The Super Intelligence brand: strategy brief for Jason

The date: September 29, 2026

“The government made that change” is Executive Order 14434, “Inaugurating the Era of Super Intelligence.” President Trump signed it on Tuesday, September 29, 2026.

Date forms for the brand: 9.29.26 · 09.29.2026 · SEP 29 2026 · “EO 14434” · “Day One of the SI era”. Year One runs to September 29, 2027.

Sources: the Federal Register, whitehouse.gov and GPO govinfo (bill texts and status).

Angus Sewell: the creator we model, with our spin

Angus Sewell (@angus.sewell, “Angus the Nontechnical”) explains AI for business and works as a consultant. He's the main model for the short-form side. We take his stories, angles and pace, not his style, and add what his posts mostly lack: the receipts.

Figures come from our October 3, 2026 study of 264 of his public posts (private dossier, available on request); they describe that sample. Our line: some of his growth recipes rely on personal-data scraping. We don't copy those. Business data only.

Names: concept families first

Ten families of ideas with 55 names. React to the concepts first; then the shortlist and the top 3 below.

Shortlist: are the names free?

Checked Saturday, November 14, 2026, from public pages only; nothing was claimed. ✓ means no public profile found or the .com is unregistered. “Free” on a social site still has to be confirmed at signup. Trademark means live USPTO marks in the classes for media, software and clothing.

Top 3 directions

Brand kits: compare the three

Full kits for the top 3 names: wordmarks, logo marks, palette and type, YouTube, Instagram and TikTok images, a 9:16 cover template, and tee, hat and sticker mockups built around 9.29.26.

• Year One SI recommended: a logbook, kept as the public record. Ledger paper, oxblood, a tally-stroke “Y1”.
• Fork & Ship: a shipping dock meets a code commit graph. Kraft, carbon and signal orange; a fork that turns into a shipping arrow.
• Ringmaster SI: a letterpress show bill. Show red, cream and brass; a center-ring mark. 2 flags on the kit page

Open the kits side by side →

Each kit has its own notes box on that page, with its own “Copy all my notes”.

Content: two formats, with your spin

One format is daily SI news shorts; the other is agents actually earning money or doing real demos with popular repos. Both work under any of the names.

Format A: SI news shorts

One SI story a day, 45 to 75 seconds, for Instagram, TikTok, Shorts and X. Every claim gets an on-screen verdict: legit or hype.

We take: their topics, hook shapes and Angus's pace, plus his finding that a re-angled story with a working build beats being first. We change: proof on every post, faster sourcing, no comment-for-a-link bait, and clear labels on paid posts.

Format B: agents at work your spin

“The spin I want to do on the repo round-ups is agents producing income or actually doing demos.” Other creators list repos you could “run your business on”. We give those repos to our agents with a real job: make money with it, or do a real demo of what it promises. “Does it run?” is just the first gate.

• The income run: “make money with this repo”. The scoreboard is income, cost and time. In the sandbox it's simulated or paper money only; real accounts or money need approval first.
• The real demo: “do what the README promises, on a real task”, such as a finished Short, an SEO audit of a real public site, or a working sales pipeline. Did a stranger get something they'd pay for?

The stamps: IT EARNS IT DELIVERS RUNS, DOESN'T PAY DOESN'T RUN. The same stamps work on merch.

Experiments to run first (from the Angus study)

1. Story radar: a daily scan of X launches, Hacker News, GitHub trending and arXiv. Backtested against Angus's and seb.ai's posts: how many days earlier would we have seen each story? Free.
2. Fresh vs old-with-a-build: 20 posts alternating “fresh launch within 24 hours” and “resurfaced story plus a working build”, same format both ways. This settles speed vs angle for us.
3. Claim-check scorecard: his legit-or-hype tallies, but every verdict backed by a run we did, with the cost shown.
4. Build-it-ourselves, for real: his best series, made with builds we've actually shipped, showing the real output, not just the prompt.

