正在加载视频...

视频加载失败

返回首页

A 2-year study just proved Gigacasting repairs can cost LESS than traditional bodies 💡 Thatcham Research conducted a rigorous study on Tesla Model Y repairability. At 15 km/h impacts: Crash cans absorbed energy. Gigacasting sustained zero damage. Standard repair. At 25 km/h impacts: Casting cracked. Full replacement required. The...

lucagrecoita's profile picture

Luca Greco

9,293 subscribers

94,749 次观看 • 7 个月前 •via X (Twitter)

科学技术
LIVE

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

Dongfeng's 16,000-ton Gigacasting machine just entered production - the largest die casting machine ever built 🏭 It is now producing battery trays. For scale: before 2018, a "large" die casting machine was 4,000 tons. Tesla raised the bar to 6,000 tons for the Model Y in 2020. Then 9,000 tons for the Cybertruck. Now Dongfeng is producing at 16,000 tons. Four times the pre-Gigacasting standard. At this tonnage, Dongfeng can cast integrated chassis and battery tray structures in a single shot - components that require multiple machines and assembly steps at any other OEM. The evolution in 7 years: -> Pre-2018: 4,000T (industry ceiling) -> 2020: 6,000T (Tesla Model Y) -> 2023: 9,000T (Tesla Cybertruck) -> 2026: 16,000T (Dongfeng, in production) No Western machine is operating above 9,000 tons today. March 25-28th I will be in Wuhan, China where I will see this 16,000 T machine If you want to follow my trip, foIIow me on Instagram, I’ll post stories of the factory and event
0:38

Dongfeng's 16,000-ton Gigacasting machine just entered production - the largest die casting machine ever built 🏭 It is now producing battery trays. For scale: before 2018, a "large" die casting machine was 4,000 tons. Tesla raised the bar to 6,000 tons for the Model Y in 2020. Then 9,000 tons for the Cybertruck. Now Dongfeng is producing at 16,000 tons. Four times the pre-Gigacasting standard. At this tonnage, Dongfeng can cast integrated chassis and battery tray structures in a single shot - components that require multiple machines and assembly steps at any other OEM. The evolution in 7 years: -> Pre-2018: 4,000T (industry ceiling) -> 2020: 6,000T (Tesla Model Y) -> 2023: 9,000T (Tesla Cybertruck) -> 2026: 16,000T (Dongfeng, in production) No Western machine is operating above 9,000 tons today. March 25-28th I will be in Wuhan, China where I will see this 16,000 T machine If you want to follow my trip, foIIow me on Instagram, I’ll post stories of the factory and event

Luca Greco

111,133 次观看 • 5 个月前

Ford’s Unicastings are 27% lighter than Tesla’s 2022 Model Y 💡 Weight is one of the greatest efficiency robberies in EVs. Ford’s engineering team has a clear philosophy: “Best part is no part.” The Universal EV Platform uses front and rear Unicastings that deliberately prioritize weight over maximum part consolidation. Wide flat surfaces are excluded from the rear casting for this reason. The result is a 75% reduction in parts compared with a conventional pickup truck body, with the two castings eliminating 146 individual components. Welds drop by 66% and fasteners by 25%. Ford’s own data show the castings are 27% lighter than those on Tesla’s 2022 Model Y, with the front piece estimated under 65 kg and the rear under 45 kg. This design doesn't achieve the biggest possible consolidation. It is the lightest structure that still delivers major part and process reduction. The platform also changes how the vehicle is built. A three-branch assembly line replaces the traditional single conveyor — one branch for the front, one for the rear, and one for the structural battery, seats, consoles and carpets. The modules are then integrated, cutting workstations by 40% and raising production speed by 15%. This architecture underpins Ford’s $2 billion transformation of the Louisville plant. The first vehicle, a mid-size electric pickup, is planned for 2027, followed by crossovers and ride-hailing models on the same platform. When weight reduction and part reduction move together, the economics of affordable EVs start to look different. 📊 Ford program tracked in The Gigacasting Database and forecast report: ❌ Don't leave your insights to chance with the X algorithm ✅ Subscribe for free to my weekly newsletter about all things Gigacasting and magnesium Thixomolding: 📬
1:28

Ford’s Unicastings are 27% lighter than Tesla’s 2022 Model Y 💡 Weight is one of the greatest efficiency robberies in EVs. Ford’s engineering team has a clear philosophy: “Best part is no part.” The Universal EV Platform uses front and rear Unicastings that deliberately prioritize weight over maximum part consolidation. Wide flat surfaces are excluded from the rear casting for this reason. The result is a 75% reduction in parts compared with a conventional pickup truck body, with the two castings eliminating 146 individual components. Welds drop by 66% and fasteners by 25%. Ford’s own data show the castings are 27% lighter than those on Tesla’s 2022 Model Y, with the front piece estimated under 65 kg and the rear under 45 kg. This design doesn't achieve the biggest possible consolidation. It is the lightest structure that still delivers major part and process reduction. The platform also changes how the vehicle is built. A three-branch assembly line replaces the traditional single conveyor — one branch for the front, one for the rear, and one for the structural battery, seats, consoles and carpets. The modules are then integrated, cutting workstations by 40% and raising production speed by 15%. This architecture underpins Ford’s $2 billion transformation of the Louisville plant. The first vehicle, a mid-size electric pickup, is planned for 2027, followed by crossovers and ride-hailing models on the same platform. When weight reduction and part reduction move together, the economics of affordable EVs start to look different. 📊 Ford program tracked in The Gigacasting Database and forecast report: ❌ Don't leave your insights to chance with the X algorithm ✅ Subscribe for free to my weekly newsletter about all things Gigacasting and magnesium Thixomolding: 📬

Luca Greco

39,955 次观看 • 1 个月前

Here’s my written & video review of the new 2026 Tesla Model Y Performance after driving it for a week. This is the best-value new Tesla you can buy. Crazy performance, no real drawbacks, and all for just $57,490. Let’s dive in. Price: I haven’t seen many others mention this: every option on the Model Y Performance is included at no extra cost in the US (except FSD). So if you spec a Model Y Premium AWD with an upgraded paint color, tow package, white interior, and upgraded 20" wheels, a fully loaded Model Y Premium AWD ends up only about $2,500 less expensive than a fully loaded Model Y Performance. Ride Quality: I thought I might feel worse ride quality vs my Premium AWD Model Y with 20" wheels, but I struggled to find any real difference, despite the larger 21" wheels, firmer suspension setting, and 0.6" lower ride height on the Performance trim. A true testament to Tesla's engineering magic on this thing. Exterior Design: Unlike the previous Model Y Performance, Tesla made some exterior design tweaks with a new front and rear fascia to spice things up a bit. The result, in my opinion, is the best-looking Model Y trim you can buy. It definitely has a more aggressive presence in person, even if it's subtle. The carbon-fiber spoiler boosts high-speed stability and cuts aerodynamic drag by 10%. The new 21’’ Arachnid 2.0 wheels look fantastic in person, one of my favorite designs ever from Tesla. Staggered wheel and tire fitment provides better grip and steering. The beefier 275mm rear tires (255mm in the front) give the vehicle a better stance from behind. Vehicle-to-Load (V2L): For the first time on a Model Y in North America, you can now plug in anything you want to the exterior charge port with an adapter, even a campsite! It provides up to 2.4 kW of power (120V at 20A) from two household outlets. It's a great feature. Interior: It’s what you know and love, but with a few changes that elevate the ownership experience. New with the Performance is a larger 16" center screen (vs. 15.4" on non-Performance models), with thinner bezels and higher resolution. It’s not a huge difference on paper, but you definitely notice it in daily use. The carbon-fiber décor on the door cards and dash is a nice touch, though I would like to see it extended to the center console. The new performance seats are the best seats of any Tesla I've ever experienced. While they retain aggressive bolstering in the torso area, the bottom seat cushion has less aggressive bolstering than on the Model 3 Performance seats, making it easier to get in and out. It also doesn’t squeeze your thighs too tightly. The powered thigh extenders add comfort on longer drives, especially for taller people who want extra support. And of course, they’re heated and ventilated. The headrests also feel more comfortable than the ones in my Model Y. I want these seats. Unlike the old Model Y Performance, the new one has no Track Mode. Why? Because nobody used it lol. No point in putting engineering resources into something people won’t use. The refreshed Model 3 Performance still has it, though. Cabin Quietness: Despite the thinner-profile tires, there is no noticeable difference vs my Premium Model Y. Decibel reading results at highway speeds were visually the same compared to my 2026 Model Y Premium (65-66). Driving Impressions: It’s amazing. Sharp, precise, and agile. The vehicle feels stable at all times. Acceleration is blistering (3.3s 0–60 mph), with plenty of punch even at higher speeds. The tires offer good grip, and cornering is fantastic for an SUV. The suspension setup is great. More steering wheel feedback would be nice, though. Cruising around traffic is a joy. The brakes are much improved over the previous Model Y Performance and are far better suited for spirited driving. There are three acceleration modes: Chill, Standard, and Insane. Just stay in Insane. You’d be insane not to lol. The car also lets you switch between two ride and handling modes: Standard and Sport. The difference isn’t huge, but Standard is better if you’ve got passengers. FSD: I unfortunately wasn’t able to get FSD V14 on this car. It had V13.2.9, so I didn’t use it much. But in the little time I did, it was smooth and comfortable. It didn’t bother me because V14 will perform just as well here as it does on my 2026 Model Y Premium AWD. Conclusion: You won’t find another new SUV today that offers this level of performance for the price. Back in 2022, when Tesla couldn’t build Model Ys fast enough, a fully loaded Model Y Performance cost over $90,440. Today, the refreshed and far more capable 2026 Model Y Performance is just $57,490 fully loaded, and you can simply subscribe to FSD for $99/month. The 2026 Model Y Performance delivers utility, great performance, comfort, tech, self-driving and everything else people love about the Model Y. It’s a no-brainer purchase. I want one badly, but I'll need to show restraint, as I’m saving up for a house lol.
15:48

