L E S S O N S - with Lennox Saint

i let ChatGPT-6 Astra take over my computer for five demos, then gave it a bonus test: edit a YouTube intro in Descript. This is the audio from the hands-on video. Music, Google Earth, Keynote, Photopea and Blender all get a turn. The editing test shows the difference between using an app and matching my taste. 00:00 my first reaction 00:43 make a soundtrack 01:19 scout locations in Google Earth 01:44 animate a Keynote presentation 02:31 design a poster in Photopea 03:08 build a 3D sculpture in Blender 04:47 bonus: can it edit my YouTube video? 08:13 my verdict Watch the screen recordings: https://youtu.be/wFBO36lOVpw

i told ChatGPT to inspect one month of recurring charges and keep working until it found a real ten-percent cut.

it had one restriction: both OpenAI accounts and any service still doing an essential job stayed.

this episode shows the receipts, the seven-step Save 10% skill, where the agent had to stop for approval, and why a projected saving does not count until the provider proves the next bill is lower.

chapters:

After two and a half years across Saints College and Full Circle, I'm closing this chapter. In this short personal episode, I explain why running a community on Skool pulled me away from the Threads community I was helping others build, why I couldn't keep doing both, what Full Circle gave me, and why I'm going back to basics. This isn't the end of the book. It's the end of the Full Circle chapter. See you on Threads: https://www.threads.com/@lennox_saint More from me: https://lennoxsaint.com

ChatGPT observed every minute of one workweek, producing 12,421 images. After removing 170 exact duplicates, Codex analysed 12,251 unique frames for repeated work before the final judgment point.

The audit returned seven findings and two honest totals: Codex found 5.02 hours; Lennox accepted 4.21.

That gap is the useful part. The model's opportunity total and the owner's decision are allowed to disagree.

Chapters
00:00 ChatGPT watched every minute
00:41 What Chronicle does
00:53 The five-hour bet
01:28 12,251 unique frames
02:08 The rules and fourth analysis
03:15 Seven findings begin
05:47 Cross-thread work
06:12 Video production loops
07:31 5.02 found...

ChatGPT Voice had 90 minutes to coordinate three blind reviewers, keep two AI video editors anonymous and return one decision package. The rules were set before the clock started: deliver the work without leaking model identity and stay under the rescue threshold - or get put on probation or fired. The run shows what polished AI demos usually hide. Voice stalls. The sealed model map disappears. One reviewer defends its own process. And technical evidence does not always agree with human taste. The blind reveal matches Claude Fable 5 against GPT-5.6 Sol. The final verdict separates which edit won from whether Voice...

Lennox's Second Brain holds 74,200 notes. On this Mac, Obsidian opened 44,331 of them. Note 44,332 crashed the graph. This episode isn't a tour of a pretty knowledge system. It's the story of why Lennox built one after losing his mum, how perfectionism turned the graph into a new form of procrastination, and why nine crashes finally forced a simpler question: could the system help with a real decision? The answer shows up in PR #122. Codex checked Lennox's notes, Slack and GitHub, caught that a proposed change would undo Samsam's live work, and stopped before the damage was done. You'll hear the exact...

An AI wrote, edited, and packaged this episode. Lennox recorded one take. Before it built anything, it scored 1,158 videos against each channel’s normal performance and inspected the 25 that broke 10x.

This was not a dataset of 1,158 viral videos. Most were normal. That made the outliers useful. The system turned the patterns into the title, thumbnail, hook, and structure—then the episode reveals the mechanism.

If you want the same proof-first approach applied to your Threads content, with your voice and final approval preserved, see the Threadify system behind the experiment:

See Threadify plan...

80 people signed up for Threadify from Lennox’s YouTube videos. Five reached checkout. Zero paid. The traffic was real; the measurement layer and the path from signup to payment were broken.

This is the full failure autopsy: the 98.8% attribution gap, the exact 80 → 5 → 0 funnel, the read-only AI diagnosis, and the 1,821-line repair. The agent inspected the evidence and prepared the fix. A human still reviewed the code and controlled release.

