๐ˆ๐ฏ๐จ (@FLASHevropa) header
๐ˆ๐ฏ๐จ (@FLASHevropa) profile picture

Cutting through the AI noise. Daily clips on where tech is headed. ๐Ÿค–

Joined August 2026
Photo by ๐ˆ๐ฏ๐จ (@FLASHevropa)Photo by ๐ˆ๐ฏ๐จ (@FLASHevropa)Photo by ๐ˆ๐ฏ๐จ (@FLASHevropa)Photo by ๐ˆ๐ฏ๐จ (@FLASHevropa)Photo by ๐ˆ๐ฏ๐จ (@FLASHevropa)Photo by ๐ˆ๐ฏ๐จ (@FLASHevropa)Photo by ๐ˆ๐ฏ๐จ (@FLASHevropa)Photo by ๐ˆ๐ฏ๐จ (@FLASHevropa)Photo by ๐ˆ๐ฏ๐จ (@FLASHevropa)Photo by ๐ˆ๐ฏ๐จ (@FLASHevropa)
A FOUR YEAR OLD CHIP STARTUP HAS $1 BILLION IN ORDERS. INVESTORS ARE OFFERING TO VALUE IT AT $40 BILLION. That is 40 times its order book. Look at the staircase. Etched in December 2025: $5 billion. July: $10.3 billion. August 18: $21 billion, one month after the last round. This week: bids at $40 to $50 billion. Four steps. Ten months. Eight to ten times. Now look at who sits on both sides. Jane Street tested the chip, bought it, installed a rack in its own data center, then led the $700 million round that doubled the price. Same name as customer and as lead investor. One honest caveat. These are offers, not a closed round. The $40 billion comes from top-tier firms. The $50 billion comes from lesser-known backers. Etched has accepted neither. Now the other side of the same week. An Nvidia-backed AI infrastructure company pulled a multibillion dollar IPO after public investors balked at the price. Private money is paying 40 times backlog. Public money just said no. Both are looking at the same industry. Only one of them has to explain the number to shareholders every quarter.
34
Steve Wozniak literally explained why he walked away from a fortune most people can't even picture. It is not that he lost the money. It is not that someone cheated him out of it. It is that he gave most of it away on purpose, while it was still entirely his to give. Wozniak and Steve Jobs each held 45 percent of Apple at the start, after five separate rejections from HP turned him toward building his own machine instead. Ron Wayne held the other 10 percent, and sold it back within days for 800 dollars, a stake that could be worth hundreds of billions today. By the 1980s, Wozniak started selling off his own stake too. Some went to early employees who had missed out on equity entirely when the company first took off. Some went straight to charity, no fanfare attached to any of it. He still technically works at Apple today. Officially the only person who has received a paycheck there every single week since the company began in 1976. After taxes, it comes to roughly 50 dollars. Had he held everything from the 1980 IPO without selling a share, he would likely be a trillionaire by now. Instead he says the quiet part out loud, decades after he had every reason to stop meaning it. I didn't want to be near money, because it could corrupt your values.
23
Markets never sleep, and neither does the signal in the noise. Every candle on the chart tells a story โ€” fear, greed, conviction, doubt. The best traders don't chase the green; they read the rhythm underneath it. Crypto taught us that volatility isn't the enemy โ€” it's the price of asymmetry. Stocks taught us that patience compounds louder than any single trade. Zoom out. The chart always tells the truth eventually.
42
THE MAN WHO SELLS THE CHIPS SAYS CEOs ARE LYING ABOUT WHY THEY FIRE PEOPLE Jensen Huang runs the most valuable chip company on earth. Every AI data center on the planet depends on what he sells. In a late May interview with CNA, he was asked about CEOs who blame AI for layoffs. He called that narrative "too lazy." Then he did the math out loud. AI only became genuinely useful about six months earlier. So how could layoffs from two years ago be its fault? His message to workers was blunter. You are not losing your job to AI. You are losing it to someone who uses AI better. Gartner found no link between AI-driven cuts and better returns. Nearly 6 in 10 companies admitted the AI framing was covering a financial reason. The one person with every incentive to sell you the fear is telling you not to buy it.
1
48
THE CEO OF GOOGLE WAS ASKED IF HIS COMPANY IS SAFE IF THE AI BOOM COLLAPSES Sundar Pichai sat down with the BBC at Google's California headquarters. He called the current wave of investment an extraordinary moment. Then he said the part most people skipped. Parts of it are irrational. He compared it to the internet. Clearly a lot of excess investment, and the internet was still profound. Both things at once. Then the BBC asked the direct question. Would Google be immune if it all burst? "No company is going to be immune, including us." At that moment Alphabet stock had doubled in seven months to $3.5 trillion. Ten months later, five companies spent $194 billion on AI infrastructure in a single quarter. Everyone quotes the boom. The man running one of its biggest engines used the word irrational.