Here’s my written & video review of the new 2026 Tesla Model Y Performance after driving it for a week. This is the best-value new Tesla you can buy. Crazy performance, no real drawbacks, and all for just $57,490. Let’s dive in. Price: I haven’t seen many others mention this: every option on the Model Y Performance is included at no extra cost in the US (except FSD). So if you spec a Model Y Premium AWD with an upgraded paint color, tow package, white interior, and upgraded 20" wheels, a fully loaded Model Y Premium AWD ends up only about $2,500 less expensive than a fully loaded Model Y Performance. Ride Quality: I thought I might feel worse ride quality vs my Premium AWD Model Y with 20" wheels, but I struggled to find any real difference, despite the larger 21" wheels, firmer suspension setting, and 0.6" lower ride height on the Performance trim. A true testament to Tesla's engineering magic on this thing. Exterior Design: Unlike the previous Model Y Performance, Tesla made some exterior design tweaks with a new front and rear fascia to spice things up a bit. The result, in my opinion, is the best-looking Model Y trim you can buy. It definitely has a more aggressive presence in person, even if it's subtle. The carbon-fiber spoiler boosts high-speed stability and cuts aerodynamic drag by 10%. The new 21’’ Arachnid 2.0 wheels look fantastic in person, one of my favorite designs ever from Tesla. Staggered wheel and tire fitment provides better grip and steering. The beefier 275mm rear tires (255mm in the front) give the vehicle a better stance from behind. Vehicle-to-Load (V2L): For the first time on a Model Y in North America, you can now plug in anything you want to the exterior charge port with an adapter, even a campsite! It provides up to 2.4 kW of power (120V at 20A) from two household outlets. It's a great feature. Interior: It’s what you know and love, but with a few changes that elevate the ownership experience. New with the Performance is a larger 16" center screen (vs. 15.4" on non-Performance models), with thinner bezels and higher resolution. It’s not a huge difference on paper, but you definitely notice it in daily use. The carbon-fiber décor on the door cards and dash is a nice touch, though I would like to see it extended to the center console. The new performance seats are the best seats of any Tesla I've ever experienced. While they retain aggressive bolstering in the torso area, the bottom seat cushion has less aggressive bolstering than on the Model 3 Performance seats, making it easier to get in and out. It also doesn’t squeeze your thighs too tightly. The powered thigh extenders add comfort on longer drives, especially for taller people who want extra support. And of course, they’re heated and ventilated. The headrests also feel more comfortable than the ones in my Model Y. I want these seats. Unlike the old Model Y Performance, the new one has no Track Mode. Why? Because nobody used it lol. No point in putting engineering resources into something people won’t use. The refreshed Model 3 Performance still has it, though. Cabin Quietness: Despite the thinner-profile tires, there is no noticeable difference vs my Premium Model Y. Decibel reading results at highway speeds were visually the same compared to my 2026 Model Y Premium (65-66). Driving Impressions: It’s amazing. Sharp, precise, and agile. The vehicle feels stable at all times. Acceleration is blistering (3.3s 0–60 mph), with plenty of punch even at higher speeds. The tires offer good grip, and cornering is fantastic for an SUV. The suspension setup is great. More steering wheel feedback would be nice, though. Cruising around traffic is a joy. The brakes are much improved over the previous Model Y Performance and are far better suited for spirited driving. There are three acceleration modes: Chill, Standard, and Insane. Just stay in Insane. You’d be insane not to lol. The car also lets you switch between two ride and handling modes: Standard and Sport. The difference isn’t huge, but Standard is better if you’ve got passengers. FSD: I unfortunately wasn’t able to get FSD V14 on this car. It had V13.2.9, so I didn’t use it much. But in the little time I did, it was smooth and comfortable. It didn’t bother me because V14 will perform just as well here as it does on my 2026 Model Y Premium AWD. Conclusion: You won’t find another new SUV today that offers this level of performance for the price. Back in 2022, when Tesla couldn’t build Model Ys fast enough, a fully loaded Model Y Performance cost over $90,440. Today, the refreshed and far more capable 2026 Model Y Performance is just $57,490 fully loaded, and you can simply subscribe to FSD for $99/month. The 2026 Model Y Performance delivers utility, great performance, comfort, tech, self-driving and everything else people love about the Model Y. It’s a no-brainer purchase. I want one badly, but I'll need to show restraint, as I’m saving up for a house lol.

Sawyer Merritt

225,461 次观看 • 9 个月前

AI spend is now one of the fastest-growing line items at every company, but here's the problem: it's invisible to traditional spend management. We experienced this firsthand at Brex, so we built Magpie: an internal tool that sources and normalizes AI spend across every vendor so we can break it down by model, customer, employee, and workflow. What Magpie has already unlocked: → ~85% cost reduction on an audit agent through prompt caching → Real customer cost-to-serve numbers that inform pricing and margin → Model routing governance we couldn't enforce before → A surprise: operational AI is much cheaper than we assumed So why is AI spend so hard to track? Because each one exposes usage data differently (some not at all). At Brex, tool-use costs weren't tracked alongside tokens, customer attribution was inconsistent, and a meaningful chunk of our daily AI spend was unattributable. Now with Magpie, we have reliable visibility. AI spend is more complex than anything finance teams have had to manage before, and it's exactly the kind of problem Brex is built to solve, shaping how we think about the future of our platform. More below.
1:56

AI spend is now one of the fastest-growing line items at every company, but here's the problem: it's invisible to traditional spend management. We experienced this firsthand at Brex, so we built Magpie: an internal tool that sources and normalizes AI spend across every vendor so we can break it down by model, customer, employee, and workflow. What Magpie has already unlocked: → ~85% cost reduction on an audit agent through prompt caching → Real customer cost-to-serve numbers that inform pricing and margin → Model routing governance we couldn't enforce before → A surprise: operational AI is much cheaper than we assumed So why is AI spend so hard to track? Because each one exposes usage data differently (some not at all). At Brex, tool-use costs weren't tracked alongside tokens, customer attribution was inconsistent, and a meaningful chunk of our daily AI spend was unattributable. Now with Magpie, we have reliable visibility. AI spend is more complex than anything finance teams have had to manage before, and it's exactly the kind of problem Brex is built to solve, shaping how we think about the future of our platform. More below.