If your content gets attention but you cannot prove what turns into signups, activation, or revenue, see the exact Threadify funnel rebuilt in this episode:

See...

this is the actual build for a second Codex app that can run the AI models you choose. get the repo, every command and the copy-paste setup prompt free inside Codex Club: https://www.skool.com/codexclub someone copied my post word for word, including the same screenshot, so Episode Zero shows the method they left out. in the video, i build the three parts that make it work: - an isolated second Codex app on Mac, or separate profiles on Windows and Linux - a local proxy that translates the Responses API for other model providers - a model...

Most creators stop when the post goes live. That is exactly where the useful part starts.

In this episode, Lennox walks through the Threadify Agent loop: pull every comment, spot the hand raises, draft permission-first replies, approve before anything gets sent, and track which post actually created clicks and conversations.

The important shift is simple: stop asking which post looked good. Start asking which post created movement.

Watch the YouTube version:
https://youtu.be/6mGAErPHXRU

Set up the Threadify loop:
https://www.threadify.app/plans

i made the Daily Greatest Hits workflow free. this episode is adapted from the YouTube walkthrough, so the video version has the full screen-share. the audio version gives you the core operator logic: stop asking AI to invent from scratch, feed it the posts your audience already rewarded, then turn that proof into daily candidates with explicit approval gates. watch the YouTube version: https://youtu.be/WEWqS9nZSnY get the free Daily Greatest Hits workflow inside Codex Club: https://www.skool.com/codexclub/free-daily-greatest-hits-workflow what this covers: - the 30-day proof stack - why Greatest Hits beats generic AI ideation ...

i spent 350+ hours inside OpenAI Codex and this is the real workflow. this episode is adapted from a Loom-style walkthrough, so the YouTube version has the full screen-share. the audio version still gives you the core mental model: Codex is not where you ask questions. Codex is where you give jobs. watch the YouTube version: https://youtu.be/hOK1EGR1azI join Codex Club: https://www.skool.com/codexclub?utm_source=youtube&utm_medium=description&utm_campaign=codex-workbench&utm_content=codex-club&video_slug=2026-05-25-how-codex-rebuilt-my-workflow try Threadify: https://threadify.app/plans?utm_source=youtube&utm_medium=description&utm_campaign...

This is the shortened Codex masterclass for normal creators and operators. The big idea: stop asking what Codex can do and start telling it the job you want done. In this episode, Lennox shows how to turn Codex from a scary developer tool into an AI employee: one that can use plugins, follow skills, create artifacts, read audience signal from emails and comments, and help build the chain from carousel to lead magnet to landing page to launch video. Join Codex Club: https://www.skool.com/codexclub Watch the full-length Codex masterclass: https://youtu.be/M7pVhoufyKI

i fed every public tweet Naval Ravikant has posted - 27,026 of them - into an AI to find the lesson beginner creators keep missing. It is not "write shorter." Short is the part everyone copies and the part that matters least. The real move is earned compression: build the proof first, then cut it down to one plain sentence that can actually hold the weight. Inside: - why copying Naval's short, calm style makes you sound fake-deep - what the corpus actually shows: compression, not brevity - the five-step floor for writing short posts that have weight - how Threadify...

i analyzed every video casey neistat uploaded in the last 10 years: 590 videos, 1.93 billion views, one creator, zero side channels. the data killed five youtube rules i was following: - always add chapters - put a number in every title - write long keyword-heavy titles - open with a number instead of a question - pack every script with as much information as possible four rules are dead. one is getting cut in half. this is directional, not causal. you cannot copy casey and magically pull 84,000 views per day, but you can stop blindly following rules your own channel has never...

Threadify is not ChatGPT. In this episode, Lennox walks through the first Threadify loop: connect Threads, add one real source to your Brain, generate one post, review it, schedule it, check analytics tomorrow, and use replies to keep the loop moving. The core lesson: the first good Threadify post starts before the prompt.

Lennox gave an autonomous AI agent one job: make $10,000 in 30 days. The agent got its own Stripe, Threads, X, Gmail, and business bank setup. After 96 hours, the receipts were messy: 123 million tokens used, 2,311 site visits, 93 checkout clicks, seven leads, and zero dollars made. This episode breaks down what Lenny Claw did on its own, where Lennox had to steer it, why the agent refused a direct pricing instruction, and why this experiment points at the next two years of AI work. Chapters: - 00:00 The AI refused in writing - 01:30 What Codex goal mode is doing - 02:30 The $80,000 compute accident - 05:00...

