SI Brief

The Super Intelligence brand: strategy brief for Jason

Everything put together so far, on one page: the date, the names, the content, the tools we could build, and the open decisions. Read it, type your tweaks in each notes box, then copy them all back to the lead.

9.29.26 Executive Order 14434 · AI became “Super Intelligence”

Your notes save in this browser as you type. When you're done, press Copy all my notes and paste the result to the lead. Last updated Saturday, November 14, 2026, 8:30 PM EST.

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.

FactWhat the record says
SignedTuesday, September 29, 2026 (“The White House, September 29, 2026.”)
PublishedFriday, October 2, 2026, Federal Register 91 FR 63129 (document 2026-20321)
In effectFrom signing. No separate effective date is given.
What it doesFederal agencies use “Super Intelligence” and “SI” in place of “Artificial Intelligence” and “AI” in official communications, and the executive branch “will not acknowledge” the old terms. For now, SI means exactly what the law already calls AI: it's a rename, not a new technology.
Next dateSaturday, November 28, 2026: within 60 days, the President's science adviser must propose a legal definition of “Super Intelligence”.

One wording fix: it's an executive order, not a law. Congress hasn't passed anything on superintelligence. Two bills to ban it (S. 5493 and H.R. 10538, introduced September 23 and 24, 2026) are still in committee. So the line is “the day AI became Super Intelligence”, not “signed into law”.

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.

What we studiedWhat we found (264 posts, catalogued October 3, 2026)
Size174K on Instagram, 42K on TikTok and about 15K on Substack
GrowthThe account started in September 2025. Median plays per post went from about 540 that month to 31K in January 2026, 142K in March and 174K in July. He posts 15 to 22 times a month.
FormatTalking head plus screen recording, about 65 seconds. Strong one-line hooks (“Forbes just bamboozled all of us.”).
Best series“Can we just build this ourselves?” / “[X] are looting [big company]”: 11 posts from July to October 2026, at a median of 193K plays. The incumbent's price goes on screen, then he builds the thing with an agent and a prompt.
News as the hookOnly about 1 in 6 posts hangs on a news event. The news frames a build the viewer can copy. Workflow tutorials (86K median) and lead-gen builds (97K) outperform plain takes (25K).
Old stories winSince March 2026, resurfaced stories with a build angle get about 227K median plays, against about 155K for fresh news. He's drifting toward older viral stories, and his median lag went from 20 to 57 days this year.
Where his stories come fromX around launches, the platform's own feed, Reddit, and Google AI Overview screenshots (his most-used on-screen “source”). Almost never Hacker News, arXiv or GitHub trending.
The gapOnly 14% of his posts show a real result. Most show the prompt and the idea, not a finished system producing a measured outcome. About a quarter of his recent posts are sponsored, and most gate the prompt behind a comment.
How he earnsComment-for-the-prompt funnels into a Substack, paid “skill packs”, sponsors, and consulting that comes in from the content.

Your spin on him: a mix of Angus and seb.ai, doing the AI news and taking the content rather than the style, with a close look at how they source it. Then the twist: our agents actually run the build and show the result, income or a real demo, receipts on screen.

Our edge, in one line: his own numbers say speed isn't what pays; a re-angled story with a verified build is. So we go faster at finding stories and add the proof he leaves out.

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.

The Era

We're living in Year One of the SI era, and this channel keeps the record.

Year One SIThe SI EraDay One SIAnno SINine Twenty-NinePost-AI

The Rename

The government renamed AI. What the rename means, and what's real under the new name.

Formerly AINot ArtificialSay SIUnacknowledgedAI (Retired)The Rename

The Spectacle

In the spirit of Showtime's The Circus: SI is the greatest show on earth, with politics, money and machines.

Super Intelligence CircusSI CircusRingmaster SIThree Ring ComputeCenter RingThe Big Top

The Lab

We don't relay repos, we run them: agents in a sandbox, on camera.

