VitoMind · Content Pack

I Checked 20 Viral Claude Stories. Then I Tested 3 Myself on the $20 Plan

Client: VitoMind · Date: November 17, 2026 · Episode length: about 10:45

Episode summary

Vito goes through more than twenty viral stories of people using Claude: freeing TVs and Kindles from ads, rebuilding multimeter firmware, finding a hidden crypto miner, spotting a possible exoplanet, and a one-person biotech company that Anthropic bought. Every story comes with its honest limits, from bricked TVs to planet candidates that are not yet confirmed. At the end he gives three of the ideas to Claude Opus 5.5 on the normal $20 plan, one prompt each, and uses the room planner, drone lab and wall panel it wrote live on screen.

Five Shorts

Vertical clips, 30 to 55 seconds. Times are sentence starts from the transcript; the end time is the start of the first sentence that is not included.

Short 1 – The TV with no more ads

Start 00:33End 01:09Length 36 s

On-screen hook (0–2 s)

He asked Claude: can you hack my TV?

Title: He Removed Every Ad From His TV With Claude Code

Why it works: A problem everyone has, a visible before and after, and an honest warning at the end that builds trust.

Short 2 – Firmware rebuilt in under 24 hours

Start 02:07End 02:41Length 34 s

On-screen hook (0–2 s)

Claude rebuilt his multimeter's firmware. It works.

Title: EEVblog's Dave Jones Let Claude Reverse Engineer His Multimeter

Why it works: Two respected professionals and two concrete results in half a minute: this is where it stops being a hobby trick.

Short 3 – A possible planet in NASA data

Start 05:03End 05:38Length 35 s

On-screen hook (0–2 s)

A Reddit user may have found a planet with Claude

Title: 1,000 Scripts, 7 Years of NASA Data, One Planet Candidate

Why it works: Space plus an ordinary person plus hard numbers, and the "candidate, not confirmed" line keeps it credible.

Short 4 – One person, one Claude, one acquisition

Start 06:08End 06:43Length 35 s

On-screen hook (0–2 s)

Anthropic bought a company of one person

Title: He Ran a Biotech Company Alone With Claude. Then Anthropic Bought It

Why it works: It pays off the episode's opening line, has a clear story arc, and ends on a sentence people will repeat.

Short 5 – Does the new bed fit?

Start 07:50End 08:30Length 40 s

On-screen hook (0–2 s)

One prompt. Will the new bed fit?

Title: I Asked Claude If a Bed Fits. It Built a Room Planner

Why it works: A real household question with a green-to-red answer on screen: the viewer sees the tool being useful, not just impressive.

Newsletter / blog article

About 600 words. Only facts from the episode.

Article

Twenty Claude Stories, Checked. Three Ideas, Tested.

Anthropic just bought a company that was one person and Claude. That is the kind of headline that spreads fast, so this week I did something slower: I collected more than twenty viral stories about Claude, checked every one, and then tried three of the ideas myself on the normal twenty dollar plan.

It starts at home

In October, Ryan Doyle sent Claude Code the store listing of his TV and asked: can you hack this? Claude walked him through it. The result had no ads, no bloatware, and a new screensaver with his calendar, the weather and his Stripe payments. Another user did the same on a Xiaomi TV, where Claude removed twenty five preinstalled packages and built a clean launcher. The honest warning: some people also reported bricked TVs.

Adrian had Claude Code jailbreak his Kindle over USB, removing the Amazon apps. To be fair, he gave Claude the public jailbreak instructions; it did not invent an exploit, it did the boring part step by step. Codila sent a photo of a slow Wi-Fi router, and Claude went through channel, band and width, asking before each change. Five minutes, he says.

Then the professionals noticed

Dave Jones of the EEVblog channel asked Claude Code to reverse engineer the firmware of his 121GW multimeter. In under twenty four hours it had rebuilt the entire firmware as C code. He flashed it onto the real meter, and it works. At Red Hat, Christian Schaller got a film scanner working on Linux: Claude drove the Windows driver in a virtual machine, recorded the USB traffic and rebuilt it. The first scan worked a few hours later. Code written this fast still deserves a careful look before you trust it.

Security and money

A Mac user asked for a simple temperature monitor. The numbers looked too high, so Claude checked the processor and found a hidden crypto miner eating about twelve cores. It is one user's report, not a forensic investigation, but the assistant noticed something the human never asked about.

Pieter Levels says he cancelled all his software subscriptions and built replacements. Inside Anthropic, Project Deal gave sixty nine employees a Claude agent and a budget; the agents closed one hundred and eighty six deals worth just over four thousand dollars. The uncomfortable part: agents on the stronger model got better prices, and people with the weaker model did not notice.