Pedro Franceschi

28,580 次观看 • 3 个月前

John Ternus, Apple's SVP of Hardware Engineering, explains why Apple deliberately made the iPhone harder to repair, and why the math says it was worth it: In a conversation with MKBHD, John frames the design challenge by asking you to imagine two extremes: "Sometimes for me I find it helpful to kind of think about the book ends. Like if you imagine a product that never fails, right? That just doesn't fail. And on the other end, a product that maybe isn't very reliable but is super easy to repair." His position is clear: "Product that never fails is obviously better for the customer. It's better for the environment." When pushed on whether infinite repairability and infinite durability have to be mutually exclusive, John acknowledges they aren't always, but explains why the tension is real, using the iPhone battery as an example. Batteries wear out. If you want to extend the life of the product, they need to be replaced. But in the early days of iPhone, one of the most common failures wasn't the battery, it was water: "Where you drop it in the pool or you, you know, spill your drink on it and the unit fails. And so, we've been making strides over all those years to get better and better and better in terms of minimizing those failures." That work led Apple to an IP68 rating, the point where customers fish their phones out of lakes after two weeks and find them still working. But there was a cost to achieving that level of durability: "To get the product there, you've got to design a lot of seals, adhesives, other things to make it perform that way, which makes it a little harder to do that battery repair." That's the deliberate tradeoff. Apple chose tighter seals and stronger adhesives, knowing it would make battery replacement more difficult, because the reliability gains were worth it. John argues the math backs this decision: "It's objectively better for the customer to have that reliability and it's ultimately better for the planet because the failure rates since we got to that point have just dropped. It's plummeted, right? The number of repairs that need to happen and every time you're doing a repair, you're bringing in new materials to replace whatever broke." His conclusion reframes the entire repairability debate: "You can actually do the math and figure out there's a threshold at which if I can make it this durable, then it's better to have it a little bit harder to repair because it's going to net out."
2:27

John Ternus, Apple's SVP of Hardware Engineering, explains why Apple deliberately made the iPhone harder to repair, and why the math says it was worth it: In a conversation with MKBHD, John frames the design challenge by asking you to imagine two extremes: "Sometimes for me I find it helpful to kind of think about the book ends. Like if you imagine a product that never fails, right? That just doesn't fail. And on the other end, a product that maybe isn't very reliable but is super easy to repair." His position is clear: "Product that never fails is obviously better for the customer. It's better for the environment." When pushed on whether infinite repairability and infinite durability have to be mutually exclusive, John acknowledges they aren't always, but explains why the tension is real, using the iPhone battery as an example. Batteries wear out. If you want to extend the life of the product, they need to be replaced. But in the early days of iPhone, one of the most common failures wasn't the battery, it was water: "Where you drop it in the pool or you, you know, spill your drink on it and the unit fails. And so, we've been making strides over all those years to get better and better and better in terms of minimizing those failures." That work led Apple to an IP68 rating, the point where customers fish their phones out of lakes after two weeks and find them still working. But there was a cost to achieving that level of durability: "To get the product there, you've got to design a lot of seals, adhesives, other things to make it perform that way, which makes it a little harder to do that battery repair." That's the deliberate tradeoff. Apple chose tighter seals and stronger adhesives, knowing it would make battery replacement more difficult, because the reliability gains were worth it. John argues the math backs this decision: "It's objectively better for the customer to have that reliability and it's ultimately better for the planet because the failure rates since we got to that point have just dropped. It's plummeted, right? The number of repairs that need to happen and every time you're doing a repair, you're bringing in new materials to replace whatever broke." His conclusion reframes the entire repairability debate: "You can actually do the math and figure out there's a threshold at which if I can make it this durable, then it's better to have it a little bit harder to repair because it's going to net out."

Big Brain Business

385,849 次观看 • 4 个月前

Sam Altman made the case for open-source harnesses in July. a month later, someone shipped it, and it's more efficient than most managed harnesses. here is the problem it was aimed at: a large share of your agent's token bill is the model rereading things it already read. that isn't the model's doing. the runtime around it decides what goes into every prompt and how often the model gets called. for example, an agent queries a CRM at step four and gets back 400 rows. those rows get piled up in the conversation history. by step nineteen, the model has to read those rows fifteen times unnecessarily, and every token read is billed at input rates. it happened because your harness assembled that prompt on every turn and kept the rows in it. that gives you two levers: how much context the harness carries forward, and how often it calls the model. there are four practical ways to keep the prompt from growing unnecessarily: → load tool schemas on demand. a server with 100 tools doesn't need to put all 100 into every prompt when the agent only calls two. → offload large results to disk. turn a large response into a short preview and a file path instead of replaying the entire result on every turn. → delegate to subagents. let a subagent spend thirty tool calls in its own context and return one summary to the root agent. → run toolchains in code. one script calls three tools, joins the results, and returns a table instead of three turns each dragging a full response. but reducing context is only half the job. you also need to control how often the model gets called. a good harness should avoid unnecessary planning, verification, and reflection when the work can be completed in fewer steps. TrueFoundry's open-source agent harness, TrueForge, is built around both of those controls. it sits between the model and the tools, deciding what goes into every prompt and when another model call is actually needed. it also breaks token usage down across the harness, skills, instructions, tools, and messages. DevRev's Enterprise-Bench is where this gets tested, on multi-step tasks of the kind where an agent pulls records from one system and reconciles them against another. TrueFoundry ran TrueForge there against Claude Managed Agents, both on the same model, and both finished the same number of tasks. the tie is the part that matters, because it means the gap underneath is not a quality tradeoff. TrueForge reached that score on close to a third of the tokens, with roughly 40% fewer trips back to the model. for the same result, that comes out around 2.7x cheaper than Claude Managed Agents. swapping in an open model made it sharper still. TrueForge with GLM-5.2 scored a little higher than either setup above, and the entire benchmark run cost about $3 at list prices. being open source matters beyond the license here. the model underneath can be swapped without rewriting the agent, and the whole thing can run inside your own environment when the data cannot leave it. all of this comes down to the runtime around the model, the context it carries, the tools it exposes, and how many times it goes back to the model. that is what a production harness actually owns. the full task list, the per-run numbers, and the MIT-licensed code are on GitHub: (don't forget to star 🌟) you can read more about the same in the article quoted below. thanks to the TrueForge team for working with me on this one.
0:39

Sam Altman made the case for open-source harnesses in July. a month later, someone shipped it, and it's more efficient than most managed harnesses. here is the problem it was aimed at: a large share of your agent's token bill is the model rereading things it already read. that isn't the model's doing. the runtime around it decides what goes into every prompt and how often the model gets called. for example, an agent queries a CRM at step four and gets back 400 rows. those rows get piled up in the conversation history. by step nineteen, the model has to read those rows fifteen times unnecessarily, and every token read is billed at input rates. it happened because your harness assembled that prompt on every turn and kept the rows in it. that gives you two levers: how much context the harness carries forward, and how often it calls the model. there are four practical ways to keep the prompt from growing unnecessarily: → load tool schemas on demand. a server with 100 tools doesn't need to put all 100 into every prompt when the agent only calls two. → offload large results to disk. turn a large response into a short preview and a file path instead of replaying the entire result on every turn. → delegate to subagents. let a subagent spend thirty tool calls in its own context and return one summary to the root agent. → run toolchains in code. one script calls three tools, joins the results, and returns a table instead of three turns each dragging a full response. but reducing context is only half the job. you also need to control how often the model gets called. a good harness should avoid unnecessary planning, verification, and reflection when the work can be completed in fewer steps. TrueFoundry's open-source agent harness, TrueForge, is built around both of those controls. it sits between the model and the tools, deciding what goes into every prompt and when another model call is actually needed. it also breaks token usage down across the harness, skills, instructions, tools, and messages. DevRev's Enterprise-Bench is where this gets tested, on multi-step tasks of the kind where an agent pulls records from one system and reconciles them against another. TrueFoundry ran TrueForge there against Claude Managed Agents, both on the same model, and both finished the same number of tasks. the tie is the part that matters, because it means the gap underneath is not a quality tradeoff. TrueForge reached that score on close to a third of the tokens, with roughly 40% fewer trips back to the model. for the same result, that comes out around 2.7x cheaper than Claude Managed Agents. swapping in an open model made it sharper still. TrueForge with GLM-5.2 scored a little higher than either setup above, and the entire benchmark run cost about $3 at list prices. being open source matters beyond the license here. the model underneath can be swapped without rewriting the agent, and the whole thing can run inside your own environment when the data cannot leave it. all of this comes down to the runtime around the model, the context it carries, the tools it exposes, and how many times it goes back to the model. that is what a production harness actually owns. the full task list, the per-run numbers, and the MIT-licensed code are on GitHub: (don't forget to star 🌟) you can read more about the same in the article quoted below. thanks to the TrueForge team for working with me on this one.

