The 10M+ AI / tech video

Hafu Go · November 1, 2026 · the career video for December, before CES

The short answer

• The hard part first.
  • Since 2025, 7 of your 33 long-forms cleared 10M: Dirt Into a $1,000 Ball, two puzzle videos, $1 vs RC Plane, Banned, Unbreakable, and Level 1 to 100 Science Tricks.
  • None of the 7 was about AI or futuristic tech. Your AI, futuristic and how-it's-made videos since 2025 all landed below your normal (How Smartphones Are Made is at 0.36x). Gadget tests did fine (Banned, Unbreakable), but they weren't AI or futuristic.
  • 10M is about 2.4x your 2026 median (4.2M).
  • So the AI video has to ride one of your proven 10M shapes. A new topic alone won't get it there.
• Pick: "I Tried to Turn Sand Into a Computer Chip." It's your biggest shape: "How I Turned Dirt Into a $1,000 Ball" is your most-viewed long-form ever, 37.9M, 6x your 2025 median. Applied to the thing AI runs on:
  • It's positive, it's impact (the foundation of AI), and it's a natural chip and PC sponsor story.
  • It has one honesty problem, covered below: your own sand can become real silicon, but not a whole chip. The title says "Tried" for that reason.
  • It needs a lab partner, so it may land in January.
• Fastest to ship by Dec 31: "Can AI Solve Puzzles That Look Easy?" Your puzzle franchise with AI as the contestant, and an AI company as the natural sponsor. Your latest two puzzle videos were under your normal, though.
• The brand piece: "Can AI Replace Me?" The Mediastorm idea at US scale. It's the best for impact and AI sponsors, and the least likely to clear 10M.
• Dropped: Kung Fu Master vs Robot. Fight robots are steered by people with controllers. Unitree and AgiBot already did kung fu robots at scale, including with monks at the Shaolin Temple. And the US banned imports of new Chinese humanoids in July.

Your 10M math

All numbers are fresh as of November 1.

Your AI, futuristic and how-it's-made long-forms since 2025 all landed under your normal. Plain gadget tests sat around normal (I Tested Every Science Gadget on Amazon 1.20x, I Tested Every Satisfying 3D Print 1.02x).

"I Tested Impossible Product Claims" got 9.66M, and the later AI version got 4.50M. But follow-ups to your biggest hits usually drop even without AI (Puzzles IMPOSSIBLE 26.4M, then HARD 5.21M; RC Plane 17.0M, then RC Boat 5.35M; Banned School Gadget 12.6M, then Banned School Hacks 1.88M, though that one is only three weeks old), so the AI effect can't be separated from the follow-up effect.

What actually reaches 10M in AI and tech

I pulled the latest 400 uploads from 33 big creator and tech channels (MrBeast, Mark Rober, Dude Perfect, Ryan Trahan, Nick DiGiovanni, Airrack, Mrwhosetheboss, MKBHD, Veritasium and more), plus 34 searches through YouTube's own search. "x" means views compared to that channel's median for the same year.

1. For mega-channels, 10M is normal, so their robot videos prove less than they look.
   • The 58 robot-titled videos from these channels have a median of 0.96x.
   • The 27 AI-titled videos have a median of 0.74x, with 8 at or above normal. The top one is a Kurzgesagt AI explainer (3.15x), from a channel built for explainers.
2. Person-vs-machine often works for them, but at their scale.
   • 4 of the 7 person-vs-robot titles beat their channel's normal: Dude Perfect "Dude Perfect vs Robots" 32.3M (3.2x), Nick DiGiovanni "I Cooked Against Robots" 80.0M (2.3x), Mark Rober's Ronaldo goalie robot 51.7M (1.7x) and his NFL kicker robot 102.4M (1.7x).
   • The biggest, MrBeast "World's Strongest Man Vs Robot" (150.8M), was 0.74x.
   • Outside these channels: ZHC "Human VS Robot For $100,000" got 8.9M, and a creator-made human-vs-robot cage fight got 1.2M.
3. What over-performs is a big, simple shape with tech inside it. The topic alone doesn't carry a video.
   • Veritasium's chip story "The Closest Thing We Have to Alien Technology" (65.0M, 5.2x).
   • Ryan Trahan's survival challenge "I Survived 50 Hours in Apple Vision Pro" (40.1M, 1.2x).
   • A big tech channel's chip-making video, Linus Tech Tips "I made my own RAM! - Micron Factory Tour", got 2.97M.