Crash Test AgentsDoes It RunThe Proving GroundBench DayRun LogSandbox Season

The Verdict

Receipts, not hype. Every claim gets a verdict.

Legit or HypeGround Truth SIIt RunsThe VerdictReceiptsVerified SI

The Insider

The backroom politics of SI: who writes the rules, and who profits.

Order 14434Back Channel SIThe BriefingExecutive SIThe Cabinet

The Builder

Agents test the repos, and we fork the winners into our own tools.

Fork & ShipRepo to RevenueShipped by AgentsAgents for HireThe Agent ShopForked

The League echoes the separate League project

Agents compete head to head, with scores and seasons. Use only if you want the link to your separate League idea.

Agent LeagueHead to Head SIBox Score SIAgent Draft

The Field Guide

SI in plain English for people who aren't developers.

SI Field GuidePlain SISI for HumansThe SI ManualSI Explained

The Ascent

The big-picture, superhuman-capability story told long form.

OvermindAbove HumanThe UpgradeBig Brain EraAscent

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.

NameHandleYouTubeInstagramTikTokX.comTrademark
Year One SI@yearonesi✓✓✓✓✓ free“Year One” alone exists in software, boots and events; the full name has no hit
Fork & Ship@forkandship✓✓✓✓✓ freeNone found
Ringmaster SI@ringmastersi✓✓✓✓✓ free“Ringmaster” exists for software; none for entertainment
Crash Test Agents@crashtestagents✓✓✓✓taken“Crash-Test Dummies” is registered for AI research
Order 14434@order14434✓✓✓✓✓ freeNo hit, but hard to own; better as a series name
Three Ring Compute@threeringcompute✓✓✓too long; @3ringcompute✓ freeSimilar marks in software and studios
Formerly AI@formerlyai✓taken✓✓takenNone found
SI Field Guide@sifieldguide✓✓✓✓takenNone found
Shipped by Agents@shippedbyagents✓✓✓✓takenNone found
Repo to Revenue@repotorevenue✓✓✓✓takenNone found
The SI Era@thesiera✓✓takentakentakenNone found
Ground Truth SI@groundtruthsi✓✓✓✓takenCrowded in software
Day One SI@dayonesi✓✓✓takentaken“Day One” is registered to Amazon for entertainment
Legit or Hype@legitorhype✓taken✓takentakenNone found; better as a segment
Super Intelligence Circus@sicircustakentaken✓takentaken (since July 2025)Showtime holds no “Circus” trademark, but the show title still counts; a “Circus” streaming-series filing is pending

Rule for every name: “Super Intelligence” on its own can't be ours. It's now the government's word for AI, and others rushed to file it after the order, including SpaceX for clothing and entertainment. The other word has to carry the brand. Before any merch spend, a trademark lawyer should clear the pick.

Top 3 directions

Year One SI

recommended The Era

The first year of the SI era, logged as it happens: “Day 47 of SI.” The date is the name, and it's free on all four platforms plus the .com. It has a built-in finale on Wednesday, September 29, 2027.

Models: Angus Sewell (hooks and pace), seb.ai (topics), Johnny Harris (a source doc for every video).

First 5 videos

  1. Day 1: the government renamed AI. What EO 14434 actually says.
  2. A bot turned $68 into $750K? Our agents checked.
  3. Agents try to run a real sales pipeline on the free Salesforce (Twenty).
  4. Can a laptop run Super Intelligence?
  5. Day 60: the November 28 definition deadline.

Merch: a “YEAR ONE · 09.29.26” tee, a “DAY 1” dad hat, an “SI ERA · EST. 9.29.26” sticker, and anniversary drops.

Fork & Ship

The Builder

Agents put the repos everyone recommends to work, and we fork the winners into our own tools. The show and the product share one name. It's clean on every check, but it doesn't say “SI”, so the tagline has to: “Super Intelligence, put to work.”

Models: seb.ai (the repo audience, but with proof), long-form agent-tool reviewers.