Akshay 🚀

76,552 次观看 • 20 天前

🇨🇳 Many people don't know this but Chinese high speed trains are now the fastest in the word and officially faster than even the 🇯🇵 Japanese Shinkansen Chinese trains go 350-400km/h (217-250mph) depending on the model While the Japanese Shinkansen goes 320km/h (200mph) max So I went to try it from Chengdu to Chongqing (just $80 for biz class) This one is the slightly older 350km/h model The business class cabin has 5 big seats that can lay flat with a stewardess dedicated to the cabin who serves you drinks like coffee, water, orange juice etc. I guess if it's a longer trip you probably get food served too The journey was very smooth and very fast, it maintained close to 350km/h at all times and there wasn't many stops or breaks (good) It didn't have WiFi, which is kind of a miss but it seems Chinese just use 5G for everything which worked fine for me without interrruption too I'd say the interior could be a bit more beautiful and modern. I usually say things in Japan look ancient but I think the Shinkansen interior there looks a bit more modern than here The trip took 1 hour and it passed about 300km in that time, roughly equivalent to the width of my country the Netherlands 😂 Anyway solid experience (and cheap!) 👍
0:33

🇨🇳 Many people don't know this but Chinese high speed trains are now the fastest in the word and officially faster than even the 🇯🇵 Japanese Shinkansen Chinese trains go 350-400km/h (217-250mph) depending on the model While the Japanese Shinkansen goes 320km/h (200mph) max So I went to try it from Chengdu to Chongqing (just $80 for biz class) This one is the slightly older 350km/h model The business class cabin has 5 big seats that can lay flat with a stewardess dedicated to the cabin who serves you drinks like coffee, water, orange juice etc. I guess if it's a longer trip you probably get food served too The journey was very smooth and very fast, it maintained close to 350km/h at all times and there wasn't many stops or breaks (good) It didn't have WiFi, which is kind of a miss but it seems Chinese just use 5G for everything which worked fine for me without interrruption too I'd say the interior could be a bit more beautiful and modern. I usually say things in Japan look ancient but I think the Shinkansen interior there looks a bit more modern than here The trip took 1 hour and it passed about 300km in that time, roughly equivalent to the width of my country the Netherlands 😂 Anyway solid experience (and cheap!) 👍

@levelsio

2,013,282 次观看 • 9 个月前

Last week at Annecy - the 'Cannes of Animation' - I watched the future of entertainment unfold. While most Hollywood executives networked over formal dinners, the CEOs of Mediawan and Gameloft were doing something different. They were partying with Claynosaurz holders, talking about how this entertainment model can revolutionize IP creation and incubation. The numbers had caught their attention, but the sophistication of the model and community kept them there. As someone who works in media & entertainment, investing in production studios and co-financing movies, I've never seen anything like this. Here's why it matters. The traditional entertainment industry is broken. I spoke with six independent animation studio heads at Annecy - all successful studios who've produced for major platforms. They're fed up. Netflix and Disney are slashing budgets. Cost-plus deals have dropped from +25% to +5%. Creators have zero ownership. Even when you succeed, you're building someone else's empire. The reality? Social video and YouTube are eating traditional long-form content alive. Gen Z doesn't care about Hollywood's gatekeepers. They want authentic content and real connection with creators. The executives know this - they're desperate for a new model but can't let go of their old playbook. Then there's Claynosaurz. A few weeks ago, they generated $18M in demand from a single Popkins raffle - more than most animation studios see from a year of content. When I shared this with studio heads at Annecy, they couldn't believe it. But what happened next proved this wasn't just about numbers. The Claynosaurz party became THE event of Annecy. The CEOs of Mediawan and Gameloft weren't just making appearances - they were actively seeking out Clayno whales like CryptoKid, 𝙋𝘼𝙉𝙄𝘾, myself and others. They wanted to understand what we see in the brand that they're missing in traditional entertainment. These weren't courtesy meetings - these were strategy sessions that turned into 4AM conversations about the future of media. In our conversations, these CEOs weren't just excited about the Claynosaurz partnership - they were fascinated by how completely Claynosaurz has flipped the traditional Hollywood playbook. While studios spend millions trying to find audiences, Claynosaurz has built a passionate community that markets itself. While traditional IP takes years to validate, Claynosaurz is applying the lean startup model to IP incubation, focusing on social video, testing their content and building an audience. This is unprecedented in entertainment. Even the creator of Paw Patrol - who built a $10B+ global franchise the traditional way - came to understand what makes Claynosaurz different. The President of Animation Magazine was there too, recognizing this wasn't just another Web3 project - this was the future of franchise building. These weren't courtesy meetings. These were industry leaders realizing that while they've been fighting over streaming deals, Claynosaurz has cracked the code on something more valuable: genuine community engagement. The old model is dying. Streaming economics are broken. But while Hollywood wrestles with these existential threats, Claynosaurz is thriving. Their community-first model isn't just working - it's about to be supercharged. Why? Because Claynosaurz understood something fundamental: in the age of infinite content, superfans are everything. While Hollywood sees AI as an existential threat, Claynosaurz sees it as rocket fuel for community creativity. Imagine thousands of holders empowered to tell their own stories in the Claynosaurz universe, with the quality of studio animation but the authenticity of user-generated content. That's not just a different distribution model - that's a fundamental reimagining of how entertainment is created. My conviction has never been stronger: Claynosaurz will be a household name and multi-billion dollar franchise. The blueprint is clear. The partners are committed. The community is unmatched. And most importantly - the team is world class. They don't just understand entertainment - they understand how to build authentic connections in a digital age. When the entertainment industry looks back, they won't just remember Annecy 2023 as the moment Web3 crashed Hollywood's gates. They'll remember it as the moment the future of entertainment became clear. Traditional studios will either adapt to this new model or become irrelevant. Claynosaurz isn't just building another entertainment company. They're building Web3 Disney. And we're just getting started. 🌋🦖
1:22

Last week at Annecy - the 'Cannes of Animation' - I watched the future of entertainment unfold. While most Hollywood executives networked over formal dinners, the CEOs of Mediawan and Gameloft were doing something different. They were partying with Claynosaurz holders, talking about how this entertainment model can revolutionize IP creation and incubation. The numbers had caught their attention, but the sophistication of the model and community kept them there. As someone who works in media & entertainment, investing in production studios and co-financing movies, I've never seen anything like this. Here's why it matters. The traditional entertainment industry is broken. I spoke with six independent animation studio heads at Annecy - all successful studios who've produced for major platforms. They're fed up. Netflix and Disney are slashing budgets. Cost-plus deals have dropped from +25% to +5%. Creators have zero ownership. Even when you succeed, you're building someone else's empire. The reality? Social video and YouTube are eating traditional long-form content alive. Gen Z doesn't care about Hollywood's gatekeepers. They want authentic content and real connection with creators. The executives know this - they're desperate for a new model but can't let go of their old playbook. Then there's Claynosaurz. A few weeks ago, they generated $18M in demand from a single Popkins raffle - more than most animation studios see from a year of content. When I shared this with studio heads at Annecy, they couldn't believe it. But what happened next proved this wasn't just about numbers. The Claynosaurz party became THE event of Annecy. The CEOs of Mediawan and Gameloft weren't just making appearances - they were actively seeking out Clayno whales like CryptoKid, 𝙋𝘼𝙉𝙄𝘾, myself and others. They wanted to understand what we see in the brand that they're missing in traditional entertainment. These weren't courtesy meetings - these were strategy sessions that turned into 4AM conversations about the future of media. In our conversations, these CEOs weren't just excited about the Claynosaurz partnership - they were fascinated by how completely Claynosaurz has flipped the traditional Hollywood playbook. While studios spend millions trying to find audiences, Claynosaurz has built a passionate community that markets itself. While traditional IP takes years to validate, Claynosaurz is applying the lean startup model to IP incubation, focusing on social video, testing their content and building an audience. This is unprecedented in entertainment. Even the creator of Paw Patrol - who built a $10B+ global franchise the traditional way - came to understand what makes Claynosaurz different. The President of Animation Magazine was there too, recognizing this wasn't just another Web3 project - this was the future of franchise building. These weren't courtesy meetings. These were industry leaders realizing that while they've been fighting over streaming deals, Claynosaurz has cracked the code on something more valuable: genuine community engagement. The old model is dying. Streaming economics are broken. But while Hollywood wrestles with these existential threats, Claynosaurz is thriving. Their community-first model isn't just working - it's about to be supercharged. Why? Because Claynosaurz understood something fundamental: in the age of infinite content, superfans are everything. While Hollywood sees AI as an existential threat, Claynosaurz sees it as rocket fuel for community creativity. Imagine thousands of holders empowered to tell their own stories in the Claynosaurz universe, with the quality of studio animation but the authenticity of user-generated content. That's not just a different distribution model - that's a fundamental reimagining of how entertainment is created. My conviction has never been stronger: Claynosaurz will be a household name and multi-billion dollar franchise. The blueprint is clear. The partners are committed. The community is unmatched. And most importantly - the team is world class. They don't just understand entertainment - they understand how to build authentic connections in a digital age. When the entertainment industry looks back, they won't just remember Annecy 2023 as the moment Web3 crashed Hollywood's gates. They'll remember it as the moment the future of entertainment became clear. Traditional studios will either adapt to this new model or become irrelevant. Claynosaurz isn't just building another entertainment company. They're building Web3 Disney. And we're just getting started. 🌋🦖