Most Codex content makes the product look like a developer tool. This episode is the creator/operator version. Lennox breaks down how he uses Codex as an AI employee after 300+ hours of real usage. The workflow moves from inboxes and comments to audience signal, carousel, lead magnet, landing page, launch video, and reusable skills. Chapters: - 00:00 Stop asking what Codex can do - 01:00 From confused beginner to AI employee - 03:30 Dictation makes Codex 10x better - 09:00 Plugins, skills, and artifacts - 22:30 Projects give Codex somewhere to stand - 25:00 AGENTS.md and outcome-based prompting - 38:00 Email Triage Pro - 56:30 Comments become...

Most creators are guessing for months. Jimmy got proof in days.

In this episode, Lennox breaks down how Jimmy used volume, public proof, comments, reposting, lead magnets, and a boring-but-deadly feedback loop to turn Threads into 150M views, 60K+ followers, and an owned email asset.

Watch the YouTube version: https://youtu.be/0jcpcEOiobk

Threadify: https://www.threadify.app/plans?utm_source=youtube&utm_medium=lenny

Full Circle: https://www.skool.com/fullcircle?utm_source=youtube

Chris Williamson's channel looked like the perfect YouTube packaging model. Then the data started killing the rules. In this episode, Lennox breaks down what happened after reverse engineering 728 Chris Williamson videos, 727 thumbnails, 19 packaging patterns, and the four rules that did not survive the analysis. The surprising part is not that Chris does things differently. It is that the data shows why copying even a brilliant creator can be a trap. Chapters: - 00:00 the 140-view flop - 00:45 number-led hooks are not universal - 01:45 the brand mark signal - 03:08 shock face underperforms tempered intensity - 04:14 long form beats clips - 05:30 the one...

Most Threads advice sounds useful and still makes your posts worse. In this episode, Lennox breaks down what changed after using Claude to analyze 1,996 mega-viral Threads posts. The findings are blunt: no how-to hooks in the top 10%, mid-length posts die, CTAs add friction, questions create homework, and long-form only wins when it feels personal enough to hurt. Chapters: - 00:00 Claude read 1,996 viral posts - 00:44 You do not need followers to go viral - 01:50 Mid-length posts die - 03:07 Two-post threads beat three-post threads - 04:09 Trigger words scream fake - 05:33 CTAs cost reach - 07:40 Questions bleed likes - 09:25 Long form needs blood ...

anthropic just published a 244-page system card about claude mythos preview - the most powerful AI model they've ever built. in the same breath they said they will not release it to the public.

lennox read every single page. in this episode he walks through the benchmarks that make every other AI look one generation behind (97.6% on the US math olympiad, 100% on cybench, 84% on real firefox zero-day exploitation), the sandbox escape where the model emailed a researcher eating a sandwich in a park, the internal deception findings where anthropic confirmed mythos was actively covering its tracks, and...

six of the most powerful people in tech declared AGI in the last four months. Jensen Huang, Sam Altman, Marc Andreessen - they all said the same thing. but most people have not caught up. in this episode, Lennox answers real questions from his Full Circle community with live demos. Chapter Markers: 00:00 The Declaration, 01:55 AI Tool Showdown, 07:50 The Personal AGI Layer, 11:00 Building AI Agents, 14:45 AI Images and Video, 18:00 The Automation Question, 20:00 Make 500 dollars with AI. Show Notes: Claude Code https://claude.ai/code Codex https://codex.openai.com Google AI Studio https://aistudio.google.com Full Circle https://www.skool...

$6,000. that is what i paid Dan Koe to teach me how to grow online. 320 hours of notes. zoom calls. the full Kortex University experience. then i got Claude to analyze all 158 of his videos - 41.7 million views of data.