First 5 videos

  1. Agents get MoneyPrinterTurbo and one goal: make a Short that earns.
  2. The free Semrush, run on a real local business site.
  3. Agents edited this whole video, so now we're forking the tool.
  4. Agents vs the free Salesforce: book 10 meetings.
  5. Season 1 scoreboard: what earned, what didn't.

Merch: “IT EARNS” / “DOESN'T RUN” stamp stickers, a “FORKED 09.29.26” tee, and a fork-glyph hat.

Ringmaster SI

The Spectacle

The insider-docuseries tone of The Circus without borrowing its name. Politics, money and machines, with agent runs as “acts”. It's free on YouTube, Instagram, TikTok and X, and the .com is unregistered.

Models: The Circus (Showtime) for tone, Johnny Harris and Wendover for craft, Angus Sewell for shorts.

First 5 videos

  1. The big top opened on 9.29.26: inside the order that renamed AI.
  2. Who wants to ban superintelligence? The two bills nobody passed.
  3. Act 1: three agent teams, one repo, one hour, one dollar goal.
  4. The trademark rush on “Super Intelligence”.
  5. Who's the ringmaster: the government or the labs?

Merch: a top hat embroidered “EST. 9.29.26”, “ADMIT ONE · SI ERA” ticket stickers, and a circus-bill poster.

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.

Angus Sewellseb.ai
Size174K on Instagram108K on Instagram, grown in about 5 months
FormatTalking head plus screen recording, about 65 secondsSelfie cam over a laptop showing the source, 25 to 60 seconds; also carousels
HooksOne-line provocations; a recurring “are looting [big company]” series“SOMEONE JUST…” caption card, then “I'm gonna give it $X and see…”
Cadence15 to 22 posts a month128 to 202 posts a month at peak
Shows real proofAbout 14% of postsAbout 2% of posts

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

Season 1: ten repos and the job each one gets
RepoWhat it claimsThe job we give the agentsRuns in the no-keys sandbox?
twentyhq/twentyOpen-source SalesforceSet up a sales pipeline and work a public lead listYes (AI parts need a local model)
harry0703/MoneyPrinterTurboOne click, a finished ShortMake 5 Shorts; estimate what they'd earnOnly with a local model; stock-footage key needed
heygen-com/hyperframesHTML to video, built for agentsProduce a real episode's graphicsYes
every-app/open-seoOpen-source SemrushAudit a real local business site; would they pay?No: needs a paid data account (your call)
TauricResearch/TradingAgentsA multi-agent hedge fund30-day paper-trading runOnly with a local model; paper only
virattt/ai-hedge-fundAn AI hedge fund teamHead-to-head paper run vs TradingAgentsOnly with a local model and a data key
Panniantong/Agent-ReachAgent eyes on the internetFind 3 stories before the big accounts doYes, public sources only
Egonex-AI/Understand-AnythingAny codebase as a teaching mapExplain a famous repo in 60 secondsYes
DeusData/codebase-memory-mcpMemory graph for coding agentsAgents fix a real bug faster with it than withoutYes
AlexsJones/llmfitWhich models run on your hardware“Can a $500 laptop run SI?”Yes

Scorecard: 8 criteria from 0 to 5 (install, time to first result, does the job, income or demo result, claim match, agent-friendliness, real-world cost, health and license). Every run gets a full run log. Nothing runs until the lead sets up the sandbox.

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.

Why V1: every episode leaves a lasting score, the forked tools are income beyond ad money, and showing proof is something relay accounts can't copy. V2 works as the Season 1 theme; V4 waits for your go-ahead on money.