m3taversal

37,309 次观看 • 1 年前

A Tesla is the #1 safest vehicle on Earth and I would never let my kids and the people I love drive anything other than a Tesla. Independent safety agencies around the world all come to the same conclusion. Teslas consistently earn the highest possible safety ratings and even set records no other car has beaten. Many still may not believe this, but safety is the #1 priority behind every single design decision at Tesla. The Tesla Model 3 and Y both hold 5-star overall ratings from U.S. regulators in every category: frontal crash, side crash, and rollover protection. The Model 3 also holds the lowest probability of serious occupant injury ever recorded in government testing - 5.7%, compared to 7-15% for most sedans and SUVs. No other vehicle has ever surpassed that. In Europe, the results are just as strong. The Model Y earned 98% adult occupant protection, 89% child protection, and one of the highest active safety scores ever measured. Also, in the real world, using data collected from billions of miles driven, Tesla’s own safety reports show: • 1 crash every ~6.36 million miles when Autopilot or supervised FSD is active • 1 crash every ~3.85 million miles with standard active safety features • All while the U.S. average is 1 crash every ~670,000 miles! Bro… this means driving with Tesla’s safety systems is about 9x safer than the national average! This is not a coincidence. Teslas are designed for safety first from day one. 1/ The battery sits low in the floor, giving the car a low center of gravity and dramatically reduces rollover risk 2/ There are massive crumple zones to absorb energy before it reaches the cabin 3/ A rigid safety structure keeps the passenger space intact 4/ Cameras and AI software react faster than humans, cutting rear end crashes nearly in half 5/ On top of that, Teslas get safer over time with software updates continuously improving braking, pedestrian detection, and crash avoidance and more. People can debate their opinions on the internet all they want, but as a parent, all I care about are the facts and outcomes. And when something goes wrong, I want my family in the car with the lowest injury risk ever measured and the best real world safety record on the road. That’s why I choose Tesla. I don’t care about how it looks… even though I think they are the sexiest cars on the planet. I simply choose a Tesla for protection.
2:42

A Tesla is the #1 safest vehicle on Earth and I would never let my kids and the people I love drive anything other than a Tesla. Independent safety agencies around the world all come to the same conclusion. Teslas consistently earn the highest possible safety ratings and even set records no other car has beaten. Many still may not believe this, but safety is the #1 priority behind every single design decision at Tesla. The Tesla Model 3 and Y both hold 5-star overall ratings from U.S. regulators in every category: frontal crash, side crash, and rollover protection. The Model 3 also holds the lowest probability of serious occupant injury ever recorded in government testing - 5.7%, compared to 7-15% for most sedans and SUVs. No other vehicle has ever surpassed that. In Europe, the results are just as strong. The Model Y earned 98% adult occupant protection, 89% child protection, and one of the highest active safety scores ever measured. Also, in the real world, using data collected from billions of miles driven, Tesla’s own safety reports show: • 1 crash every ~6.36 million miles when Autopilot or supervised FSD is active • 1 crash every ~3.85 million miles with standard active safety features • All while the U.S. average is 1 crash every ~670,000 miles! Bro… this means driving with Tesla’s safety systems is about 9x safer than the national average! This is not a coincidence. Teslas are designed for safety first from day one. 1/ The battery sits low in the floor, giving the car a low center of gravity and dramatically reduces rollover risk 2/ There are massive crumple zones to absorb energy before it reaches the cabin 3/ A rigid safety structure keeps the passenger space intact 4/ Cameras and AI software react faster than humans, cutting rear end crashes nearly in half 5/ On top of that, Teslas get safer over time with software updates continuously improving braking, pedestrian detection, and crash avoidance and more. People can debate their opinions on the internet all they want, but as a parent, all I care about are the facts and outcomes. And when something goes wrong, I want my family in the car with the lowest injury risk ever measured and the best real world safety record on the road. That’s why I choose Tesla. I don’t care about how it looks… even though I think they are the sexiest cars on the planet. I simply choose a Tesla for protection.

Teslaconomics

14,952 次观看 • 7 个月前

The most valuable skill you can teach yourself right now is agent management. Nobody hands you a system for that, so here is mine. It runs at 22 to 25 pull requests a day, and sometimes 40. The agents scale without limit. They move fast, then they stop, because they need a yes, a no, or an actually-do-it-this-way from me. 1) Pin the threads that matter today. Everything else runs on its own and I get back to it when I can. 2) Check the pinned ones every 25 minutes. Slow is smooth and smooth is fast, and if I click through threads all day I get wiped out. 3) Keep the day's real priorities on paper. About 10 of those PRs are putting out fires, and the list is what stops the fires from eating the day. 4) Run one premium model as the parent thread and let it spin up children on a cheaper model. The catch: a few years ago you made 2 or 3 high-stakes decisions in a day. Now you make 10 to 20 before lunch, and you work a lot more. On the 18th I climbed Mount Washington with my son and had no signal, and I still shipped 8 PRs that morning while he was asleep. Your homework: do more than 50% of your work from your phone.
9:45

The most valuable skill you can teach yourself right now is agent management. Nobody hands you a system for that, so here is mine. It runs at 22 to 25 pull requests a day, and sometimes 40. The agents scale without limit. They move fast, then they stop, because they need a yes, a no, or an actually-do-it-this-way from me. 1) Pin the threads that matter today. Everything else runs on its own and I get back to it when I can. 2) Check the pinned ones every 25 minutes. Slow is smooth and smooth is fast, and if I click through threads all day I get wiped out. 3) Keep the day's real priorities on paper. About 10 of those PRs are putting out fires, and the list is what stops the fires from eating the day. 4) Run one premium model as the parent thread and let it spin up children on a cheaper model. The catch: a few years ago you made 2 or 3 high-stakes decisions in a day. Now you make 10 to 20 before lunch, and you work a lot more. On the 18th I climbed Mount Washington with my son and had no signal, and I still shipped 8 PRs that morning while he was asleep. Your homework: do more than 50% of your work from your phone.

The Startup Ideas Podcast (SIP) 🧃

11,736 次观看 • 15 天前

China just made Silicon Valley's entire AI industry look like a scam. The US government spent 3 years trying to stop China from building competitive AI. But this backfired HORRIBLY. Here's what happened: Yesterday, a Chinese startup called DeepSeek released a new AI model called V4. It matches the performance of OpenAI and Anthropic's best models. At 1/7th the price. And for the first time ever, it was built on Chinese chips. NOT American ones. That last part is the one that terrifies the west. For context: Since 2022, the US has banned the export of advanced AI chips to China. The entire strategy was built on the assumption that if China can't access Nvidia's best hardware, they can't build frontier AI. But DeepSeek just proved that assumption wrong. Their V4 model was trained and runs on Huawei's Ascend chips. Huawei spent months working directly with DeepSeek to make sure V4 runs across their entire line of AI processors. Jensen Huang even predicted this on a recent podcast: "The day that DeepSeek comes out on Huawei first, that is a horrible outcome for our nation." That day was yesterday. And the numbers are crazy: DeepSeek V4 costs $3.48 per million output tokens. OpenAI's latest model GPT-5.5 costs $30. Anthropic's Claude charges $25. Same ballpark performance. 7x cheaper. Uber's CTO just admitted they burned through their ENTIRE 2026 AI budget in 4 months using Anthropic's tools. If Uber had used DeepSeek instead, that same budget would have lasted 7 YEARS. 4 months vs 7 years. Same work getting done. But the pricing isn't even the big thing here. The real story is what DeepSeek did with their technical report: They published the benchmarks where they LOSE. Every AI company cherry-picks the tests where their model wins. DeepSeek ran the full comparison against GPT-5.4 and Google's Gemini, found they trail frontier models by 3 to 6 months, and printed it anyway. They literally don't care because the price gap makes the performance gap irrelevant for 90% of use cases. So the US export controls didn't slow China down. They ACCELERATED China's independence. Because Chinese developers were FORCED to train models with limited resources, they had to figure out how to make AI radically more efficient. That constraint became their competitive advantage. Every generation of DeepSeek has gotten dramatically cheaper to train. V4 continues the trend. Meanwhile US companies are going the OPPOSITE direction: OpenAI's GPT-5.5 Pro costs $180 per million output tokens. That's 51x more expensive than DeepSeek V4 for comparable work. The Commerce Secretary confirmed this week that ZERO Nvidia advanced chip shipments have actually gone through to China despite being approved in January. So China built frontier AI anyway. Without American chips. At a fraction of the cost. And the market response tells you everything: Chinese chipmaker SMIC surged 10%. Huahong Semiconductor jumped 15%. DeepSeek's Chinese AI competitors Zhipu AI and MiniMax dropped 9% because V4 is destroying them too. DeepSeek is making Silicon Valley's pricing model look like a scam. US tech companies spent $650 billion on AI infrastructure this year. DeepSeek just showed the world you can match their output for pennies. The export controls were supposed to be America's ace card. Instead they taught China how to win without American chips, at American prices nobody can compete with. Jensen Huang was right. This is a horrible outcome. But it's the outcome America built for itself.
1:06

