pulmencrFOMO

Apple has mapped over 400,000 miles of roads with camera-equipped cars None of them ever touched this street in Delhi Because cars can't fit down narrow alleys, market lanes, or footpaths So a guy walks it instead Strapped to his back is a rig worth roughly $50,000 Six cameras and lidar sensors spinning in a full 360 degrees, mounted on a curved pole rising straight off the backpack Scanning every building, every sign, every corner in real time while he just walks iPad in one hand tracking the route Weaving past tuk-tuks and motorbikes like it's a normal Tuesday The street behind him just became data Every zoomed-in street view you've ever scrolled through probably started with someone doing exactly this

@Bober_smart yes, they're moving insanely fast in AI

@pulmencr I don't know why, but Chinese people are very smart

A guy from China is making $2,700 a month by destroying Super Mario 64 with Claude He doesn't make mods. He just takes normal gameplay, runs it through Claude to set up the process, then feeds it into an AI video model. The game completely loses its mind Turtles pull out guns and shoot at Mario. Penguins turn into castles while flying. Every jump turns into a glitchy nightmare that breaks the original physics The wildest part - it takes him less than 5 minutes per video No editing. No complicated setup. Claude handles the pipeline, the AI model handles the visuals, and he just posts whatever comes out While most creators spend hours making content, this guy casually ruins childhood classics in under 5 minutes, and gets paid for it

@XXIfomo yeah, we're entering a really weird era

@pulmencr game took years to build. the ruin takes 5 minutes to render. one required a team. the other requires a prompt

@hitu_monke sounds way more fun than a normal 9 to 5 work

@pulmencr making cash just because an AI breaks the game is the best job in the world

@0xdimix imagine where this will be in six months

@pulmencr watching it panic in gta was gold, but building a real-time agent in 38 mins with no coding background is actually insane. can't wait for v2

A 21-year-old guy from South America built an AI driving coach in 38 minutes and decided to test it the hardest way possible He put on a blindfold and let the AI guide him through GTA 5 with zero vision, just voice commands First instruction sent him straight into a wall The AI immediately started second-guessing its own directions, mixing up left and right, contradicting itself mid-sentence At one point it admitted it didn't actually know how to drive either By the time police showed up in-game, they'd already hit dozens of pedestrians without either of them realizing it The AI's escape plan under pressure was just one word repeated - run He built this in under an hour using Claude Code to handle the voice logic and real-time response generation, no coding background going in The test wasn't really about driving It was seeing how an AI handles total pressure with zero visual data and everything moving in real time This version was a mess, and honestly kind of hilarious to watch fail But it's version one. The next one will actually work

@0xkerazcity kinda reminds me of people selling minecraft servers before everyone else

A 20-year-old is already making money from GTA 6. The game hasn't even launched yet. While millions are waiting to play, a small group is already building products they'll eventually sell to those players. Rockstar quietly opened an official marketplace for paid mods, and it's still early enough that almost nobody is competing there. Claude can turn plain English into working Lua, making game development faster and dramatically cheaper than it was a year ago. By the time GTA 6 reaches tens of millions of players, the best creators could already have customers, reviews, and recurring revenue. Most people will play GTA 6. A much smaller group will build businesses around everyone else playing it.

@exploraX_ everyone laughed at chinese models a year ago

china isn’t playing, kimi-3 re-made it own marketing video in just 20mins+. 🤯🤯 maybe this model is the actual chinese version of fable 5.

@DataChaz now i know what i'm doing this weekend

YOU CAN NOW LEARN FULL CINEMATIC AI PRODUCTION FOR FREE 🚨 Higgsfield just dropped `Higgsfield Academy` to help you skip the guesswork and start producing. Instead of just guessing prompts, you learn the actual pro workflow. What’s inside: > 2 totally free courses (Onboarding + Middle Guide) > a certificate for your LinkedIn > FREE credits to put what you learn to work immediately Stop poking at a prompt and hoping for the best. Master the craft 👀↓

@insomnia_vip indie devs keep carrying the gaming industry

THIS WEIRD $6 INDIE GAME IS TAKING OVER THE INTERNET What looks completely random at first turns into one of the most creative hide and seek mechanics released this year where every match becomes a battle of camouflage imagination and pure chaos The entire game revolves around painting your own character to perfectly blend into the environment which creates moments that are somehow hilarious satisfying and surprisingly competitive at the same time Big studios keep chasing bigger budgets while small teams keep proving that one original gameplay idea can capture millions of players around the world Sometimes the simplest ideas create the biggest games