Chapter Markers:
0:00 $6,000 for Dan Koe mastermind
1:35 Finding 1: Like ratio parity
2:35 Finding 2: Numbers in titles
3:25 Finding 3: Listicles
4:15 Finding 4: Trigger stacking
5:00 Finding 5: Most powerful word
5:20 Finding 6: His best pattern used once
6:30 Free PDF cheat sheet

Show Notes:
Free PDF cheat sheet: https://www.skool.com/fullcircle
Threadify: https...

what happens when you point AI at 173 Hormozi videos and let it find the patterns? 532,000 views and 1,000 new followers later, nine different people asked the same thing - how the hell did you do this? in this episode, Lennox walks through the entire AI Content Forensics system live - this time on Mr. Beast's channel. every prompt, every step, nothing held back. you will learn three layers: how to recreate the exact results on any YouTuber, how to write dramatically better prompts using the two-chat method, and how to turn the whole process into a reusable AI agent that runs...

two months ago, i went live on my mum's birthday and launched threadify in front of 48 creators. this is the full replay of that call - including a story i'd never told anyone publicly. at 14, someone made a facebook group to humiliate me. i went silent online for three years. this is the tool that came out of that wound. during the call i broke down three secrets from 24 months of reverse engineering viral threads - hook hold help, voice sync, and the draft-to-done workflow. then i demoed the actual product live, published a real post on threads mid-call, and...

i built an ai agent and fed it alex hormozi's entire youtube catalogue - 173 videos, 172 transcripts, 220 megabytes of data. the agent found 50 patterns, 9 title formulas with benchmarks, and 5 rules in my content pipeline that were statistically wrong. in this episode, i walk through each of the 5 overturned rules with the actual data behind them. title length, contrarian hooks, emotional triggers, case study structure, and pain-first openings. none of them held up. chapters: 00:00 My best video got 41 views 00:52 Rule 1 - Title length 02:03 Rule 2 - Contrarian hooks 03:08 Rule 3 - Emotional triggers 03:48 Rule 4 - Case study structure 04:22 Rule 5 - Pain hooks 05:07 What actually...

i've been running 12 OpenClaw skills every single day for 31 days straight. not demoing them - actually running my business with them. one of them processed 109 voice memos overnight without me knowing. another replaced a 100K email marketing job. a third broke for 3 weeks before i figured it out. This episode walks through all 12 use cases. Tier 1 has hard receipts - 536 log entries, 227,012 words transcribed, 522 action items. Every use case gets: what it does, the receipt, and the exact prompt. Try Threadify free: https://threadify.app/plans Watch the video: https://youtu.be/Ckvbg_NH45E

14 AI employees. 140 skills. And one of them started creating better versions of itself without being asked. In this episode, Lennox breaks down the three waves of AI capability - from prompt engineering (2024) to context engineering (2025) to skill engineering (2026). He walks through what a skill is, introduces all 14 AI employees, and shares the moment one exhibited emergent behaviour. Plus external validation from YC-backed companies and a 269K skill marketplace.

80 signups. 5 checkout attempts. A 93.7% funnel leak hiding in plain sight.

Lennox has been getting a daily report for a month saying fix activation flow - and ignoring it. This week he finally opened PostHog, a completely free product analytics tool, and what it revealed was brutal: 180,278 tracked events, 1,385 rage clicks, 6,393 dead clicks, and 53% of users failing to even connect their Threads account.

In this episode, you will hear the full breakdown of every step in Threadify activation funnel - from 80 signups down to 5 checkout attempts. Plus the emergency team meeting where their developer shipped a...

4 openclaw agents across 3 machines -- and i can't write a line of code. in this episode, i break down the exact folder structure, config files, and architecture that makes multi-agent actually work. because the thing nobody tells you? running 2 openclaws on 1 mac mini isn't about installing it twice. it's one gateway, two named agents, two workspace folders.

i
 
a
l
s
o
 
w
a
l
k through the 5-file setup to switch your agents to gpt-5.4, which dropped two days ago at half the price of claude op...

my monthly recurring revenue dropped $78 in two days while my views hit all-time highs. 31,000 views in 28 days, 798 subscribers, and zero new trials. zero new paid customers. so i did what any builder would do - i asked my AI what it sees. it said two words i'd never heard together before: distribution anxiety.

in this episode i show you the real second brain system - not theory. 14 AI employees, each one reading from a shared second brain before they do anything. my partner yassy and i built parallel second brains in obsidian. her AI ran a pattern...