The video you liked, decoded

Mediastorm's "If AI can replace me, what's my meaning?" has 440K views on their own YouTube channel (984K subscribers) and 11.9M on Bilibili, where their main audience is.

• What happened: a one-minute AI-vs-artists cold open sets up the question. Then a true story carries it: a singer losing his hearing, and the music video the studio built for him by hand (about a year of building scenes).
• The answer at the end: "maybe only lived experience."
• The ending is acts, not arguments: a concert, a donation, and the singer, recovered, proposing to his girlfriend.
• What made it work: a host admitting fear ("it frightens us"), and real stakes for a real person.
• What to keep for yours: one real person with something at stake, and an ending that's an act.

Where the tech sponsor money actually is

• Products that help the creator pull off the feat. Google's Tensor chip and 5G "help make this possible" in Mark Rober's Ronaldo video. T-Mobile "supported" his space video. Google Fitbit was one of four paid segments in a MrBeast video. Build 2 to 4 of those moments into the film and sell them before shooting.
• Robot makers send hardware, not cash. 1X approached Nick DiGiovanni with its NEO robot and sent one to MKBHD. I found no paid deal from a robot maker in 2025-26.
• Chinese robots carry a US risk now. The FCC banned imports of new Chinese humanoid models in July 2026, so US tech sponsors may avoid a video built around one.
• Chips are a clean sponsor story. Intel has made its own "From Sand to Silicon" videos (5.5M), and every PC, phone and AI company depends on chips.

The three concepts

1. I Tried to Turn Sand Into a Computer Chip ⭐ pick

• The shape: your Dirt Into a $1,000 Ball transformation. One handful of beach sand, and a question: how far can you take it toward the chip inside your phone?
• The honest part (say it on camera): you can turn sand into real silicon yourself, but only a few grams of powder mixed with leftovers. Chips need silicon about 99.9999999% pure, grown into a big crystal. So your sand goes as far as it truly can, and the chip stage happens at a real lab on commercial silicon, shown openly. That's why the title says "Tried."
• Stages (draft):
  1. Dig the sand on a beach.
  2. Burn it into silicon yourself. Sand plus magnesium, outdoors, with professionals. This is the fire beat. The acid wash afterward releases a gas that can ignite in air, so it's a lab-supervised step.
  3. Prove your silicon is real. Use it as the detector in a homemade crystal radio. When a station comes through, beach sand is now a working electronic part. This is the mid-video payoff, and it's yours. Early radios used silicon crystals this way. Test it before the shoot: fine powder may need pressing into a pellet (melting it into a lump takes a lab furnace at about 1,400°C), and if it fails, that failure is honest content too.
  4. Take it to the factories. See the 99.9999999% purification, the glowing crystal being pulled, and the wafer slicing. Those happen at polysilicon and wafer makers, not at chip labs; a crystal-pulling plant on your China trip can show it (the ones there mostly make solar-grade silicon, so say which).
  5. Make your own chip in a university cleanroom on a commercial wafer: a simple transistor chip, not a phone processor. It switches an LED on.
  6. The final shot: your sand, your silicon, your radio and your chip, next to the chip inside your phone.
• Retention: a progress counter on screen ("how pure is it now?"), a fresh visual at every stage (fire, the radio coming alive, the glowing crystal, cleanroom suits, microscope shots), and the payoff planted in the first minute.
• Comps:
  • Yours: "How I Turned Dirt Into a $1,000 Ball" 37.9M.
  • Veritasium "The Closest Thing We Have to Alien Technology" 65.0M.
  • Branch Education "How are Microchips Made?" 11.9M.
  • Smaller homemade-chip videos: Sam Zeloof's homemade silicon chips 2.5M, Dr.Semiconductor "Making RAM at Home" 2.7M.
  • A chip-making visit on a big tech channel: Linus Tech Tips "I made my own RAM! - Micron Factory Tour" 2.97M.
• Sponsor moments: the chipmaker or university lab, the PC or phone brand where the chip "ends up," and the microscope and camera gear.
• Risks:
  • The demand evidence is one huge Veritasium video plus your own transformation hit, not a pile of 10M+ creator versions.
  • The fire and acid steps need professional safety.
  • Lining up a lab takes weeks, so this is likely a January video unless a partner is locked by mid-November.

