

Luna's First Reel, Mouth Synced to Her Voice (Studio Preview, Sound On)

10.2

What Happens Next

LONG FORMAT

50 to 60 Seconds, One World

Mira's talk keeps its length and cost; the drawn part grows into a world the camera dollies through, with the pictures inside it as prints and windows. The depth test cost 0 credits.

ETCHINGS

One World for Drawing and Pictures

Approved October 29: every new picture is a hand-coloured etching graded to our paper. The test cost 0.25 credits.

OPENINGS

The First Three Seconds Must Move

Viewers said the openings felt flat. Four openings built from one Reel's own parts wait for my pick: as posted, the world first, a snap, the painting first.

TOPICS

Beyond Decisions

Fun facts, myths and slow news, each fact-checked with a safe line, to find an audience first. Every Reel becomes a test of skip rate, watch time and sends.

11

What It Runs On

AGENTS

Claude, in Two Seats

  • Claude Code (the motion seat, on my Mac)
  • Claude Cowork (a Linux sandbox)
  • CLAUDE.md and path-scoped rules
  • permission deny lists
  • a hook that blocks em dashes
  • MCP connector to Higgsfield
  • git: 464 commits
  • a paste-in handover for every fresh session
GENERATION

AI Models

  • Higgsfield nano_banana_2 (stills)
  • Higgsfield Soul (casting, Luna's looks)
  • seedance 2.0 (talk plates)
  • sync_so, Sync Lipsync 3 (mouth)
  • ElevenLabs Eleven v3 (voice)
  • ElevenLabs API: Laura, then Sapphire
  • Higgsfield gpt_image_2_5 (pictures)
PERCEPTION

Models That Measure

  • faster-whisper base.en (word timing)
  • InsightFace ArcFace (identity)
  • OpenCV (mouth and lip tracking)
  • CMUdict (visemes)
  • MediaPipe (skin finishing)
MOTION

Drawing and Motion

  • Remotion 4
  • React 19 and TypeScript
  • perfect-freehand
  • roughjs
  • design tokens
  • Superbusy Activity and Inter
  • the v2 design system: 31 primitives, 16 components
  • Instrument Serif, Geist and Lucide icons
EDIT

After Effects

  • After Effects 2026 (optional since October)
  • ExtendScript (ES3, ASCII)
  • a CEP bridge (Node inside AE)
  • strict offline AE mocks
  • golden master plus emitters
QA

Quality and Sound

  • Python: 91 tools, 37,287 lines
  • gate.py: 45 stages
  • 60 self-tests
  • regress.py (known good)
  • ffmpeg loudnorm
  • a synthesized sound kit
  • golden stills and a layout probe
  • voice_check (plain words)
SHIPPING

One Command, One Folder

  • reel2.py build: render and master
  • reel2.py post
  • -14 LUFS, -1.5 dBTP masters
  • Apple super-resolution (free)
  • four apps, posted by hand

Keep browsing

More AI Engineering

All projects
  • OCBC LIFRR SimulationA live decision simulation for OCBC's graduate trainees: one facilitator, one room screen, a phone in every hand.
  • Nod - Said, Checked, SignedA builder says a job change into his phone. AI drafts it, state rules check it, and the client signs before the work starts.
  • Mini Business SimA pixel-art business game where learners run a notebook shop for 90 days and find out why growth is more than revenue.