1
37
FIVE COMPANIES SPENT $194 BILLION ON AI INFRASTRUCTURE. IN ONE QUARTER. Amazon: $54.21B Alphabet: $44.92B Microsoft: $35.80B Meta: $30.12B Oracle: $28.5B Combined, that's more than the GDP of most countries on earth, spent in three months, just on building capacity. Private data center construction has already outpaced every other kind of private construction in the country combined, through July alone. The whole buildout is projected to hit $10 trillion and 3.6% of US GDP. Nobody has fully answered who is actually paying for all of it.
2
31
128,536 TECH WORKERS LOST THEIR JOBS THIS YEAR. "AI" WAS THE REASON FOR MOST OF THE PRESS RELEASES. Gartner checked the actual numbers behind the wave. 80% of companies that ran autonomous AI systems did cut their workforce afterward. Zero correlation between those cuts and any improvement in ROI. Then a separate survey landed. Nearly 6 in 10 companies admitted, in the same breath, that they frame layoffs as AI-driven when the real reason is financial. Oracle alone let go of 21,000 people this year while spending billions building AI data centers. The technology is real. The efficiency story attached to a specific layoff often isn't.
2
71
IN 2021, A COMPANY PAID $2.4 MILLION FOR LAND THAT DOES NOT PHYSICALLY EXIST 116 plots in Decentraland, a metaverse nobody visits anymore. That was the top of the cycle. Plot #1 on Claim Avenue just listed at $978. Plot #5, the Central Command Tower, starts at $247. Same idea. Buy a permanent digital address, brand it, let people click through to your actual site. Different price by four orders of magnitude. One SUI-branded plot has already logged over 17,000 visits. Nobody is calling this the metaverse this time. That might be exactly why it's cheap enough to actually work.
2
93
New research tackles why AI "deep research" agents fall apart on long tasks. When an LLM agent has to search, connect, and write a long report step by step, two common approaches both break down: linear search-then-write loses context as evidence piles up (the classic "lost in the middle" problem), while outline-first planning can't see its own knowledge gaps clearly enough to know what to search for next. A new ICML 2026 paper (arXiv:2602.13830, Microsoft + collaborators) proposes DualGraph memory โ€” two graphs that evolve together: an Outline Graph (the report's structure) and a Knowledge Graph (facts and relations). By comparing the two, the system spots gaps and generates targeted search queries on its own. Result: it beats prior methods across three benchmarks (DeepResearch Bench, DeepResearchGym, DeepConsult) โ€” including a 53.08 RACE score on DeepResearch Bench using GPT-5. No "models crashing," just a solid architectural fix for a real weak spot in research agents.
105
Jim Simons never worked a day on Wall Street. He ran the most profitable trading operation in history anyway. His Medallion Fund averaged 66% annual returns for 30 years. No hedge fund, no bank, no algorithm since has come close. Here's the detail that breaks people's brains. Simons refused to hire finance people. No MBAs. No Wall Street veterans. Not one. He hired physicists, codebreakers, and linguists instead. People who had never placed a trade in their life. His reasoning: finance people already believe they know how markets work. That belief is the thing that loses money. Scientists don't trust a winning pattern until the data forces them to. That instinct built a fund that made more money than Wall Street's biggest banks combined, some years. While traders were reading charts and trusting gut instinct, a room of mathematicians was quietly outperforming all of them without ever predicting a single headline. The market didn't reward experience. It rewarded people willing to admit they knew nothing and let the data decide. Most people still haven't figured that out.
25
๐ˆ๐ฏ๐จ retweeted
Replying to @FLASHevropa
Interesting @GoldmanSachs ๐Ÿคฏ
2
1
2
45
AI is not helping traders anymore. It is replacing them. In 2000, Goldman Sachs's New York trading floor employed over 600 traders. Today there are just two. The rest of the work is handled by algorithms, overseen by 200 engineers. This is not a forecast. It already happened. In currency trading, Goldman Sachs found a simple ratio: one engineer replaces four traders. A human cannot process news in a second. An algorithm places the order before you finish reading the headline. The question is no longer whether AI will replace traders. It is how much time is left for the ones still working. The labor market in finance does not shift gradually. It breaks in jolts, and each jolt empties out an entire floor. Those who can work with algorithms stay in the game. The rest become statistics.
1
1
98
THE COMPUTE BILL NOBODY IS TRACKING โ€” FIVE DEALS, FIVE MONTHS April: Amazon, +5 gigawatts of capacity added to an existing deal. May: SpaceX, $1.25B every single month, through 2029. July: AMD, $5B. Late August: Nscale, $45B, powered by Nvidia's next-gen Vera Rubin chips, live in 2027. Early September: Volta, $10B, six years of Norwegian data center power. Five commitments. One lab. The same five months its valuation went from $380B to $965B. Everyone quotes the valuation number. Almost nobody adds up the monthly bill sitting underneath it.