2. Can AI Solve Puzzles That Look Easy?

• The shape: your puzzle franchise (26.4M and 11.9M in 2025), with AI as the contestant.
• How it plays: the smartest AI models, seeing through a camera and answering by voice, against your puzzles, round by round, against the smartest humans you can find. There's a scoreboard and viewers play along.
• Comps: your puzzle hits above. Zach King "Can VFX Artists Beat AI?" 2.7M (1.97x his normal).
• Sponsor: an AI company showing off its model.
• Why it's here: no access needed, so you can film and ship by Dec 31.
• Risks:
  • Your two most recent puzzle long-forms were under your normal ("Puzzles That Look Easy But Are Actually HARD!" 5.21M, "I Bought Every Impossible Puzzle!" 5.17M).
  • A sponsor's model may lose on camera, or get shown next to rivals.
  • AI answering on a screen is less visual than a physical puzzle.

3. Can AI Replace Me?

• What it is: an AI Hafu built from what you already have (your approved voice clone, your head scan, your AI editing pipeline). You test it on harder and harder audiences: your team, your family, fans at a meet-up, strangers, and finally an AI-made video on your channel.
• The ending: the Mediastorm answer, as an act only the real you can do.
• Comps:
  • Danny Gonzalez "I Made a Fake AI" 11.3M (a comedy version).
  • GothamChess "Gotham vs. Gotham-Bot: Can I Beat My AI Clone?" 9.6M.
  • Sambucha "I Replaced Myself With AI For An Entire Month And Nobody Noticed" 6.4M.
• Sponsor: the voice, video and writing tools are all products that make the feat possible.
• Risks:
  • YouTube requires an AI label at upload for a realistic AI version of a person, so "reveal at the end" can't work on your channel.
  • Testing the clone on fans and strangers needs their consent.
  • The comps top out around 10M. It's the best impact and branding video here, and the least likely to clear 10M.

Considered and dropped

• Kung Fu Master vs Robot:
  • Fight robots are steered by people.
  • It's crowded: Unitree's "Kung Fu Bot" has 27.2M, and AgiBot trained robots with monks at the Shaolin Temple in February.
  • The US import ban on Chinese humanoids.
  • Only 4 of your 10 kung fu long-forms reached 3x.
• $1 vs $1,000,000 futuristic tech: MrBeast did "$1 vs $1,000,000,000 Futuristic Tech!" in January 2026 (217.4M).
• A day in a futuristic city or robot hotel: your Futuristic City video is at 0.68x.
• A straight AI or chip explainer: that works for Kurzgesagt and Veritasium, but your how-it's-made video is at 0.36x.

Timeline

• By Nov 15: pick the concept and lock partners. For #1, that's a polysilicon or wafer maker plus a university cleanroom or chipmaker; for any of them, 2 to 4 sponsor moments.
• If #1's lab isn't locked by Nov 15: ship #2 by Dec 31 so a tech video is live before CES, and use CES (Jan 6-9) to meet chip partners for #1.
• December: shoot (China trip in early December). Edit through the holidays.

Your calls

1. Sand-to-chip as the career video, even if it lands in January? Or #2 to hit Dec 31?
2. Should I build the partner list now (silicon and wafer makers, university cleanrooms, chipmakers, and the 2 to 4 sponsor moments with brands to pitch)?

Numbers and method: details page