Video backlog

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

Video ideaWhy nowWhat's changedFormat
Viral “bot turned $68 into $750K” trading claims, checkedProof is the brand. A perfect first verdictClaims like this have multiplied since springA
The open-source personal assistant everyone's posting (OpenClaw)The most-named open agent in your savesRenamed twice; big claims never verifiedB
Running big models at homePairs with the “Can my computer run SI?” toolThe models in the old posts are outdatedA + B
Vibe-coded apps leaking dataA safety angle against the hypeRetest with today's coding agentsB
Free replacements for Adobe, AutoCAD, Screen Studio and paid scrapers“Free replacement for an expensive tool” is the top-performing hookLicenses and support need recheckingB (Replace the Incumbent)
Agents driving creative tools (Unreal, Blender, video)Agent-on-camera footage that looks greatSeveral are sponsored posts; we stay neutralB
“1,000 videos for near zero” with code-made videoIt's literally our own production stackAn Apache-licensed alternative now existsB + our first fork
An AI second brain from your own notesEvergreen and very saveableBuilt-in memory features now compete with DIYA
“13,000 agents bought their own servers”A wild claim worth investigatingUnverifiedA
Companies gaming what chatbots saySI-era misinformationSame tech, new name, same tricksA
The law as a diff: “AI” to “SI”Show the November 28 definition proposal as a code diffDue Saturday, November 28, 2026A
Repo roundups that name no reposOur contrast: we name it, run it, no comment gateComment-gating is now near-universalPositioning

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

What we drew on (counts only)
  • 1,357 GitHub stars going back to July 2023, busiest July to September 2025. The biggest groups are coding agents and skills (222), multi-agent frameworks (157), learning lists (167) and MCP and agent tooling (76).
  • About 1,865 saved Instagram posts, 413 of them about SI, agents or repos: SI news (80), agents and automation (75), GitHub repos (71), image and video generation (37), coding agents (35), prompts (31). Saves on these topics jump from March 2026.
  • Creator studies of Angus Sewell and seb.ai, plus the old Closed Zone research.

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.

#Repo → our toolWho'd use or payLicenseHealthThe episode
1hyperframes → the SI video factoryUs first; then creators and agencies (template packs, hosted rendering)Apache-2.0: fork and sell, keep the noticesVery active“Agents edited this whole video, so now it's ours”
2open-seo → SI SEO DeskLocal businesses and small agenciesMITVery active“The free Semrush, then one a plumber can use”
3llmfit → “Can my computer run SI?”Hobbyists; a lead magnetMITVery active“Can a $500 laptop run Super Intelligence?”
4Agent-Reach → SI RadarCreators who want stories firstMIT (public sources only)Very active“Beat the AI influencers to the story by 3 days”
5Understand-Anything → Repo X-RayOur show, educators, teamsMITSlowing“Any repo in 60 seconds”
6open-code-review → SI Code ReviewSmall teams shipping with agentsApache-2.0Active, big-company backed“Code review for non-coders”
7codebase-memory-mcpWe already use itMITVery activeContribute fixes instead of forking

First pick: the SI video factory (hyperframes). It pays off on day one because it makes our own videos, needs no paid keys, and its look becomes the brand's. If you want revenue first, pick open-seo instead, but it needs a paid data account (your call).

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: any real money, account, spend or sign-up is your call.

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.

Run these first (free, sandbox-safe, great on camera):

  1. Video clipping service: agents cut a long public-domain video into ranked Shorts with local tools; we show the clips and what clippers charge.
  2. Automation for a small business: agents build a receptionist, follow-up and intake system for a fictional business and count the hours saved, on camera.
  3. Reports from free public data: agents turn SEC, permit or registry data into a weekly niche report that someone would subscribe to.

From the Angus Sewell study we add: verified skill packs (sell agent skills with their test results), a priced-stack tracker (live prices of AI tools, a lead magnet), and a written podcast-pitch desk (pitches by email, no camera or phone; you approve every send).

AI automation and web services for SMBs 64 saves fit: high sandbox-safe first version

What agents would do: Agents audit a small business's workflows, build the matching automation such as a receptionist, CRM follow-up, intake or website, and track hours saved. Demo the build on a fictional business first.

What it needs: Claude or Claude Code; n8n, Vapi or GoHighLevel (your call); client access; hosting; outreach list

Possible income: Setup fees plus monthly retainers. Sources cite roughly $500/month per client and $1,500-10,000 per project; closing clients is the bottleneck.