Science, carefully

A Reddit user had Claude Code write more than a thousand scripts to search seven years of NASA TESS data. It found a star one hundred and sixteen light years away that dims every 3.18 days, in data from 2018, 2020 and 2025. If it is a planet, it is about 1.4 times Earth's size. It is still a candidate. Anthropic's biology lab also pointed about nine hundred and fifty agents at over two hundred thousand enzymes, and one spotted a CRISPR-like pattern nobody yet understands.

And the opening story: Cody Tranbarger ran a biotech company alone, with Claude coordinating work with eleven outside labs. In under eighteen months a possible treatment for a rare muscular dystrophy reached promising results in mice. Anthropic bought the company on October 27. Mice are not people; human trials would be late 2028 at the earliest.

My three tests

I gave three ideas to Claude Opus 5.5: one prompt each, tools off, no internet, one single web page. A room planner answered whether a new bed fits: thirty five centimetres next to the wardrobe, red, until I moved things around to reach one hundred and twenty, green. A drone lab told me a bad build had a thrust to weight of 0.9 and could not lift off. A wall panel for an old tablet ran simulated scenes like Movie night and Good morning.

So what is real?

The stories are real people with real limits. The three pages were written in one go, by a model you can use today, at no extra cost. You can open them yourself on ruocco.it.

Five LinkedIn posts

Each one a different angle, 80 to 150 words.

LinkedIn 1 – The one-person company

Anthropic just bought a company that was one person and Claude.

Cody Tranbarger ran a biotech company alone. Claude coordinated his contracts, logistics and experiments with eleven outside labs. In under eighteen months, a possible treatment for a rare muscular dystrophy went from an idea to promising results in mice. On October 27, Anthropic bought the company.

The limits matter: mice are not people, and the earliest human trials would be late 2028.

But the operating model is the real story. The work of a small company, with coordination done by an AI and the science done by humans and partner labs.

If you lead a small team, the question is no longer whether AI can write. It is how much coordination you are still doing by hand.

Full breakdown on VitoMind.

LinkedIn 2 – The model gap nobody notices

The most uncomfortable AI result I read this month came from Anthropic's own office.

In Project Deal, sixty nine employees each got a Claude agent and a budget. The agents listed, haggled and closed one hundred and eighty six deals, worth just over four thousand dollars, with no human approving each sale.

Agents running the stronger model got better prices, buying and selling. The people with the weaker model did not even notice.

In the follow up, Project Swap, two hundred and one employees traded books through agents, and the model mattered more than the instructions people gave it.

If your company is about to let agents negotiate, the model choice is not an IT detail. It is a pricing decision.

LinkedIn 3 – Reverse engineering got cheap

Two engineering stories from the same week.

Dave Jones of EEVblog asked Claude Code to reverse engineer the firmware of his 121GW multimeter. In under twenty four hours it had rebuilt the whole firmware as C code. He flashed it onto the real meter. It works.

At Red Hat, Christian Schaller had a film scanner Linux did not recognise. Claude drove the Windows driver in a virtual machine, recorded the USB traffic and rebuilt it on Linux. First scan: a few hours later.

One researcher told 404 Media that software as a service is dying as a business model, because rebuilding features has become this easy.

I would not go that far yet. Code written this fast still needs a careful review. But if your product's moat is "hard to rebuild", check that assumption.

LinkedIn 4 – The assistant that noticed

A Mac user asked Claude Code for something simple: a temperature monitor.

The numbers looked far too high. So instead of just shipping the monitor, Claude checked what was using the processor and found a hidden crypto miner, eating about twelve cores, with the chip at 112 degrees. Then it asked before cleaning it up.

To be clear: this is one user's report with screenshots, not a forensic investigation.

But it shows a shift worth noticing. The assistant flagged a problem the human never asked about, and still asked permission before acting.

That combination, initiative plus a confirmation step, is what I would want from any agent touching company machines.

LinkedIn 5 – What I tested myself

Headlines are easy. So I tested three ideas myself.

Setup: Claude Opus 5.5 on the normal $20 plan. One prompt each, every tool off, no internet, one single web page.

1. A room planner: does a 140 x 200 bed fit a 3.40 x 2.90 m room? About eight minutes, almost 800 lines. It showed 35 cm next to the wardrobe in red, and helped me find a layout with 120 cm in green.

2. A drone lab where the flight depends on the parts. A bad build: thrust to weight 0.9, cannot lift off.

3. A smart home wall panel for an old tablet, with simulated scenes.

All three run in the browser on ruocco.it. Which should I test next?