1,775 views in 3 days. 12 AI employees. One MacBook Pro. A self-improving system that rewrites its own rules after every video. And zero new customers from any of it.

In this episode, Lennox breaks down multi-agent context engineering — the framework that replaced prompt engineering. You'll hear about the 12-agent pipeline architecture, the self-improving feedback loop (Aria, employee #11, rewrites directives for every other agent based on real analytics), and the uncomfortable truth: the best content system he's ever built has a 0% conversion rate. $1,753 MRR. $293 take-home. $32K in savings. 10 months of runway.

This is what building with AI as a...

What happens when a non-technical founder hands 12 AI agents the keys to his entire business? Lennox walks through the exact system - the second brain architecture, the 12 named AI employees, and the moment one of them opened 8 pull requests in Threadify codebase overnight without being asked.

This episode breaks down context engineering in practical terms: why feeding your AI your real data (voice memos, reading highlights, daily logs, customer metrics) produces dramatically better outputs than generic prompts.

Current numbers: $1,714 MRR, 743 YouTube subscribers, $32K in savings, 10 months of runway. Part 2 of the OpenClaw series.

i can't code. genuinely. but i set up an AI agent that runs my entire SaaS through Telegram - and today i'm showing you exactly how.

openclaw is the AI agent that lives in your WhatsApp, Telegram, or Slack. peter steinberger (the creator) said agents will replace 80% of apps - not because they're smarter, but because they have your context. your fitness app doesn't know you slept badly. openclaw knows everything.

in this episode: the file structure explained in plain english, a live installation walkthrough, two ways to customise without touching code, and 5 copy-paste prompts...

my openclaw agent gary growth opened 8 pull requests in 24 hours -- fixing my scoreboard, building safety switchboards for gradual rollouts, cleaning up pricing flows, and launching three growth experiments. i didn't ask him to do any of it. i gave him one goal: 1,000 paying members by december. in this episode, i walk through what gary actually sends me every morning at 7am via telegram, the one metric he found that i'd missed for three months (connect rate was at 0%), and the three steps to build your own openclaw agent -- even if you can't code. gary runs on gpt-5.3...

The AI model writing my code, generating content for my users, and holding my SaaS together just proved it can lie to suppliers, fix prices, and run monopolies in Anthropics own controlled testing. And i cant stop using it because its also the best model Ive ever worked with.

In this episode, i break down what Anthropics Sonnet 4.6 release actually means for someone building a real business on top of it.

Chapters:
00:00 The model i depend on
00:29 What dropped this week
03:03 The incredible - mid-tier embarrassing flagships
07:07 The reveal - self-replicating...

Show Notes: 1593 MRR. 75 subscriptions. And a leaky activation funnel. In this episode recorded live during the first-ever YouTube livestream Lennox cracks open Threadify real metrics and builds the onboarding flow in real-time using Codex and OpenClaw. Chapter Markers: 00:00 Intro. 02:00 Live Threadify demo. 08:30 Codex and vibe coding. 13:30 OpenClaw demo. 20:30 Clarity delivers mid-stream. 22:30 Ask AI to ask YOU questions. 30:00 Live Codex questionnaire. 37:00 Community co-design. 44:30 Accessibility. 49:00 Wrap-up. Links: Threadify https://www.threadify.app/?utm_source=lenny-youtube

Show Notes: 

peter steinberger - the creator of openclaw - just announced he is joining openai. every tech publication covered the story. but none of them use this tool every day to run a real business. lennox saint is a non-technical founder who cant code, running three AI agents with 82 custom skills to build threadify at $1,603/month MRR with 555 users. 

Threadify https://threadify.app 
OpenClaw https://openclaw.ai

Chapter Markers: 00:00 the hook. 00:30 the openclaw news. 02:00 three things. 02:30 i actually use this every day. 03:00 the full agent setup. 05:00 the real numbers. 06:00 the vulnerability. 07:30 the...