Watch out: Client data handling; overpromising results; spammy SEO or fake-review tactics are policy risks.

Agent tooling, coding swarms and AI SaaS 42 saves fit: high sandbox-safe first version

What agents would do: Agents build, benchmark and self-improve agent harnesses, skills, memory layers and small apps. Results are published as repos, demos and cost or quality scores that can be shown on camera.

What it needs: Claude Code and Agent SDK (Max sub); GitHub; eval task suite; optional Higgsfield or Browserbase accounts (your call)

Possible income: Consulting, selling skill or template packs, app subscriptions or managed setups. Mostly qualitative and uncertain.

Watch out: Cloning products can raise IP and licensing issues; autonomous spend loops; platform terms.

Trading signals: futures, options, crypto, betting 34 saves fit: high sandbox-safe first version

What agents would do: Agents scan markets for rule-defined setups such as GEX, ORB, VWAP, orderflow, wallet tracking or odds mispricings, log every alert and paper trade, and publish a transparent scoreboard. Alerts only and paper trading by default, with no autonomous live orders.

What it needs: Market, options, on-chain or odds data feeds (Jason for paid feeds); charting platform; paper broker; Telegram or Discord bot; live capital only later (your call)

Possible income: Trading profits or paid alert subscriptions. Results are unproven and most retail traders lose money.

Watch out: Securities and advisory rules for selling signals; gambling and crypto rules; memecoin scams; reputational damage if the track record is poor.

Quant strategy backtesting and research 23 saves fit: high sandbox-safe first version

What agents would do: Agents read papers, code the strategies, backtest them with realistic costs and out-of-sample checks, and publish a comparison report. The research side can run on free historical data.

What it needs: Python in uv, free or low-cost historical data, backtest library, optional GPU; live brokerage only later (your call)

Possible income: Sell research reports or templates, or trade your own capital. Most backtests do not survive live trading.

Watch out: Not financial advice; selling signals may trigger advisory regulation; overfitted results.

Courses, prompt packs and content strategy 16 saves fit: high sandbox-safe first version

What agents would do: Agents research a topic, draft modules, prompt libraries or playbooks, test them against sample tasks, and package them with a lead magnet and funnel. A build-in-public series about the process doubles as the content.

What it needs: Claude; NotebookLM; course or storefront platform (your call); DM automation optional; audience

Possible income: Course, prompt-pack and coaching sales. A saturated market where income depends on audience size.

Watch out: Income-claim and FTC disclosure rules; quality of AI-written courses.

Video clipping and AI editing services 15 saves fit: high sandbox-safe first version

What agents would do: Agents transcribe long videos, find viral moments, cut silence and filler, add captions and zooms, and render vertical or platform-specific exports. A first version can run on local FFmpeg and Whisper using sample footage.

What it needs: FFmpeg, Whisper, Remotion or open-source editor; source footage; optional OpenCut/OpusClip account (your call); client or Whop accounts (your call)

Possible income: Per-clip or monthly packages for creators, Whop per-view payouts, or freelance editing retainers.

Watch out: Clipping others' content without rights; payout-platform terms.

AI creative production: film, 3D, game, music 44 saves fit: high needs real accounts

What agents would do: Agents turn a brief or prompt into storyboards, shots, 3D scenes, music or game builds through Claude plus MCP-connected tools such as Blender, Unreal, Higgsfield or Suno. Show the prompt-to-finished-asset run on camera.

What it needs: Claude Code; Blender, Unreal or ComfyUI (free); Higgsfield, Runway, Suno or OpenArt accounts (your call); GPU

Possible income: Client video and 3D work, asset packs, streaming royalties or game sales. Streaming royalties and asset packs are usually small.

Watch out: Streaming and distributor AI-music rules; asset and sample licensing; AI-disclosure rules.

Faceless AI video and YouTube channels 41 saves fit: high needs real accounts

What agents would do: A research agent picks niches and studies winning formats, a writer agent scripts, a production agent generates voice, visuals and edits, and a scheduler uploads and tracks retention. Weekly, the team logs which formats gain views and doubles down on them.