Two more from the study live in the Workflows card: company-level lead routing (business data only), and a written podcast-pitch desk.

The bigger version

These stand on their own; they don't depend on the League.

• V1, Test → Fork → Ship strongest: agents work a repo on camera, it gets a public score, and the winners get forked into our own branded tool, which ships on the show. Next season our tool goes back on the bench against rivals. Viewers vote on the next repo.
• V2, Replace the Incumbent: each season targets one expensive tool (Salesforce, Semrush, a Bloomberg Terminal) and builds the free replacement, with a running “you'd save $X a year” counter.
• V3, the Public Bench: a standing website that scores every repo we test, updated weekly. The index becomes an asset in its own right.
• V4, agents run a real micro-business: live profit-and-loss. It's the most gripping version, but it needs real accounts and money, so it needs approval and comes later.

Our own SI news show

“We probably could do our own AI news show too.” A recurring, faceless show fed every day by our story radar, where every verdict comes from a run we did.

Segments

1. Radar Top 5: the week's highest-ranked stories from our radar, each with when we first saw it (“we had this 4 days before the big accounts”, when true).
2. Legit or Hype? One viral claim, with the verdict coming from a run we did, not a quote.
3. We Built It: the week's agents-at-work run (income or a real demo), ending on “want one for your business? apply”.
4. SI Policy Watch: EO 14434, the Saturday, November 28, 2026 definition deadline, the two ban bills, and new Federal Register items, from primary sources only.
5. Day counter: “Day N of the SI era”, plus the next date to watch.

How it's made, with no camera

Radar digest → an agent proposes 3 stories a day → a person picks → a script agent writes it (every claim sourced) → a fact-check agent verifies against primary sources → voice-over (yours on a mic, or a licensed AI voice) → visuals in hyperframes (stamps, headline cards, day counter) plus screen recordings of the agents → captions → you approve and post. Agents never post.

Starting at 3 a week is our call: Angus's numbers say a verified build beats volume. The radar is being built by a separate agent.

Creator remix: one piece from each

Take one thing from each creator you save, and combine them into something none of them does. Five more of your top creators are being studied in depth right now, so their entries will sharpen.

Named here: creators with 100K+ followers. Smaller accounts are described, not named.

Creators to study

Ranked by how often you saved their posts. Saves undercount creators you like but rarely save, so add names in the notes box below as you find them; they'll be studied in the next batch.

1. marc.kaz: open-source repos and tools being studied
2. An open-source-alternatives account being studied
3. Angus Sewell studied
4. A money-loophole account being studied
5. A side-hustle deals account
6. godofprompt: prompts and AI tips being studied
7. realrobertgutierrez: AI business news and reports being studied
8. theaifield: AI news
9. A day-trading news account
10. sina.growthtech: agents running social pages
11. …then aitoolhub.co, theartificialintelligence, edward.builds, seb.ai studied, a research-repo spotter, a public-records lead account studied, and more.

The full ranked list, with names, is in the private notes.

Build challenge episodes

A cooking-competition format for agents: a mission, random constraint cards and a clock. Our agent teams compete, we score them, and the real run logs become the story. A recurring segment, roughly monthly to start.

• Constraint cards drawn on camera: tool, data, interface, forbidden, twist. The budget card always says $0.
• Teams for one episode only: 2 or 3 agent teams on different stacks (a frontier model, a cheaper one, a solo agent), in a sandbox with no keys and no spend.
• Judging: it has to run first; then mission met, cards obeyed, time, human interventions (aim for zero), cost and demo quality. A judge agent scores and a person makes the call.
• The story: a story agent turns the logs (plans, errors, fixes, the final sprint) into the arc, and the taste-trained editor cuts it. Real events, nothing invented.

