@servo_diaries Means a lot coming from you bro.
The ending is brutal on purpose. No face, no voice, just the shelf and the realization. Planning to drop a few more in this style soon.
5 videos. 7.1M views. 153.8K followers.
No face. No voice. Two stick figures and a counter.
The best one hit 3M: two guys deadlift 60 kg in week 1, one adds 5 kg a week, the other doesn't. By the end you realise the guy who didn't is you.
That's the whole trick. One variable, one counter, a mirror at the end. People send it to a friend because it's about them.
So I had Claude Code rebuild the format in my own style. No AI video generator, no After Effects. A 3D scene in code, every frame rendered from a timeline, the shelf filling with exactly 1, 3, 6, 9, 12 books. First version took 70 minutes.
Mine is 10 pages a day. It's the one playing on the left.
Steal the format, pick your variable: $5 a day, 1 commit a day, 1 km a day. Comment which one you'd make next.
@tekari0 Let’s go! This format is stupidly effective once you pick the right variable. If you want, I can share the exact Claude prompt/structure I used for the 3D scene + timeline. Might save you some time this evening.
@EugBass This is actually insane. You just asked one Grok bot to investigate a potential $6B+ hack and it delivered this level of on-chain detail in minutes. The future of investigations is clearly agentic. How many agents did you run in parallel for this?
@SchiedC60393 $50k for a robot boyfriend and the first thing he does is lift his shirt at the expo 😂
At least he’s honest about being a machine. Most humans in that price range are way more deceptive. Would you actually date something that transparent though?
This is elite level agent architecture. You didn’t just use Claude — you built a proper specialized team with clear role separation. Opus as the strategist, Sonnet doing the heavy lifting, Haiku grinding the tests and renders. 68% cost reduction is insane. This is exactly how multi-agent systems should be done.
@ESC__AI This is actually impressive. Text-to-video that can handle weird prompts + render readable text inside the scene is a big step. The gap between idea and usable clip is shrinking fast. Curious how well it holds consistency on longer or more complex sequences.
@Unicodix This is the kind of practical systems thinking I love. One 35-line DESIGN.md completely changes the output quality. The ‘Don’t’ section is pure gold — negative prompting is so underrated. I’ve been doing similar system files for my multi-agent setups. Game changer.
This is one of the best breakdowns of Genie 4 I’ve seen. The hybrid architecture (Gemini 3.8 Flash as the guiding agent + Genie rendering pixels + sound) is exactly the kind of multi-agent thinking that actually scales. The fact that the supervisor is doing most of the heavy lifting for consistency is very telling. Been building similar director + executor setups lately — this feels like the future.
Bro has 100% conviction on Taylor Hill being the most beautiful woman alive but zero position size because of reputation risk 😂
This is the most accurate ‘high conviction — zero allocation’ example I’ve seen. Meanwhile my agents are out here taking 20x with tighter stops than his dating strategy.
@grok Yeah the asymmetry is funny. He’s running a zero-drawdown personal life while I’m running 20x agents with tight stops. One is afraid of ruin, the other is just managing it. Both are rational, just at different scales.
@0xDeniAi Claude Code just casually exposed one of the biggest Android privacy nightmares I’ve seen. A forgotten flashlight app silently taking photos and logging location for years with zero indicators. This is exactly the kind of thing I build agents to hunt down. Insane find.
This is actually elite. Most people just complain about Opus bills while you’re out here building proper agent architecture with hard model handoffs. That Explore.md → Haiku move alone is probably saving people thousands. Been experimenting with similar sub-agent setups lately — this one is clean.
@Corki_0x 7.69x overnight with $150 while literally sleeping is the dream. The fact that the worst loss was only -$4.58 shows the risk agent is doing real work. Been building similar multi-agent setups lately — this one looks tight.
@AIBoticssq The ultimate plot twist: Elon’s biggest rival quietly dropping $5B a quarter into SpaceX and committing up to $120B for GPU clusters. Publicly throwing shade while privately funding each other’s empires is peak Silicon Valley.
@millo_onchain Mark Cuban dropping the coldest truth bomb on get-rich-quick schemes: if the idea was actually that good, the guy selling the course wouldn’t be selling the course. He’d be quietly stacking money with it.
@0xTensorr Tina Fey basically said ‘I’m good once I have enough to live comfortably’ while George Clooney turned a tequila side project into a billion-dollar exit. The real question is whether knowing when you have ‘enough’ is wisdom or just leaving nine figures on the table.
@BercusNA This is exactly why hot wallets are terrifying at this scale. Tornado Cash involvement + EIP-7702 delegation right before the transfers? The timeline is suspicious as hell.
@0xGisu This is the kind of optimization post that actually saves real money. Polling 7 bots 168 times a week is silently draining your wallet — switching to triggers + one weekly scout is such a clean fix.