Tyap Lyap

tracking the AI shift before it becomes obvious

[Tyap Lyap @tyaplyap_ai]

@Aspect0F yea.. the answers were already there

[Tyap Lyap @tyaplyap_ai]

I GAVE CLAUDE EVERY CUSTOMER MESSAGE I HAD. ON DAY THREE IT FOUND A PRODUCT I WASN'T SELLING

my store looked organized from the outside. inside, reviews were on the product pages, questions were in email, refund reasons were in support chats. i read all of it separately and remembered almost none of it

an engineer from anthropic helped me set up claude for the store. he looked at the mess and said: "don't ask ai for product ideas. let it read what people already want"

we didn't build anything complicated. i just dumped the reviews, emails, and support exports into one obsidian folder

claude read all of it in a minute. grouped it. tagged it. connected questions from people who had never spoken to each other

on day two i asked: "what are customers trying to buy from me that i don't sell?"

it found 173 requests for the same thing - a compact travel kit made from products i was already selling separately

people had been asking me for this for two years. and i had no idea

on day three i asked claude to write one product page using their exact words. i sent it to my existing customers and the first batch sold out that evening

the video shows my obsidian folder right now. these aren't random messages anymore. it's a product map built from my own customers

it hasn't even read competitor reviews yet

anyone here run a store or a business? drop in the comments what would be the first thing you'd ask claude to find - i'm curious

[Tyap Lyap @tyaplyap_ai]

https://t.co/NhxG1rEdYm

[Tyap Lyap @tyaplyap_ai]

@kai_benetti Exactly - K3 is seriously underrated right now.
Pairing it with Hermes is where it gets really interesting: K3 handles the thinking, while Hermes turns it into a repeatable workflow. Glad you found it useful!

[Tyap Lyap @tyaplyap_ai]

https://t.co/NhxG1rEdYm

[Tyap Lyap @tyaplyap_ai]

@misat0x To my surprise, few people take advantage of the negative reviews.

[Tyap Lyap @tyaplyap_ai]

20 year old Japanese developer made $180,000 using Hermes Agent.

Every night, Hermes read thousands of bad App Store reviews and found problems people were still paying to solve.

How he did it:

He built a virtual office inside Hermes and filled it with AI agents.

Each agent had its own role: one researched ideas, another designed products, a third built prototypes, and a fourth handled marketing and customer support.
They shared the same memory, passed tasks to one another, and saved successful workflows as reusable skills.

That’s how one person ran an entire app studio like a real team.

The first idea was a design app that cleaned up rough clothing sketches and generated realistic product mockups for small fashion brands.

It made $11,000 in its first month.

Less than a year later, his one-person studio had passed $180,000 in revenue.

[Tyap Lyap @tyaplyap_ai]

@tolik12308 Once every agent has a clear role, shared memory, and its own workflows, one person can operate like a small company.

[Tyap Lyap @tyaplyap_ai]

@Aspect0F yea, madness

[Tyap Lyap @tyaplyap_ai]

15 year old kid made $28,000 selling AI-powered 3D menus to restaurants.

He built a tool that turned food photos into 3D models customers could view by scanning a QR code.

Pause at 0:04 - a full size pizza appears right on the table.

He charged restaurants $299 to set up the menu and $39 a month for updates.
His first client was a small pizza shop that added its 20 most popular dishes.

Four months later, 74 restaurants were using it, bringing him just over $28,000.

[Tyap Lyap @tyaplyap_ai]

@bygodgiven 100% he is a smart kid

[Tyap Lyap @tyaplyap_ai]

@BratDotAI The models improved enough that people stopped treating them as the main story, but the problem never disappeared - we just quietly made “verify everything” part of the workflow.

[Tyap Lyap @tyaplyap_ai]

A 14 year old student in China made $35,000 selling AI-designed Tokyo Ghoul keychains

His classmates chose their favorite Tokyo Ghoul characters, and he used AI to create each custom design.

A local shop made every keychain for $0.80, and he sold them for $8.

His first batch of 50 sold out before lunch ended.

Three months later, students from 11 schools had ordered 4,400 of them.

[Tyap Lyap @tyaplyap_ai]

Yesterday, I heard about a 14 year old student in China who made $35 000 using AI.

I’m going to speak with him personally in the next few days.

His story is incredible - I’ll share the full post soon.

[Tyap Lyap @tyaplyap_ai]

Yesterday, I heard about a 14 year old student in China who made $35 000 using AI.

I’m going to speak with him personally in the next few days.

His story is incredible - I’ll share the full post soon.

[Tyap Lyap @tyaplyap_ai]

THIS 22-YEAR-OLD CHINESE DEV TURNED 64 SCREENLESS ANDROID BOARDS INTO AN AI TESTING LAB MAKING $18K/MONTH.

Bare motherboards, no screens, no batteries — just one control monitor.

Hardware cost $1,280, software cost $49, and first-month billings hit $18,720.

Pause at 0:07 — every phone instance appears on one screen while the same app runs across all 64 boards.

Studios pay $240 per test: rack the boards, sync the tests, sell the reports.

[Tyap Lyap @tyaplyap_ai]

@MalMposha @VadimStrizheus What were you doing? I have never spent my weekly limit in 12 hours 😅

[Tyap Lyap @tyaplyap_ai]

@0xRicker Honestly, fixing the loop probably saves more tokens than switching to a better model.

[Tyap Lyap @tyaplyap_ai]

@VoltexGar The scary part is how fast confidence grows while the sample size stays tiny.

[Tyap Lyap @tyaplyap_ai]

@ScottyBeamIO The real flex isn’t Obsidian, it’s actually sticking with it for 869 days

[Tyap Lyap @tyaplyap_ai]

@kocer_eth The images are the easy part. Keeping the personality believable is way harder.

[Tyap Lyap @tyaplyap_ai]

@kocer_eth Exactly 😂 the setup itself is probably a better ad than any of the content it creates.

[Tyap Lyap @tyaplyap_ai]

THIS 21 YEARS OLD CHINESE STUDENT TURNED 24 BROKEN PHONES INTO A $22K/MONTH AI PHOTO STUDIO

24 cracked Androids. Aluminum frame. Rotating sneaker. Shenzhen dorm room.

Broken phones $86. AI software $0. First-month billings $22,140.

Pause at 0:06 — all 24 cameras fire at once.A
47 seconds later, one shoe becomes 12 product photos, three ads and a vertical video.
Local sellers pay him $18 per product. He processed 1,230 items in 30 days using hardware nobody wanted.

Buy the phones. Mount the cameras. Sell the content.
Full setup in the video.

[Diam @diamai_]

https://t.co/kyJOW5TqHP