1
205
TWO LAUNCHES, ONE AFTERNOON โ€” AND TWO BETS ABOUT WHAT A MODEL IS ACTUALLY FOR At 10am Anthropic cut Opus straight to $4 and $20 per million tokens. 90 minutes later OpenAI answered with Sol at $2 and $10. Half the price. Same afternoon. One bet says the ceiling sells. Opus scores 51.2, $1.34 a task, 52.5% on Terminal-Bench. The other says the floor sells. Sol scores 47.5 at max effort, $1.06 a task, matches the ceiling on business workflows at 40% of the cost. On coding specifically the floor drops to 43.9% against the ceiling's 52.5%. Nine points of index between them, for less than half the bill. Two labs, one day, zero agreement on what you're actually paying for. Which bet are you already paying for?
2
82
TWO LAUNCHES, ONE AFTERNOON โ€” AND TWO BETS ABOUT WHAT A MODEL IS ACTUALLY FOR At 10am Anthropic cut Opus straight to $4 and $20 per million tokens. 90 minutes later OpenAI answered with Sol at $2 and $10. Half the price. Same afternoon. One bet says the ceiling sells. Opus scores 51.2, $1.34 a task, 52.5% on Terminal-Bench. The other says the floor sells. Sol scores 47.5 at max effort, $1.06 a task, matches the ceiling on business workflows at 40% of the cost. On coding specifically the floor drops to 43.9% against the ceiling's 52.5%. Nine points of index between them, for less than half the bill. Two labs, one day, zero agreement on what you're actually paying for. Which bet are you already paying for?
3
53
A $1,999 Mac Mini Turned Into an AI Studio That Bills 6 Clients $9,400 Every Month For 7 months I rented other people's computers. $180 a month on GPU hours. $60 on an image API. $220 on a video credit plan that expired whether I burned it or not. In February the card declined at 2am mid render. I bought a Mac Mini instead. Ran everything local. Storage, models, the works. No more metered hours ticking against me while I slept. Client one paid $800 a month for local video edits nobody else could touch without uploading their raw footage somewhere. Client two paid double that for the same reason, privacy, not speed. Six clients now. $9,400 a month. The machine paid for itself inside the first eleven days. Everyone is renting the same rack in the cloud right now. I bought the box.
1
3
149
There Are Only 3 Ways to Make Money From AI. Everything Else Is Noise. You are chasing one of them right now. Probably the loudest one. And nobody has told you the other two exist because the people using them are too busy building to write threads about it. Build the tool. A tiny fraction of people. Massive payoff, massive risk. This is the one everyone romanticizes and almost nobody actually does. Sell the picks and shovels. Infrastructure, data, compute, training. Less glamorous. Far more predictable. This is where the quiet fortunes are getting made right now, away from the timeline. Use it to do old work faster. The most accessible path of the three. Not sexy. Rarely gets a screenshot. Compounds silently while everyone else argues about AGI timelines. Everything else, the prompt threads, the courses, the AI side hustle screenshots, is entertainment dressed up as a strategy. You are probably doing the third one badly instead of doing it on purpose.
1
2
127
๐ˆ๐ฏ๐จ retweeted
this guy showed how he makes $9,800 a month filming jungle survival builds that never happened. the non-obvious part: there is no jungle. there is no cave. there is no builder. he generates the entire "primitive building" video as if someone discovered a mangrove cave, cleared it out, framed the walls, laid insulation, and built a hidden shelter. 64 seconds of continuous construction. muddy water, roots, brick walls, pink fiberglass. footage that feels pulled from a survival YouTube channel with 10 million subscribers. except nobody got their hands dirty. - Claude Fable 5.1 writes the full build sequence: site discovery, clearing, framing, insulation, brick laying, reveal - Seedream generates 42 photorealistic reference frames of each construction phase - Seedance 2.5 animates the continuous build with correct tool usage, material layering, and water displacement - Cartesia places every sound: mud squelch, hammer strike, water drip, and the echo shift as the cave goes from open to enclosed - CapCut stitches the phases into one seamless 64-second cut everyone builds the shelter. he builds the video of building the shelter. and both are fake. $9,800 a month, and the cave was never real. the full method is in the article.
1
3
33
3,901
๐ˆ๐ฏ๐จ retweeted
Leverage 90:10 Payout ratio with your Sim-Funded Account and keep 90% of your simulated profits. Start your Evaluation today!

179
105
14
1,146
9,054,365
๐ˆ๐ฏ๐จ retweeted
David Jerison wrote one formula on the board and said this is the most important thing in the entire course. Then he solved in 10 seconds a problem that took a full lecture the day before. One subtraction. One line. Done. He called it the Fundamental Theorem of Calculus. If you know the pattern behind a changing quantity you do not need to measure every piece and add them up. You take the end value, subtract the start, and the answer falls out. He gave the class a thought experiment. You check your speedometer every second for an hour. 3600 readings. Multiply each by one second and add them up. You get how far you traveled. But the theorem says you do not need 3600 calculations. Read your odometer at the start and at the end. One subtraction. Same answer. Then he said he lied to them the day before. The integral is not the area under the curve. It is the area above the axis minus the area below. The net, not the total. What actually remains after the positives and negatives cancel. Your savings rate is the same subtraction. Bank balance December 31st minus January 1st divided by your income. Not what you planned to save. What actually remained. One number that tells you everything. The lecture is free. The people still adding up every small expense have never pressed play.
4
1
20
771