China just made Silicon Valley's entire AI industry look like a scam. The US government spent 3 years trying to stop China from building competitive AI. But this backfired HORRIBLY. Here's what happened: Yesterday, a Chinese startup called DeepSeek released a new AI model called V4. It matches the performance of OpenAI and Anthropic's best models. At 1/7th the price. And for the first time ever, it was built on Chinese chips. NOT American ones. That last part is the one that terrifies the west. For context: Since 2022, the US has banned the export of advanced AI chips to China. The entire strategy was built on the assumption that if China can't access Nvidia's best hardware, they can't build frontier AI. But DeepSeek just proved that assumption wrong. Their V4 model was trained and runs on Huawei's Ascend chips. Huawei spent months working directly with DeepSeek to make sure V4 runs across their entire line of AI processors. Jensen Huang even predicted this on a recent podcast: "The day that DeepSeek comes out on Huawei first, that is a horrible outcome for our nation." That day was yesterday. And the numbers are crazy: DeepSeek V4 costs $3.48 per million output tokens. OpenAI's latest model GPT-5.5 costs $30. Anthropic's Claude charges $25. Same ballpark performance. 7x cheaper. Uber's CTO just admitted they burned through their ENTIRE 2026 AI budget in 4 months using Anthropic's tools. If Uber had used DeepSeek instead, that same budget would have lasted 7 YEARS. 4 months vs 7 years. Same work getting done. But the pricing isn't even the big thing here. The real story is what DeepSeek did with their technical report: They published the benchmarks where they LOSE. Every AI company cherry-picks the tests where their model wins. DeepSeek ran the full comparison against GPT-5.4 and Google's Gemini, found they trail frontier models by 3 to 6 months, and printed it anyway. They literally don't care because the price gap makes the performance gap irrelevant for 90% of use cases. So the US export controls didn't slow China down. They ACCELERATED China's independence. Because Chinese developers were FORCED to train models with limited resources, they had to figure out how to make AI radically more efficient. That constraint became their competitive advantage. Every generation of DeepSeek has gotten dramatically cheaper to train. V4 continues the trend. Meanwhile US companies are going the OPPOSITE direction: OpenAI's GPT-5.5 Pro costs $180 per million output tokens. That's 51x more expensive than DeepSeek V4 for comparable work. The Commerce Secretary confirmed this week that ZERO Nvidia advanced chip shipments have actually gone through to China despite being approved in January. So China built frontier AI anyway. Without American chips. At a fraction of the cost. And the market response tells you everything: Chinese chipmaker SMIC surged 10%. Huahong Semiconductor jumped 15%. DeepSeek's Chinese AI competitors Zhipu AI and MiniMax dropped 9% because V4 is destroying them too. DeepSeek is making Silicon Valley's pricing model look like a scam. US tech companies spent $650 billion on AI infrastructure this year. DeepSeek just showed the world you can match their output for pennies. The export controls were supposed to be America's ace card. Instead they taught China how to win without American chips, at American prices nobody can compete with. Jensen Huang was right. This is a horrible outcome. But it's the outcome America built for itself.

Ricardo

281,190 次观看 • 4 个月前

This Chinese developer linked two $2,999 NVIDIA DGX Sparks into one box and runs the full Qwen3-235B at home, after dropping his $1,999-a-month cloud bill to zero. He wired 2 small boxes into a single computer, split a giant 235-billion-parameter model in half between them, and serves it across his own network at about 10 tokens a second, with no internet, no cloud, right there on the desk. No data center, no thousand-dollar graphics cards, no monthly cloud bill. Just him, 2 gold boxes the size of a sandwich, one cable between them, and 1 power strip. And here is the whole payoff. He used to pay the cloud $1,999 a month for the same model, and the meter ticked on every request. Now he paid $5,998 once for 2 boxes, they covered their cost in 3 months, and after that he sends as many requests as he wants for free, only electricity. The two Sparks talk over one fast cable, each holds 128GB of memory, and together they carry the whole model, about 73GB loaded per box, with the chip inside pinned near the limit at 96%. Both boxes work as one and keep trading data over the cable, with no cloud in the loop and no single word leaking out. The ready model sits on one local address, and any app on his network calls it as easily as ChatGPT. And here is how he described, in plain words, what this pair of boxes does: "this is a pair of boxes that holds the huge Qwen3-235B model and serves it to one network. the model is split in half, and each box owns its half. parts: // Box 1 (holds the first half of the model and starts the answer fast, the first word appears in under a second) // Box 2 (holds the second half and writes out the rest, about 10 tokens a second) // Cable (connects the 2 boxes and moves data between them on every step, with no lag) // Address (one local address where any app sends its request, like to a cloud model) // Test (a script that runs big prompts through and measures speed and delays) // Monitor (checks temperature, power draw, and load on both boxes every 2 seconds). the model never goes to the cloud. he only steps in when a box runs hotter than 80 degrees or the cable between them starts dropping data." So the system knows exactly what it is, what it is for, and where its limits are. It knows it has to hold the whole huge model across 2 boxes on its own. It knows it has to answer every request locally, with no meter, no limits, and no internet. It knows the human is only needed when a box overheats or the link between them stalls. → The setup runs around the clock on 2 boxes, each pulling under 60 watts → However many requests he sends, the monthly bill is $0, only electricity → The first box starts the answer in under a second → The second writes text at about 10 tokens a second → One request at a time: 838 tokens in 85 seconds, first word in 0.8s → Two requests at once: 697 tokens in 108 seconds, first word in 0.7s → Both boxes sit at 96% load and warm up to 76-78 degrees And only when a chip in a box runs hotter than 80 degrees or the cable between the 2 Sparks drops data does the system call the owner. And when he himself is out on a run or in a coffee shop, he still reaches his own model at home from his phone: sends a big prompt to the local Qwen3-235B, gets the full answer back in under a minute and a half, with no token meter ticking and no limit to hit. Here is what the test shows on his screen during one of the night runs: "one request at a time: 838 tokens in 84.9 seconds, first word in 0.8s, then 0.1s per token." "two requests at once: 697 tokens in 107.6 seconds, first word in 0.7s, then 0.15s per token." "Box 1: chip at 96% load, 76 degrees, 56 watts, 73GB used in memory." "Box 2: chip at 96% load, 78 degrees, 56 watts, the Qwen3-235B model fully loaded." And while everyone around is paying for AI by the month and bumping into limits, his top-tier model just sits on the desk and works as much as he wants: his own little power plant instead of a forever meter. He has no server rack of his own and no cloud account behind it. Just 2 DGX Spark boxes on a desk, one model split in half between them, one local address, and a folder of prompts next to it. Out of everything I have seen this year, this is the cleanest way to stop paying for AI: $5,998 of hardware on the desk once, $0 a month to the cloud, unlimited forever, and between them 2 gold boxes, 1 cable, and the full Qwen3-235B answering at home with no internet.
1:16

This Chinese developer linked two $2,999 NVIDIA DGX Sparks into one box and runs the full Qwen3-235B at home, after dropping his $1,999-a-month cloud bill to zero. He wired 2 small boxes into a single computer, split a giant 235-billion-parameter model in half between them, and serves it across his own network at about 10 tokens a second, with no internet, no cloud, right there on the desk. No data center, no thousand-dollar graphics cards, no monthly cloud bill. Just him, 2 gold boxes the size of a sandwich, one cable between them, and 1 power strip. And here is the whole payoff. He used to pay the cloud $1,999 a month for the same model, and the meter ticked on every request. Now he paid $5,998 once for 2 boxes, they covered their cost in 3 months, and after that he sends as many requests as he wants for free, only electricity. The two Sparks talk over one fast cable, each holds 128GB of memory, and together they carry the whole model, about 73GB loaded per box, with the chip inside pinned near the limit at 96%. Both boxes work as one and keep trading data over the cable, with no cloud in the loop and no single word leaking out. The ready model sits on one local address, and any app on his network calls it as easily as ChatGPT. And here is how he described, in plain words, what this pair of boxes does: "this is a pair of boxes that holds the huge Qwen3-235B model and serves it to one network. the model is split in half, and each box owns its half. parts: // Box 1 (holds the first half of the model and starts the answer fast, the first word appears in under a second) // Box 2 (holds the second half and writes out the rest, about 10 tokens a second) // Cable (connects the 2 boxes and moves data between them on every step, with no lag) // Address (one local address where any app sends its request, like to a cloud model) // Test (a script that runs big prompts through and measures speed and delays) // Monitor (checks temperature, power draw, and load on both boxes every 2 seconds). the model never goes to the cloud. he only steps in when a box runs hotter than 80 degrees or the cable between them starts dropping data." So the system knows exactly what it is, what it is for, and where its limits are. It knows it has to hold the whole huge model across 2 boxes on its own. It knows it has to answer every request locally, with no meter, no limits, and no internet. It knows the human is only needed when a box overheats or the link between them stalls. → The setup runs around the clock on 2 boxes, each pulling under 60 watts → However many requests he sends, the monthly bill is $0, only electricity → The first box starts the answer in under a second → The second writes text at about 10 tokens a second → One request at a time: 838 tokens in 85 seconds, first word in 0.8s → Two requests at once: 697 tokens in 108 seconds, first word in 0.7s → Both boxes sit at 96% load and warm up to 76-78 degrees And only when a chip in a box runs hotter than 80 degrees or the cable between the 2 Sparks drops data does the system call the owner. And when he himself is out on a run or in a coffee shop, he still reaches his own model at home from his phone: sends a big prompt to the local Qwen3-235B, gets the full answer back in under a minute and a half, with no token meter ticking and no limit to hit. Here is what the test shows on his screen during one of the night runs: "one request at a time: 838 tokens in 84.9 seconds, first word in 0.8s, then 0.1s per token." "two requests at once: 697 tokens in 107.6 seconds, first word in 0.7s, then 0.15s per token." "Box 1: chip at 96% load, 76 degrees, 56 watts, 73GB used in memory." "Box 2: chip at 96% load, 78 degrees, 56 watts, the Qwen3-235B model fully loaded." And while everyone around is paying for AI by the month and bumping into limits, his top-tier model just sits on the desk and works as much as he wants: his own little power plant instead of a forever meter. He has no server rack of his own and no cloud account behind it. Just 2 DGX Spark boxes on a desk, one model split in half between them, one local address, and a folder of prompts next to it. Out of everything I have seen this year, this is the cleanest way to stop paying for AI: $5,998 of hardware on the desk once, $0 a month to the cloud, unlimited forever, and between them 2 gold boxes, 1 cable, and the full Qwen3-235B answering at home with no internet.

