与「COOL」相关的搜索结果

Yes, the green flame is caused by burning copper from depleting fuel in the regenerative cooling circuit and going oxygen-rich (closer to stoichiometric temperature) in the main combustion chamber. Slight underperformance of engines on this flight is the reason. The booster is meant to land with tiny residuals, so the solution is to improve engine performance and performance consistency (underway with Raptor V3.2), as well as adjust flight software to do MECO1 about one second earlier to ensure enough propellant for boostback and landing (tiny change). These early ships and boosters are already outdated by improved designs in production or on the test stand, so this booster loss does not change our timeline to reach high volume, fully reusable Starship flights.

When BTC takes a deep dive, I don't want my entire market perspective to turn red with it. 😂 I've just updated my GT Watchlist on @GeckoTerminal with four memes: PEPE, BONK, Nihao and $FOOD. I'm keeping an eye on their communities, pool liquidity and trading volume. When the hype cools down, I want to know which ones still have real market activity. My routine is simple: open GT on my phone, check my saved pools, compare price action and volume, and see what's changed. The best part? Creating a Watchlist is incredibly easy. Just find the tokens or pools you want to follow, add them to your list, and they're all organized in one place. No more searching for the same coins every time I open the app. A falling market isn't a reason to stop paying attention. This isn't the time to surrender and walk away. It's time to stay alert and spot opportunities. 🐸 Do you have any suggestions for me to add to this list?

I mean how cool is this! Let's prove memory safety for all C. Sounds crazy to say out loud, and tbh this is just a start... but we have to start somewhere. Explore it @

I have a major result to share. I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China. This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model. Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves. I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems. I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used. It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20. These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs. I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.

SpaceX has just unveiled a massive new $50 billion Supercomputing Campus that they want to build in Cochise County, Arizona. All info: • Up to 6 GW across the 9 supercomputing buildings • Construction starts 2027 • Size of each supercomputing building: ~493,000 square feet. • Total supercomputing building space: 4.44 million square feet, excluding support facilities. • SSVEC residential customers are projected to see electricity bill reductions of 10–40%. • Up to 24 GWh of grid-connected battery storage to provide backup power, stabilize the regional electric grid, support local voltage, and absorb fluctuations in electricity demand. Among the largest planned battery installations in the United States. • Additional solar generation in the region is planned to supply renewable electricity during the day and reduce demand on the local grid. • No dedicated on-site power plant or fuel storage yard is planned. • Each supercomputing building will have 10 MW of natural-gas-powered standby generation for critical equipment, operating only during utility outages and periodic testing. • No gas turbines will be installed at the site. • SpaceX plans to fund $3.6 billion in grid infrastructure upgrades, transferred to AEPCO at no cost to its members. • Up to $260 million annually in additional benefits for cooperative members at full buildout, on top of residential bill reductions. • Up to $200 million annually for cooperative members statewide, in addition to SSVEC benefits. • The supercomputers will use dry cooling with a closed-loop water circulation system rather than evaporative cooling. • SpaceX projects zero ongoing industrial water consumption. • The cooling system will be filled initially, with water reused for up to 10 years. • SpaceX estimates annual water usage equivalent to approximately 20–30 Arizona residential homes. • Projected annual water use would be less than 2.5% of a single center-pivot irrigation system's annual consumption. • Permanent jobs: Approximately 1,000. • Construction jobs: Approximately 5,500. • Annual property tax revenue: Up to $103 million in estimated new annual revenue. • Regional economic benefits: SpaceX says its investment will support public education, infrastructure, essential local services, and economic development. • Local suppliers: SpaceX is encouraging suppliers to participate in construction and development of the campus. • Noise mitigation: Equipment will be positioned away from neighboring properties where possible, with setbacks and additional screening to reduce noise. • Nighttime lighting: Outdoor lighting will remain off unless work is underway, with active lights shielded and directed downward. • Dark-sky protection: The campus will comply with Cochise County's dark-sky requirements to limit light pollution. • Nighttime construction: SpaceX says it will seek waivers when nighttime construction lighting is necessary. • Wildlife protection: The campus is being designed with consideration for the region's sandhill crane wintering habitat. This campus's purpose is to build one of the world's most advanced AI supercomputing campuses to train SpaceX's superintelligence models and support scientific and engineering breakthroughs. Learn more:

A year ago I got fired from MrBeast. For no reason at all by the way. Restructuring they said, personally I just think their HR guy didn’t like me. Anywho, that really sucked and I was super depressed, it was my dream job. Last night my girlfriend looked at me after a wonderful date night. She said, “I’m so proud of how far you’ve come” I was so confused and asked why? She said “you were so hurt and depressed and I didn’t think you’d get better from it” She was right but also day one I said “I’ll figure it out” and I did (thanks Prozac) Since that awful day I’ve: - doubled my MrBeast salary - traveled to more countries than I ever have - directed numerous shoots solo - led edits and creative production on launches for multi million dollar founders - met the coolest people I get to call family at @shownmedia All this rambling is to say, even if you think you’ve found your “dream job” something better may be out there and it’s not as big and flashy. I feel so low anxiety it’s amazing. We’re hiring btw;)

I just pushed a BIG update to my 3D kart racer that I've been building with @claudeai Fable all morning. It now has: → 3 amazing tracks: beach, jungle, volcano → Online multiplayer for up to 6 friends → 6 distinct racers and cool music I wanted to make this game actually fun instead of some one shot slop and I think I did it! 📌 Play it for free in your browser now:

Fast launch times are cool, but actual usage is what matters. Check out the speed of the long-requested mobile account switcher on iOS on the new Rust client. And yes it is just as fast on Android and Desktop. We're still in testing for now but hope to get it out to you as soon as it's ready.