@mjstallone @deedydas Say more please. I'm curious to understand what you mean
English
MICHΛΞL TRΞZZΛ
367 posts
MICHΛΞL TRΞZZΛ
@MichaelTrezza
🤖 AI Transformation Partner & Automation Expert • 🧭 CEO at Lithyem • 🦉 Exited Founder • 🚀 Futurist • 🎤 TEDx Speaker
La Jolla, CA Katılım Ocak 2010
200 Takip Edilen107 Takipçiler
@deedydas The crazy thing about DeepSeek isn’t the OCR (it’s just good), it’s that they showed LLMs can use a higher-level information representation than language to store more data with equivalent accuracy.
English
DeepSeek-OCR is the best OCR ever.
It parses this extremely hard to read handwritten letter written by mathematician Ramanujan in 1913 with a frightening degree of accuracy.
Not perfect, but beats former best dots ocr. Bonus points if you can spot the errors.
Try it here:
English
credit to @OpenAI Sora2 and @pinai_io voice keyboard
My Sora User Name:
My PIN AI invitation code: pCg9Vax
PIN AI Download link: pinai.com/product
First 3 users who reply will get SORA invite code
English
Humans shape tools. Tools shape us. medium.com/p/humans-shape…
English
Every failed AI project I’ve seen has one thing in common:
Nobody defined the problem before building the solution.
It’s a clarity problem. A business process problem. It’s NOT a tech problem.
Most AI projects fail because the input was vague. The process was s**t. The logic was fuzzy. So the code just automates dysfunction.
We don’t write a single line of code, or ideate solutions until we:
→ Map the processes
→ Find the friction
→ Engineer better workflows (on paper or more typically, Lucid)
→ Test with real, messy data
Building systems that scale how your best people think, is a business process project first, NOT a technical one... and that's where most get it completely wrong.
We’ve built tools that:
→ Saved a home improvement company $180k per year by repetative sales team tasks
→ Tripled output of a manufacturing company with the same headcount
→ Recovered 20+ hours a week in admin tasks for a financial services company
Code is the last step. Clarity is the first.
English
I did 40 hours of work in a day.
Using Manus AI:
✅ Logged into a web portal for me
✅ Paginated through 49 pages of potential contacts
✅ Clicked into the profile details of one
✅ Systematically analyzed the bio of each contact
✅ Cross checked them against a deep knowledge base of the services I offer to determine a potential fit
✅ Provided reasoning for the determination, and put this all into a spreadsheet
Then I had Claude and the Gmail integration check if I'd ever emailed any of the contacts and update the spreadsheet...
Next, I built a custom MCP SalesForce integration, with the help of Claude, to check if the contacts are in my CRM already… and add them if not.
That was before lunch.
Later, I plugged in a handful of useful YouTube videos into my n8n tool which grabbed the transcripts and put them into a vector database.
I pointed Manus at a few trusted blog sources which it aggregated tidily.
Then I used ChatGPT and Manus to revise my content creation guide with the latest best practices and linkedin algorithm changes.
Next I had it look at 6 weeks worth of my medium.com posts, analyze them for new angles through my revised content guide and create a content calendar.
In a single day I accomplished easily 40 hours or more of focused work.
All of this should’ve taken a week.
Maybe two.
Instead, it took ONE day.
… and a stack of AI tools most CEOs still don’t know exist or how to use.
But sure...
AI is “overhyped.”
It “hallucinates.”
It “can’t be trusted.”
Keep saying that.
Meanwhile, I’m going to keep building.
🧠 What’s one thing in your business that would change if you could 10x your productivity?
👇 P.S. If you want this kind of leverage, you know where to find me.
English
AI isn't moving fast because it's trendy. It's moving fast because whoever builds first controls the rules.
Global power shift > hype cycle.
"Slow down AI" sounds good until you realize:
- China won’t.
- Rogue devs won’t.
- Venture-backed labs definitely won’t.
Ethical builders slowing down just hands the wheel to someone else.
There is no FAA for AI. No FDA. No central body.
It’s a free-for-all arms race. Anyone pretending there’s a kill switch is lying.
We can’t wait for safety to be perfect. We test, we monitor, we iterate.
That’s how every real-world system matures.
That’s how AI has to evolve too.
AI is non-deterministic. It won’t give the same answer. It often can't explain why it did what it did.
Designing guardrails around that? Hell of a challenge.
Distrust in tech is earned. But total cynicism isn’t a strategy.
Some of us are actually building systems that work.
Not for hype. For leverage.
AI will multiply whatever you point it at. Clean process? It scales it. Broken system? Now it breaks faster.
Be careful what you automate.
The speed of AI isn’t the real threat. Standing still is.
Ask yourself: Is your business built to keep up?
English
Found the craziest N8N AI automation… 🔥
This system converts ANY concept into trending AI videos automatically.
Zero manual work. Zero hiring. Zero headaches:
→ Brainstorms content concepts via ChatGPT integration
→ Generates insane visuals through Veo 3 + FAL API (think: Dragons doing cooking tutorials)
→ Handles lip-sync, animation, rendering & file storage automatically
→ Records every prompt, description, video link + scene data in Google Sheets
→ Can merge multiple clips into extended content
→ Primed for instant publishing across TikTok, Instagram, YouTube & more
Entire system runs on n8n automation. Zero human intervention needed.
Like + reply “VIDEO” & I’ll send you the FULL workflow + setup FREE!
English
The future is co-piloted: when expertise becomes ambient medium.com/p/the-future-i…
English
Today is the worst AI will ever be. medium.com/p/today-is-the…
English
We don’t watch movies and complain about every CGI scene, shouting, “That’s not real!” We care about the storytelling, the emotions stirred, the experience.
Why here then?
English