What it needs: YouTube/TikTok/IG channels (your call); AI video, voice and image tool subscriptions (your call); scheduler; analytics; ~$50-200/mo

Possible income: Ad revenue, sponsorships and affiliate links. Most new channels earn little for months, and monetization thresholds apply.

Watch out: Platform rules on mass-produced or reused content; demonetization; copyrighted source clips.

AI personas and multi-account social pages 32 saves fit: high needs real accounts

What agents would do: Agents generate persona or niche content, schedule posts across many pages, reply to comments and report growth by account. Run it on one page first and show the dashboard stats on camera.

What it needs: Social accounts (your call); scheduler such as Postiz; Claude and an image or video generator (OpenArt, Higgsfield); affiliate programs; VPS optional

Possible income: Sponsorships, affiliate commissions or social retainers for local businesses. Growth is slow and uneven.

Watch out: Platform bans for multi-account or automation, inauthentic-behavior rules, AI-disclosure rules, reposting rights.

E-commerce, TikTok Shop, POD and digital products 28 saves fit: high needs real accounts

What agents would do: Agents research trending products, script videos, draft listings and mockups, and track sales and margins per product. Digital-product versions can be built end-to-end with no inventory.

What it needs: TikTok Shop, Shopify, Etsy or Amazon seller accounts (your call); supplier or POD accounts (your call); ad budget (your call); AI image and video tools

Possible income: Product margin or affiliate commissions. Highly variable, and many sellers lose money on ads and returns.

Watch out: Seller-account bans, product-claim and compliance issues, trademark and merch IP, refund exposure.

AI UGC and product ad creative 22 saves fit: high needs real accounts

What agents would do: Agents take a product brief, write hook variants and scripts, generate avatar or UGC-style video and image ads, and package them with a tracked winner report. Demo before and after assets on camera.

What it needs: Arcads, Higgsfield, Seedance or Creatify accounts (your call); product photos; Claude; optional ad account (your call)

Possible income: Per-creative or monthly packages sold to DTC brands. One source cites roughly $1,500-3,000 per client package; results vary widely.

Watch out: Fake-testimonial and AI-disclosure rules; likeness consent; ad platform policies.

Security research, OSINT and privacy services 30 saves fit: medium sandbox-safe first version

What agents would do: Agents run scoped recon, scanning and triage on authorized targets or labs, summarize findings, and draft reports. Privacy variants audit exposure and file opt-outs with client consent, and secret-leak scanners run on your own repos.

What it needs: Written scope or authorization; Ghidra, Nmap, AFL++ and OSINT tools; bug bounty accounts (your call); lab environment; client consent

Possible income: Bug bounty payouts, audit fees or privacy subscriptions. Bounties are lumpy and competitive.

Watch out: Unauthorized testing is illegal; PII and OSINT lookup privacy law; decompilation licensing.

Lead generation and outbound outreach 22 saves fit: medium sandbox-safe first version

What agents would do: Agents pull public records such as permits, property data and clinic registries, score and enrich leads, draft personalized outreach and log replies in a CRM. A first version can use free public datasets and a spreadsheet.

What it needs: Public datasets (NYC permits, NPI registry, county records); CRM; email or DM tool (your call); Claude API; enrichment tool optional

Possible income: Per-lead fees, monthly lead subscriptions or retainers. It depends on contractor and clinic demand.

Watch out: CAN-SPAM, TCPA and privacy rules for outreach; scraping terms; Instagram DM automation limits.

Data, research reports and monitoring feeds 22 saves fit: medium sandbox-safe first version

What agents would do: Agents pull public data from sources such as SEC, GDELT, NASA, Git statutes and reviews, structure and refresh it, and produce reports or alert feeds. This runs entirely on free public sources.