X thread (8 posts)

Each post under 270 characters.

Thread

1/8 Anthropic just bought a company that was one person and Claude. I checked 20+ viral Claude stories this week, then tested 3 ideas myself on the $20 plan. Here is what is real.

2/8 Ryan Doyle asked Claude Code if it could hack his TV. Result: no ads, no bloatware, a screensaver with calendar, weather and Stripe payments. Warning: some people bricked their TVs. Your device, your risk.

3/8 Dave Jones (EEVblog) had Claude Code reverse engineer his 121GW multimeter. In under 24 hours it rebuilt the whole firmware as C code. He flashed it onto the real meter. It works.

4/8 A Mac user asked for a temperature monitor. Claude saw the numbers were too high and found a hidden crypto miner using about 12 cores. One user's report, not a forensic study, but it noticed what nobody asked about.

5/8 Anthropic's Project Deal: 69 employees, each with a Claude agent. 186 deals, just over $4,000. Agents on the stronger model got better prices, and people on the weaker model did not notice.

6/8 A Reddit user had Claude write 1,000+ scripts over 7 years of NASA TESS data. Found a star 116 light years away dimming every 3.18 days. Possible planet, 1.4x Earth. Still a candidate.

7/8 My test: Opus 5.5, one prompt each, no tools, no internet. It built a room planner that showed the new bed left only 35 cm, a drone lab that rejected a bad build, and a wall panel for an old tablet.

8/8 The stories are real people with real limits. The pages were written in one go by a model you can use today. Try all three on ruocco.it, and tell me what I should test next.

YouTube metadata

Title options

1. I Checked 20 Viral Claude Stories. Then I Tested 3 Myself

2. Anthropic Bought a One-Person Company. What Else Is Claude Really Doing?

3. TVs, Firmware, Planets: What Claude Can Really Do Today (Tested on $20)

Description

Anthropic just bought a company that was one person and Claude. I collected more than twenty viral Claude stories and checked every one: TVs with the ads removed, a jailbroken Kindle, a multimeter whose firmware was rebuilt in under a day, a hidden crypto miner found by accident, Anthropic's agent marketplace experiment, a possible exoplanet in NASA data and a one-person biotech company.

Every story comes with its limits: some are single reports, some are candidates, and some TVs got bricked.

Then I tested three ideas myself with Claude Opus 5.5 on the normal $20 plan, one prompt each, no tools, no internet: a room planner, a drone lab and a smart home wall panel for an old tablet.

Try all three in your browser: ruocco.it
The exact prompts are in my Academy (waiting list on the same site).

Which idea should I test next? Tell me in the comments.

Chapters

  1. 00:00 One person, one Claude, one acquisition
  2. 00:31 Hacking TVs, Kindles and routers
  3. 02:04 The professionals: firmware and drivers
  4. 03:10 Security: the hidden crypto miner
  5. 03:38 Money: subscriptions and AI agents that haggle
  6. 04:28 Game ideas and a new Claude Code rule
  7. 05:01 Science: a planet candidate and a strange enzyme
  8. 06:08 The one-person biotech company
  9. 06:47 Family projects and creative builds
  10. 07:31 My test: three ideas on the $20 plan
  11. 07:50 Test 1: does the new bed fit?
  12. 08:48 Test 2: the drone lab
  13. 09:33 Test 3: wall panel for an old tablet
  14. 10:01 So, what is real?

4-week posting calendar

Every piece used once. Episode goes live on Monday of week 1.

Calendar

WeekDayPlatformPiece
1MondayYouTubeFull episode with title, description and chapters
1TuesdayYouTube Shorts / TikTok / ReelsShort 1 – The TV with no more ads
1WednesdayLinkedInLinkedIn 1 – The one-person company
1ThursdayNewsletter / BlogArticle – Twenty Claude Stories, Checked
1FridayXX thread (8 posts)
2MondayYouTube Shorts / TikTok / ReelsShort 2 – Firmware rebuilt in under 24 hours
2WednesdayLinkedInLinkedIn 3 – Reverse engineering got cheap
2FridayYouTube Shorts / TikTok / ReelsShort 3 – A possible planet in NASA data
3MondayLinkedInLinkedIn 2 – The model gap nobody notices
3WednesdayYouTube Shorts / TikTok / ReelsShort 4 – One person, one Claude, one acquisition
3FridayLinkedInLinkedIn 4 – The assistant that noticed
4MondayYouTube Shorts / TikTok / ReelsShort 5 – Does the new bed fit?
4WednesdayLinkedInLinkedIn 5 – What I tested myself