@beamnxw that's both impressive and a little unsettling

WE HAVE IRL WALL HACKS NOW: HOW TECH STARTUPS ARE USING WIFI ROUTERS TO SEE THROUGH CONCRETE WALLS Well, if you think wall hacks are just something annoying you run into while playing Call of Duty, you need to wake up because it is officially a thing in the real world now. Tech startups are finding ways to turn your ordinary wireless router into a literal x-ray machine. They are using basic Wi-Fi signals to watch people right through solid concrete walls A company called ZaiNar is mapping the exact radio frequency distortions that happen whenever a human body moves through a wireless network. The system tracks how the signals bounce around the room, feeds that raw RF data straight into a machine learning model, and instantly spits out a highly accurate, 3D skeletal map of whoever is inside the building. No cameras required, no heavy sensors, and zero direct line of sight The real flex here is that it is a massive play for defensive tech and localized tracking pipelines. Instead of spending a fortune on complex satellite sweeps or intrusive physical cameras, operators can just monitor an entire structure using the existing mesh routers already sitting on the tables. They recently pulled in a massive 100 million dollar investment round to scale the platform, which is an insane valuation for a hardware space that most people do not even realize exists yet Tactical monitoring? Data privacy? - Bye bye.. It proves you can build high-fidelity motion tracking networks using nothing but the ambient waves floating through your house right now

@luntych13 that's probably the most overlooked part of the whole ai race

THE SCARIEST PART ABOUT THE AI BOOM ISN’T THE TECHNOLOGY. IT’S WHO’S PAYING FOR IT. Everyone celebrates another $100B investment. Another giant data center. Another funding round. Another record valuation. Very few people stop to ask the only question that actually matters: Who wrote the check? The biggest fortunes in history weren’t made by buying into every gold rush. They were made by understanding where the money flows after the headlines disappear. Every dollar burned by one company becomes revenue for someone else. The real game isn’t predicting which AI startup wins. It’s figuring out who gets paid no matter which one loses. Markets rarely collapse because people don’t see the bubble. They collapse because everyone sees it… …and believes they can leave one minute before everyone else. That’s why the smartest investors can call something a bubble while continuing to buy it. They’re not betting they’re right. They’re betting they’ll be faster. Whenever you see billions flowing into a new industry, stop looking at the companies making the headlines. Look at the businesses selling the infrastructure. The chips. The electricity. The bandwidth. The data. History has a habit of rewarding the people selling picks and shovels long after everyone forgets who found the gold.

@Durektor97 give it a few updates and it'll be a different product

@pulmencr I wonder if he'll manage to improve this

@hitu_monke at least it's being honest🤣

@pulmencr the ai admitting it doesn't know how to drive either is peak

@hitu_monke makes sense, agents need an ending, not just a task

THE THING THAT DECIDES WHAT AI CAN AUTOMATE ISN'T INTELLIGENCE, IT'S WHETHER YOU CAN MEASURE IT every agent loop is four parts. a trigger, the work, a goal you verify, an output you log. people obsess over the first two and skip the third phase three is the whole thing. not because verification makes the loop good, but because it decides whether the loop can exist at all look at the two examples. "make this python script faster" loops fine, you benchmark it, you get a number, the system knows when it's done. "is this linkedin post good" never loops, because nothing tells it to stop. same model, same tools, same effort so the line isn't drawn where models get dumb. it's drawn where the finish line stops being a number. loops eat the measurable half of your work and leave the rest untouched, no matter how smart the model gets which means the last thing to get automated is taste. the exact thing everyone promised would go first

@s4yonnara interesting take, the loop is starting to matter more than the model itself

Anthropic's Applied AI team said something that reframes the whole "just get a bigger context window" debate. Prompt engineering is over. What's left is context engineering: deciding exactly what the model sees at every step of an agent loop. The proof is in the benchmark they shared. A year and a half ago Claude scored 49% on SWE-bench. Opus 4.5 scores 80% now. That jump didn't come from a better prompt. It came from engineering the loop itself: the tools, the memory, what gets summarized, what gets dropped. Their fix for long tasks: compact before the window fills, write state to notes outside the context, hand messy sub-tasks to isolated sub-agents. ~40 mins, free. Watch it, then read why your agent forgets everything after message 50 in the article below ↓

@test0rosso exactly, most people only ever see the front end

@pulmencr most people have no idea rooms like this exist, which is exactly why they matter

This room has over 200 phones running at the same time - all doing the exact same thing, non-stop A guy built the entire wall in just two days Every phone is mounted on a pegboard, wired into one power system, and controlled from two monitors a few feet away You've probably scrolled past content that came from a setup like this without ever knowing it This is what a phone farm actually looks like Businesses use these walls to run hundreds of accounts simultaneously, test apps at scale, or automate engagement that would take one person years to do manually The guy who built it was honest about it "When I tell you this was hard, it was hard" He's already planning to build a second wall next week Right now, a room full of phones is working And most people have no idea rooms like this even exist

@hitu_monke yeah, it's oddly intimidating

@pulmencr 200 phones on a pegboard is genuinely unsettling to look at