Blaze

93,871 次观看 • 3 个月前

Elon Musk just explained why the SpaceX IPO is an energy story and the energy constraint is why he believes space becomes the only viable path for AI to scale (Save this). The argument he is making is one of the most important and least understood things happening in technology right now. The United States currently consumes roughly 500 gigawatts of electricity on average. To double that capacity which is what continued AI expansion on the current terrestrial trajectory would eventually require would mean building as many power plants as currently exist in the entire country. He is not arguing that this is technically impossible, just that communities are not willing to accept it, that permitting timelines make it unrealistic, and that the hard ceiling on Earth based power generation means the expansion of AI compute will eventually hit a wall that no amount of capital can overcome on the ground. His observation is that in space, that wall does not exist. A solar panel in orbit produces roughly five times more power than the same panel on Earth, operates in continuous sunlight uninterrupted by weather or nighttime, and benefits from the vacuum of space as a completely passive cooling system meaning the two largest operating costs of any terrestrial data center, energy and cooling, are effectively eliminated. He then said that you could theoretically increase harnessed energy by a factor of one million and still be using less than a millionth of the sun's total energy output. This is the underlying physics of why SpaceX filed with the FCC to launch up to one million solar powered AI satellites, and why they described that constellation in their own filing as a first step toward becoming a Kardashev Type II civilization capable of harnessing the full power of the sun. To understand what makes this credible rather than visionary, you need to understand what SpaceX already controls that no other company on earth possesses. Starship, once operating at full cadence, can deliver 100 to 150 tons of payload to orbit per launch, at a target cost per kilogram that is an order of magnitude lower than any existing vehicle. Musk's stated ambition is to scale Starship to 10,000 to 30,000 launches per year, a frequency that would allow the deployment of orbital compute infrastructure at a pace that is currently unimaginable with any existing rocket. He told xAI staff earlier this year that achieving space-based AI at scale will eventually require manufacturing facilities on the moon, building solar panels and heat dissipation structures from lunar silicon and aluminum, and launching them into orbit from there rather than from Earth's surface because the moon's lower gravity makes the economics of launch dramatically more favorable. SpaceX's S-1 filing explicitly states that its launch capabilities could enable massive AI compute satellite constellations with the potential for millions of satellites for orbital data centers, with the first launch potentially occurring as soon as 2028. Google and Alphabet are already in advanced talks with SpaceX about deploying space-based data centers. Starcloud, a startup running Nvidia H100 GPUs in orbit, has already validated that high-performance AI inference workloads can operate in space, with plans to scale to five gigawatts of orbital compute power by 2035. This is why Musk believes the cost crossover happens in two to three years because SpaceX's launch cost trajectory intersects with the accelerating energy constraint on the ground in a way that makes space genuinely cheaper, faster, and less regulated at exactly the moment AI demand is hitting its hardest physical limits.
1:01

Elon Musk just explained why the SpaceX IPO is an energy story and the energy constraint is why he believes space becomes the only viable path for AI to scale (Save this). The argument he is making is one of the most important and least understood things happening in technology right now. The United States currently consumes roughly 500 gigawatts of electricity on average. To double that capacity which is what continued AI expansion on the current terrestrial trajectory would eventually require would mean building as many power plants as currently exist in the entire country. He is not arguing that this is technically impossible, just that communities are not willing to accept it, that permitting timelines make it unrealistic, and that the hard ceiling on Earth based power generation means the expansion of AI compute will eventually hit a wall that no amount of capital can overcome on the ground. His observation is that in space, that wall does not exist. A solar panel in orbit produces roughly five times more power than the same panel on Earth, operates in continuous sunlight uninterrupted by weather or nighttime, and benefits from the vacuum of space as a completely passive cooling system meaning the two largest operating costs of any terrestrial data center, energy and cooling, are effectively eliminated. He then said that you could theoretically increase harnessed energy by a factor of one million and still be using less than a millionth of the sun's total energy output. This is the underlying physics of why SpaceX filed with the FCC to launch up to one million solar powered AI satellites, and why they described that constellation in their own filing as a first step toward becoming a Kardashev Type II civilization capable of harnessing the full power of the sun. To understand what makes this credible rather than visionary, you need to understand what SpaceX already controls that no other company on earth possesses. Starship, once operating at full cadence, can deliver 100 to 150 tons of payload to orbit per launch, at a target cost per kilogram that is an order of magnitude lower than any existing vehicle. Musk's stated ambition is to scale Starship to 10,000 to 30,000 launches per year, a frequency that would allow the deployment of orbital compute infrastructure at a pace that is currently unimaginable with any existing rocket. He told xAI staff earlier this year that achieving space-based AI at scale will eventually require manufacturing facilities on the moon, building solar panels and heat dissipation structures from lunar silicon and aluminum, and launching them into orbit from there rather than from Earth's surface because the moon's lower gravity makes the economics of launch dramatically more favorable. SpaceX's S-1 filing explicitly states that its launch capabilities could enable massive AI compute satellite constellations with the potential for millions of satellites for orbital data centers, with the first launch potentially occurring as soon as 2028. Google and Alphabet are already in advanced talks with SpaceX about deploying space-based data centers. Starcloud, a startup running Nvidia H100 GPUs in orbit, has already validated that high-performance AI inference workloads can operate in space, with plans to scale to five gigawatts of orbital compute power by 2035. This is why Musk believes the cost crossover happens in two to three years because SpaceX's launch cost trajectory intersects with the accelerating energy constraint on the ground in a way that makes space genuinely cheaper, faster, and less regulated at exactly the moment AI demand is hitting its hardest physical limits.

Milk Road AI

12,738 次观看 • 3 个月前

🥳OK OK ，In a small vote, seem the community prefers Steampunk pistol more. So let's cook something more special this time . The greatest welcome to our hot agent Joi，She will bring the second class today. 📑“The fantasy of steampunk is broken down into gears and trajectories. The carving knife of 0 and 1 carves the ambition of the Victorian era. Highlighting the etched numbers, the algorithm is loading the violent aesthetics”. 🔫Create a weapon, just hand it over to Joi and after she sings magic, meet the industrial grade delivery standards. 👇Let's drive deeper about 【Technical Analysis of AI-Driven 3D Weapon Pipeline】 Core Technology Stack： 1⃣ NLP-Concept Binding Using the CLIP-Vit-L/14@336px cross-modal engine, descriptive terms such as "steampunk + brass + Victorian ballistics" are mapped to a 768-dimensional semantic space. Through the Latent Diffusion Model (k=25, cfg=7.5), a 1024px concept image is generated, with a focus on the bolt locking structure (key prompt weight x1.8). 2⃣ Topology Reconstruction Based on a NeRF-Transformer hybrid architecture, 2D concept images are parsed into a 256³ voxel grid (resolution 0.2mm). A non-rigid ICP algorithm is used to align moving parts like the trigger/barrel, with topology optimization iterations exceeding 500 times (MeshLab parameters: Remeshing_VCG 0.7). 3⃣ Procedural PBR Workflow Combining MaterialGAN to generate basic metallic textures, handcrafted features are injected through Style Transfer (normal map intensity 0.85, roughness mapping range 0.3-0.7). Rust effects are simulated using the Weber-Fechner perception model to mimic a 12-year oxidation cycle. 🔥 Based on a full-link generation system integrating natural language and geometric topology, this solution reduces the traditional modeling process from 72 hours to 37 seconds, with an error rate of less than 0.3mm³ (meeting FPS game firearm assembly standards). This technology has achieved an 89% reduction in modeling costs in AAA studio prototype verification. ✍️Finally, what props would you like Joi to make for you? Looking forward to assets being put on the chain？ Just leave your thoughts here.
0:37