Three starter episodes (about $0)

1. The 60-Minute Replacement: a free booking tool for a made-up bakery. Cards: one HTML file, works offline, and at 30 minutes the client adds a waitlist. Can a stranger book in under a minute?
2. Make It Run: a popular starred repo that's stopped working, brought back to life. Cards: no new dependencies, a local model only, 45 minutes.
3. Explain It in 90 Seconds: the radar's top story, run on camera and cut into a 90-second explainer. Cards: no stock footage, one number on screen, a source line. The winning cut can air as that day's short.

Not yet: seasons, standing teams, prizes or sponsors. A real-money “first dollar” episode needs approval first.

Giving our video editor taste

From Angus Sewell's reel on the “taste book” (60 s, about 94K plays). His point: AI editing is already better than a cheap freelancer but has no taste, so have it learn somebody else's.

What he does

1. Pick 3 to 5 videos whose editing you love, and give them to the AI.
2. An “editing analyst” prompt detects every cut with ffmpeg, logs the pacing (cut length, cuts in the first 10 seconds, zooms, captions, effects) and writes the rules into a TASTEBOOK file (“cut every 1.8s, zoom on the punchline, caption the key words”).
3. The AI edits the raw video in that style.
4. Four sub-agents check the edit: cut dead air, score the hook, find B-roll, fix captions.
5. A small model marks every line keep, delete or move.
6. Render lots of versions and pick.

What he doesn't show: before-and-after numbers. The full prompt is behind a comment.

How we'd do it (no Premiere, no camera)

• A taste reference set for each format (daily short, weekly roundup, agents-at-work), analyzed only, never reused.
• A taste analyzer agent that measures each reference: cuts, timing, zooms, caption style, B-roll ratio, numbers on screen. We already have the tools (frames, OCR and Whisper).
• A TASTEBOOK per format with measured values, plus our brand rules: stamp, day counter, source line.
• Taste as code: the rules become hyperframes components and a timing grid, so every agent edit inherits them.
• Critic pass (his four checks, made measurable) plus the line-by-line pass. Then 3 to 5 variants, a judge agent, and a person picks.
• Close the loop: after posting, check which rules actually moved retention, and drop the ones that didn't. That's the receipts step he skips.

Video backlog

Drawn from your saved posts (by topic only; individual posts stay private), your GitHub stars and the order itself.

Skipped on purpose: jailbreak and refusal-removal repos (news only, never demos).

Our own tools: forks of popular repos

“Maybe we build our own custom software from some of these popular repos.” Seven picks, ranked. Plan only: nothing has been forked or created.

Not picked: tools under Elastic, Business Source or AGPL licenses (we can't sell them freely, or must publish our changes); trading tools (selling them carries financial-advice risk, so episodes only); tools for hiding AI writing (against the brand).

Workflows from your saves

“Some of the stuff I saved was for me to research how to do certain things so I could turn them into a workflow or have agents do it to try and produce income.” These are the how-to saves, grouped by idea. Plan only: anything involving real money, accounts, spend or sign-ups needs approval first.

Haiku read about 1,865 saves and found 546 how-to posts: ones that teach a method, a tool setup or a money play. They group into the 19 ideas below (526 posts; 20 one-offs were set aside). Ideas are sorted by how well they fit “agents earning on camera”, then by whether a first version can run in the no-keys sandbox.

Adopt for our stack 252 saves

Many saves are techniques to adopt into our own production and agent stack, not income ideas. Tagged by theme:

The first one is worked through in the “Giving our video editor taste” card. The full per-post list stays in the private notes.

Also from the Angus study: company-level lead routing, meaning agents score companies visiting a site (never identifying people) and route follow-ups. It's a cheap business service.

Income figures are what the saved posts or common sources claim; none are verified. Our line on personal data: business-entity data only (companies, not people). That rules out the person-level tracking some growth recipes teach.

Renaming the existing channel

A simple checklist for moving the current YouTube channel to the new name once you pick it. You do the rename yourself; nothing has been changed. The channel-specific details are in the private notes.

• Stays the same: the channel ID, its old link and its subscribers.
• YouTube's limits: you can change the handle twice in 14 days, and YouTube holds the old handle for 14 days in case you switch back. The name can also change twice in 14 days; a name change removes a verification badge.

