Today, OpenAI announced the GPT4 Vision API.
So I created a tool that uses GPT4-Vision + Facial Recognition to create real-time descriptions of the feed from my webcam
If you want to see how I did it, leave a comment below and I'll send you the github repo 👇🏻
I've built hundreds of Claude Code skills.
It feels like everybody is talking about "skills", but most of them don't even work correctly...
Because while they might seem complicated - here's what nobody explains:
A skill is just a set of saved instructions. You teach Claude how to do something once, then trigger it with a couple of words instead of re-explaining it every time.
But where you save that skill decides everything...
If you get it wrong, Claude acts like the skill doesn't exist.
Most people blame the skill, or blame Claude, when the real problem is just a simple folder.
I made a full breakdown: what skills are, where to find them, how to install them so they actually work, and how to build your own
Link's in the comments ⬇️
I turned Claude Code into my own personal Youtube team.
9 custom skills, installed via a single command.
It teaches Claude how to handle research, scripting, thumbnails, titles, descriptions, presentations, analytics, and channel audits.
Here's what it actually does:
You run /youtube-init and it launches 5 parallel sub-agents that tear through your entire channel: packaging, content strategy, competitors, voice, performance. Then it generates personalized config files that power every other skill.
From there:
/youtube-research does deep 4-phase research: audience, competitors, content gaps, idea validation
/youtube-script writes full scripts with hooks, open loops, pacing, and retention optimization
/youtube-title generates and stress-tests titles against proven CTR frameworks
/youtube-thumbnail creates 4 variations using AI image generation composited with your headshot
/youtube-preview shows you exactly how your thumbnail + title combo looks in the YouTube feed
/youtube-presentation builds 80-200 slide decks with GIFs, images, and speaker notes
/youtube-description writes SEO-rich descriptions with timestamps and keywords
/youtube-analysis pulls real analytics, generates dashboards, tracks what's working
I've been using this system to produce a ton of videos on my channel. The entire production pipeline, from idea to every asset needed for publishing, runs through these skills.
Both of the attached images were 100% generated by this system.
I filmed a video breaking down how it works. Link is in the comments ⬇️
Tomorrow I'm giving away a Mac Mini.
Everyone's talking about them right now. They've become THE machine for running Claude Code, spinning up OpenClaw, and building AI agents locally.
So I decided to give one AI Innovators member a $600 Mac Mini (they are sold out everywhere)...
On top of that, every single AI Innovators member gets a free 1:1 onboarding call with me. Which is NOT a sales call. It's a live call where we work on whatever YOU want:
Getting Claude Code set up from scratch, mapping out an automation strategy for your business, debugging a workflow that's been kicking your ass, or just asking every question you've been sitting on.
The Mac Mini is a nice bonus, but the 1:1 call and the community are what people actually stay for.
The giveaway ends tomorrow at 12pm EST. Link in the comments.
What would you use it for?
I type one command into Claude and get a polished diagram in minutes.
I built a skill inside Claude Code that generates hand-drawn diagrams from a single text prompt.
Here's how it works (check out the first image):
Phase 1: Claude Code takes your topic and generates an SVG blueprint. This controls the layout: where boxes go, how arrows connect, what text says what.
Phase 2: Nano Banana 2 (Google's Gemini image model) takes that blueprint and renders it in a hand-drawn, sketchy style. Like someone drew it on a whiteboard.
The result looks very human, and is easy to read.
I use these for LinkedIn posts, YouTube thumbnails, Instagram carousels, and video explainers. Every diagram in this post was generated with the skill.
The secret sauce is the two-phase approach..
You get the precision of structured layout PLUS the personality of hand-drawn art. Neither Claude or Nano Banana can do both alone, but together they're ridiculous.
Took me about a few hours to build the skill. Now a diagram takes less than 5 minutes.
Plus, this style is *fully customizable*...
If you want the full walkthrough on how to build this yourself, I'll leave the video in the comments 👇
One carousel got me thousands of leads.
The crazy thing is, it was 100% generated by AI. Every slide, every word, every pixel.
I built a custom skill inside Claude Code that turns any topic into an Instagram carousel, styled to look exactly like Twitter/X posts. Profile pic, handle, verified badge, the whole sha-bang.
You type a topic, it writes the thread using proven frameworks. Then it renders every slide as an image, and is usually done in like 60 seconds.
The fix-it loop is what makes it insane. Something looks off? Wrong image?
Just tell Claude what you need, and it will figure out how to fix it - quickly.
And these crush. Saves, shares, comments: the algorithm loves them because the audience loves them.
I didn't hire a designer, I didn't open Canva. I wrote a Python script, wrapped it in a Claude Code skill, and now I generate these on demand whenever I want.
This is what building with AI actually looks like. Not prompting ChatGPT and copy-pasting. Building tools that produce finished assets, and actually help you move your business forward.
Comment "IG" and I'll show you exactly how I built it
I built a Claude Code skill that makes screen recordings.
& I don't know if I'll ever film an SOP again...
This skill lets Claude make smooth, polished recordings that legitimately look like they were made by a human with something like Screen Studio.
- Cursor movements that feel natural
- Zooms and pans that hit at exactly the right moment
- Bezier curves + keyframe editing for complete customization
The kind of recordings you'd swear someone sat there and carefully produced by hand. But, Claude Code did the entire thing, from start to finish.
I use it for everything now: YouTube tutorials, product demos, SOPs, etc.
Anytime I need to show something on screen without actually sitting there and recording it manually, this skill handles it.
I broke down exactly how it works in a YouTube video. I'm dropping the link in the comments ⬇️
What's the most unexpected thing you've built a Claude Code skill for?
I replaced $300/month in SaaS tools last Tuesday night.
But not with another app, with a conversation...
Claude Code has this feature called Skills. They're basically reusable capabilities you can build by just... talking to it. Describe what you want, iterate for an hour, and suddenly you have a custom tool that does exactly what you need.
Last week I built:
- A content calendar system that replaced Notion ($12/mo)
- An SEO audit and content scoring system that replaced SEMrush ($140/mo)
- A social media repurposing engine that replaced https://t.co/eX6EWPfhXR ($35/mo)
- A thumbnail generator that replaced Photoshop ($20/mo with CC)
- An analytics dashboard that replaced three different tools
Total cost: my Claude Code subscription I was already paying for.
Total time: one night.
& these aren't janky prototypes.
They're tailored to exactly how I work (no features I'll never use) & most importantly, no "upgrade to Pro to unlock this."...
The SaaS model depends on one assumption: building software is hard enough that you'll pay someone monthly to do it for you.
That assumption is breaking in real time.
I'm not saying every SaaS company is dead. But if your entire product is a thin wrapper around a workflow that an AI can just... do? You should be nervous.
The new stack isn't 47 tabs of subscription software. It's one AI that builds what you need, when you need it.
Stop paying rent on tools you could own.
I put together a free Claude Code course so you can start taking advantage of skills like these. Link in the comments. 👇🏻
@TheShortBear Isn't this referring to work performed by Claude prior to stopping for human input? I would assume the 0.01% of users are spending waayyy more time with these tools