Square
Following
Hot
News
Profile

Yukiyu

vip
Active for: 1.7y
Peak Tier 0
No content yet
0
Following
2
Followers
17
Liked
My favorite coffees are from Thailand’s First Officer Coffee Roasters and Graph, and whenever I go to Chiang Mai, I always make sure to leave room in my suitcase for two bags of coffee beans
The coffee I usually drink☕️ generally comes from Colombia, Costa Rica, or Ethiopia. These regions are very established. You can’t say their coffee is bad, but the flavor is simply the kind you can expect.
Many of the beans sold in Chiang Mai come from Chiang Rai. You can sense their aroma before even taking a sip: fresh, delicate, with a hint of mountain fruitiness. They have layers without being overpowe
I chatted yesterday with a friend who made a tidy sum last year. For the past six months, he has been either skiing or surfing, barely looking at the markets or AI. He said that now he’s back, it feels like he didn’t miss anything.
As for me, when GPT 6 came out, my first reaction was that the little lobster I built with GPT 5 was obsolete, and that the old model workflow would have to be torn down and rebuilt from scratch. Was everything I had done before meaningless?
Let’s first look at what happened over the past few days.
On September 7, Jensen Huang posted that Astra was trained on 100,00
post-image
  • 2
Here I am, here I am💛
Come find me to check in✅
Wall Street is preparing to securitize computing power, turning it into an asset that can be traded and pledged as collateral like gold and oil.
In mid-August, Jensen Huang brought together six giants—BlackRock, Blackstone, Apollo, KKR, Brookfield, and Goldman Sachs—to announce the mobilization of $500 billion to financialize computing power.
His exact words were, "In AI, compute is revenue"—in AI, computing power is revenue.
However, as digital energy, Bitcoin has long been algorithmic pricing for computing power. During this four-day sustained rebound, hashprice jumped from $31.8 on August 1
BLK+1.13%
KKR-1.55%
GS-1.28%
BTC+0.42%
NVDA-1.42%
This time I came to Japan and found that robots really are starting to replace humans. In the mall, there is a robot security guard that’s 1.8 meters tall—an SECOM ugo company product in Tokyo—with a panoramic camera hidden in its hat. It can salute, it can press the elevator buttons by itself, and it patrols across floors.
Renting two robots costs the same as hiring one guard: no need to pay social insurance, and once charged, they stand guard for 8 hours.
At the airport, there are SECOM’s self-regulating patrol robots, cocobo, which can legally conduct night patrols on the road. Monthly rent
After the ChatGPT upgrade, talking to it basically has no delay. The latest real-time voice model’s latency is roughly 200–300 ms, and the natural response interval in human conversation is also 200–300 ms. It’s the same.
But the cost of near-zero latency is forgetting. The real-time voice model’s context window is only 32k tokens, which is 1/300 of a pure text model. Low latency = small memory + proactive forgetting. To both respond instantly and remember every thing you said—right now, it’s a dilemma.
Compute has been insufficient. Getting one SDXL image from a single 5090 takes 2 seconds; F
post-image
DRAM-5.87%
@ZF_lab Are we basically the same?
After turning 30, it feels like you’ve already lived through two-thirds of life, because day after day keeps repeating and your brain is too lazy to think anymore.
The brain encodes new information more densely, and it estimates how long time feels based on the amount of information stored in memory. During the fresh period, you record more details—so looking back makes it feel like, “This took forever.” Repetitive days leave almost no new memories, and they get compressed into a blur.
So I want to go out as much as possible, talk to more interesting people, and use fresh inputs to stretch tho
Why must robots be made humanoid? Because it eliminates the need to spend centuries transforming our current living environment. Huang Renxun says we built this world for ourselves, so he advocates for bipedal humanoids that can plug into existing infrastructure at the lowest cost.
The difficulty of building humanoid robots, according to Musk, falls somewhere between the Model X and Starship. There is no existing humanoid robot supply chain today; every component and actuator must be designed and developed from scratch.
At an annual production volume of one million units, the cost could be dri
NVDA-1.42%
GS-1.28%
New York is a city full of intense emotions.
I saw a few people sitting on the car roof, excited beyond words. I asked: Why not sit inside the car? My friend said, because it's not their car. She also said, "If the Knicks win, I’ll go climb a traffic light! 🚥"
And it turned out the Knicks really won!
The person telling me this is a friend who has a green card in New York, lives in New Jersey, commutes across the river to Manhattan every day, and works at Bloomberg headquarters.
Recently, she’s preparing to change jobs. I asked what kind of job she’s looking for? She thought for a mome
CBRE-1.77%
RWA-0.07%
USDC-0.02%
BLK+0.86%
GLDX-1.42%
Today in New York, I met with a friend who works in neurology, currently developing brain-machine interfaces.
She surprisingly said that in the academic community, Musk's Neuralink is considered outdated technology.
Biologists look down on its use of press releases to present scientific claims; Neuralink is seen as "engineering-led, science-cutting corners."
After a three-hour drive from DC, I thought it was just a catch-up, but we ended up talking all afternoon about brain-machine interfaces, AI, and storage anxieties.
They can design their own chips after raising funds, and order 22-
IBM-0.85%
NVDA-1.42%
BLK+1.13%
BTC+0.42%