1. Claim the new handle on Instagram, TikTok and X first, so nobody grabs it in between.
2. YouTube Studio → Customization → Profile → Handle, then Publish.
3. Same page → Name, then Publish.
4. Swap the picture, banner and description for the new brand kit's.
5. Check that old playlists are set the way you want.

Work with us: turning episodes into clients

Angus Sewell's format, turned into a client funnel. The episode shows a build working; business owners who want one apply, and you build theirs. Everything is written and async (no camera, no calls). Plan only: there's no live form and nothing collects data until you OK it.

Posts that lead to it

• “Can we just build this?”: the big company's price, then agents build it, then it running. Ends with “Want one for your business? Apply.”
• “We built it for a business like yours”: a build for a made-up dental office, roofer or salon, with hours saved.
• Agents at work: agents earn or demo with a repo. “Want agents running this for you?”
• Viewer's idea, scoped on screen (written permission only), and a monthly receipts recap with “N spots open”.

One call to action per post. No comment-for-a-link gates.

The application (short)

1. Your business: what you sell and to whom, industry, team size, what you want built, what it costs you today (hours or dollars), tools you use, which post brought you.
2. Fit: built for you, built with you, or a ready-made pack; budget range; timing.
3. Contact, the minimum: name, work email, website (optional). One required box to store the answers, and one optional, unticked box for future emails.

Never asked: phone, home address, birthday, payment details, logins or customer data.

How agents triage it (0 to 100)

70+ gets a drafted written scope and quote for you to review; 45 to 69 gets the pack or a clarifying question; under 45 gets a polite no with free resources. Agents only read the answers given, never look people up, and you approve every reply. Anything involving personal data, health, finance or regulated work comes straight to you.

Where it lives

Draft privacy notice: answers are used only to review the request, never sold, kept 12 months, deletable on request; marketing email only if ticked, with an unsubscribe link.

Offers for you to price

• Custom build: we build it, then hand it off with a recorded walkthrough, docs and 2 weeks of email support.
• Done-with-you: built together in their accounts over 2 to 4 weeks, async.
• Skill pack: a tested agent skill from the show, sold as an instant download (low ticket).
• Build audit: a written review of their tools plus a build plan, which could be credited toward a build.

Open decisions

Each one has a default we're using until you say otherwise. Write your call in the notes.

1. The name. Default: Year One SI.
2. The direction: Era, Builder or Spectacle. Default: Era, with “Fork & Ship” as the agents-at-work series inside it.
3. Who the brand is for. Default: AI-curious builders (small-business owners, creators, side-hustlers).
4. The voice. Default: skeptical verdict, receipts first, wit allowed, no hype.
5. How this page is built. Default: static HTML on Cloudflare Pages, hidden from search.
6. Which tool we fork first. Default: the SI video factory (hyperframes).
7. Sandbox models: may agents use a free local model on our own GPU? Three Season 1 repos need one. Default: ask the lead to allow it.
8. Simulated or real runs: agents-at-work runs stay simulated until you approve otherwise. Default: simulated only.
9. Handles and domain: claim them on all four platforms, and buy the .com? Default: you claim them yourself once you pick; we never sign up.
10. Trademark: get a lawyer's clearance before merch? Default: yes, before any merch spend.
11. Season 1 dates. Default: premiere on launch day, finale on September 29, 2027.
12. On camera or faceless? Angus is a talking head, but you've said no camera or phone. Recommendation: faceless. Screen recordings of the agents at work plus a voice-over (yours, or a licensed AI voice), with the receipts on screen. The agents are the talent and the proof does the job of a face.
13. Client funnel: where the form lives, offer prices, the privacy notice, and whether client builds can be filmed. Default: Tally, nothing live until you OK it.
14. When does the build challenge grow into its own league? Default: for now it stays a segment inside this brand, with no seasons or standing teams, until you say so.