What it needs: Python and free APIs (EDGAR, GDELT, NASA EONET); LLM API; hosting; optional paid data such as Similarweb (your call)

Possible income: Subscriptions, per-report fees or newsletter revenue. It takes real niche demand to earn.

Watch out: Scraping terms; financial-content regulation for investment reports; data accuracy.

Arbitrage flipping and auction buying 20 saves fit: medium sandbox-safe first version

What agents would do: Agents monitor clearance listings, auctions and marketplaces, estimate resale comps and margins, and alert on underpriced items. Listing and comp scoring is software-only, while buying and shipping need a human.

What it needs: Marketplace and auction accounts (your call); sourcing capital (your call); price comp data; storage and pickup

Possible income: Resale margin per item, usually thin after fees and time. A realistic range is hard to cite.

Watch out: Dealer-license, lien and title rules; platform bans on scraping and bots; inventory loss.

Local LLM and self-hosted infrastructure 14 saves fit: medium sandbox-safe first version

What agents would do: Agents install, benchmark and document local models, sandboxes and self-hosted tools such as llama.cpp, BitNet, SearXNG and Proxmox. The output is tested setup guides with speed and cost numbers.

What it needs: Local GPU or CPU server; llama.cpp; open-weight models; Docker; optional used hardware spend (your call)

Possible income: Setup fees, hosting subscriptions, guides and affiliate links for private AI rigs. The market is small and mostly technical.

Watch out: Model license limits; hardware spend; support burden.

Meta ads funnels, audits and testing 15 saves fit: medium needs real accounts

What agents would do: Agents audit ad accounts and landing pages, scrape competitor ad libraries, draft creative and funnel structures, and score the leaks. A first version can audit public ad-library data with sample funnels.

What it needs: Meta Ads account and client access (your call); ad budget (your call); Apify or ad library tooling; analytics data

Possible income: Audit fees, agency retainers or managed ad fees. One source cites around $2,000 per audit; it varies by client.

Watch out: Meta ad policy and account bans; scraping terms; client budget liability.

Government contracts, grants, legal and funding 19 saves fit: low sandbox-safe first version

What agents would do: Agents scan SAM.gov, grants.gov, tender and county lists, match opportunities to a client profile, and draft bids, applications or briefings. Legal drafting is first-pass only, for attorney review.

What it needs: SAM.gov registration and UEI (your call); attorney or CPA review; public data feeds; CRM or email delivery

Possible income: Contract or grant awards, subscription alerts, contingency or coaching fees. Win rates are low and the sales cycle is slow.

Watch out: Unauthorized legal practice, regulated brokerage, and contingency-fee rules by state.

Trades, real estate and physical businesses 23 saves fit: low needs real accounts

What agents would do: Agents handle the paperwork around a physical business, such as lead capture, quoting, deal modeling, scheduling and invoicing, and track margins. They can run demand and deal-math research before any capital is spent.

What it needs: Capital, licenses, insurance and equipment (your call); local leads; load boards or county data; attorney or CPA

Possible income: Service fees, deal assignment fees or resale margin. It is capital-heavy and highly local.

Watch out: Licensing, insurance, emissions law (derate removal) and real-estate rules; capital loss.

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 @theclosedzone

The simple version: there are no uploads and no stats, so there's nothing to hide. You do the rename yourself; nothing has been changed.

  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. Content → Playlists: confirm everything old is private.

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

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)

CriterionPoints
Problem is clear and specific20
Fits what we build20
Value at stake (hours or dollars)20
Budget fits the scope15
Timing10
Can be done in writing10
Would make a good episode (with consent)5

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

OptionCostNotes
Tally start hereFree (paid tiers optional)Live in minutes; export or webhook for agent triage
Google FormsFreeAnswers land in a Google Sheet; plain look
Our own form (Cloudflare Pages + D1)Free at this volumeFully branded on our domain; we own privacy and deletion. Move here once the name is picked.

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

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. Real money: paper or simulated income only for now? Any real account, data subscription or trading is your call. Default: paper 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. The League link. Default: kept separate; “League” names appear only as one family.