🥳OK OK ，In a small vote, seem the community prefers Steampunk pistol more. So let's cook something more special this time . The greatest welcome to our hot agent Joi，She will bring the second class today. 📑“The fantasy of steampunk is broken down into gears and trajectories. The carving knife of 0 and 1 carves the ambition of the Victorian era. Highlighting the etched numbers, the algorithm is loading the violent aesthetics”. 🔫Create a weapon, just hand it over to Joi and after she sings magic, meet the industrial grade delivery standards. 👇Let's drive deeper about 【Technical Analysis of AI-Driven 3D Weapon Pipeline】 Core Technology Stack： 1⃣ NLP-Concept Binding Using the CLIP-Vit-L/14@336px cross-modal engine, descriptive terms such as "steampunk + brass + Victorian ballistics" are mapped to a 768-dimensional semantic space. Through the Latent Diffusion Model (k=25, cfg=7.5), a 1024px concept image is generated, with a focus on the bolt locking structure (key prompt weight x1.8). 2⃣ Topology Reconstruction Based on a NeRF-Transformer hybrid architecture, 2D concept images are parsed into a 256³ voxel grid (resolution 0.2mm). A non-rigid ICP algorithm is used to align moving parts like the trigger/barrel, with topology optimization iterations exceeding 500 times (MeshLab parameters: Remeshing_VCG 0.7). 3⃣ Procedural PBR Workflow Combining MaterialGAN to generate basic metallic textures, handcrafted features are injected through Style Transfer (normal map intensity 0.85, roughness mapping range 0.3-0.7). Rust effects are simulated using the Weber-Fechner perception model to mimic a 12-year oxidation cycle. 🔥 Based on a full-link generation system integrating natural language and geometric topology, this solution reduces the traditional modeling process from 72 hours to 37 seconds, with an error rate of less than 0.3mm³ (meeting FPS game firearm assembly standards). This technology has achieved an 89% reduction in modeling costs in AAA studio prototype verification. ✍️Finally, what props would you like Joi to make for you? Looking forward to assets being put on the chain？ Just leave your thoughts here.

Kingnet AI

17,798 次观看 • 1 年前

2026 Conservative Claire Coutinho claims the UK's pursuit of Net Zero is driven by ideology, and that using less wind and solar could save the equivalent of around £500 per household a year 2023: Conservative Claire Coutinho, "Nothing will distract us from achieving Net Zero or driving forward with renewables" The claim that less wind and solar would save every household £500 a year is deeply misleading. Coutinho is effectively taking the result of a model that changes almost the entire future electricity system—less wind and solar, much more nuclear and gas, different grid requirements and, crucially, no carbon price—and presenting the resulting £500-plus difference as though it were the cost of renewables. The research simply doesn't establish that. The figure does not come from evidence that renewables currently add £500 to household bills. It comes from a Conservative-backed modelling exercise comparing two hypothetical energy systems and turning £320 billion of estimated economy-wide savings between 2030 and 2050 into an annual equivalent per household. Crucially, the supposedly cheaper alternative doesn't simply build less wind and solar. It assumes far more nuclear power, substantially more unabated gas, lower spending on networks and balancing, and the removal of carbon pricing from electricity generation. Of the £320 billion headline saving, £137bn comes from assumed lower network costs, £67bn from balancing and ancillary services, £22bn from subsidies and capacity costs, and £94bn from lower wholesale prices. That £94 billion is particularly important because much of that supposed saving is created by changing the rules of the comparison. The existing-policy scenario assumes generators continue paying a carbon price which rises substantially towards 2050; the alternative effectively abolishes that carbon price from 2031. Unsurprisingly, a model in which gas-fired electricity no longer carries that cost produces cheaper wholesale electricity. But that isn't evidence that wind and solar made electricity £94bn more expensive. It is principally the consequence of assuming a major carbon-pricing policy is abolished. Nor is removing carbon pricing simply £94bn of free money. Carbon pricing is deliberately intended to make polluters account for their emissions, discourage fossil-fuel generation and encourage cleaner alternatives, while carbon allowances also generate government revenue. Removing it means accepting considerably greater emissions and losing those revenues and incentives. The alternative model consequently produces hundreds of millions of tonnes of additional CO₂. The model also assumes Britain can successfully build around 20GW of nuclear capacity, while retaining much greater dependence on gas. Whether nuclear can be delivered at the assumed cost and timetable is far from guaranteed, while greater reliance on gas leaves households more vulnerable to international gas-price shocks—the same exposure that played such a large role in Britain's recent energy-price crisis. There is a legitimate argument about the costs associated with large amounts of wind and solar. An electricity system dependent on intermittent generation requires substantial investment in transmission, balancing, storage, interconnectors and backup capacity, and those costs shouldn't be ignored. But that is very different from demonstrating that wind and solar cost every household £500 a year. Nearly £100bn of the supposed saving comes from the decision to remove carbon pricing, not from discovering that wind and solar are £100bn more expensive. Coutinho has taken a £500-plus headline generated by a Conservative-backed model containing a whole package of assumptions—less wind and solar, much more nuclear, more gas, less grid investment and the abolition of carbon pricing—and presented it to the public as evidence that Britain's reliance on wind and solar is costing households hundreds of pounds. That's misleading because the model doesn't establish that. It establishes that, if you change the entire future energy system and the taxes imposed upon it, its authors calculate a lower overall system cost. Turning that into “less wind and solar will save your household £500” strips away the assumptions that actually produce the number.
1:23

2026 Conservative Claire Coutinho claims the UK's pursuit of Net Zero is driven by ideology, and that using less wind and solar could save the equivalent of around £500 per household a year 2023: Conservative Claire Coutinho, "Nothing will distract us from achieving Net Zero or driving forward with renewables" The claim that less wind and solar would save every household £500 a year is deeply misleading. Coutinho is effectively taking the result of a model that changes almost the entire future electricity system—less wind and solar, much more nuclear and gas, different grid requirements and, crucially, no carbon price—and presenting the resulting £500-plus difference as though it were the cost of renewables. The research simply doesn't establish that. The figure does not come from evidence that renewables currently add £500 to household bills. It comes from a Conservative-backed modelling exercise comparing two hypothetical energy systems and turning £320 billion of estimated economy-wide savings between 2030 and 2050 into an annual equivalent per household. Crucially, the supposedly cheaper alternative doesn't simply build less wind and solar. It assumes far more nuclear power, substantially more unabated gas, lower spending on networks and balancing, and the removal of carbon pricing from electricity generation. Of the £320 billion headline saving, £137bn comes from assumed lower network costs, £67bn from balancing and ancillary services, £22bn from subsidies and capacity costs, and £94bn from lower wholesale prices. That £94 billion is particularly important because much of that supposed saving is created by changing the rules of the comparison. The existing-policy scenario assumes generators continue paying a carbon price which rises substantially towards 2050; the alternative effectively abolishes that carbon price from 2031. Unsurprisingly, a model in which gas-fired electricity no longer carries that cost produces cheaper wholesale electricity. But that isn't evidence that wind and solar made electricity £94bn more expensive. It is principally the consequence of assuming a major carbon-pricing policy is abolished. Nor is removing carbon pricing simply £94bn of free money. Carbon pricing is deliberately intended to make polluters account for their emissions, discourage fossil-fuel generation and encourage cleaner alternatives, while carbon allowances also generate government revenue. Removing it means accepting considerably greater emissions and losing those revenues and incentives. The alternative model consequently produces hundreds of millions of tonnes of additional CO₂. The model also assumes Britain can successfully build around 20GW of nuclear capacity, while retaining much greater dependence on gas. Whether nuclear can be delivered at the assumed cost and timetable is far from guaranteed, while greater reliance on gas leaves households more vulnerable to international gas-price shocks—the same exposure that played such a large role in Britain's recent energy-price crisis. There is a legitimate argument about the costs associated with large amounts of wind and solar. An electricity system dependent on intermittent generation requires substantial investment in transmission, balancing, storage, interconnectors and backup capacity, and those costs shouldn't be ignored. But that is very different from demonstrating that wind and solar cost every household £500 a year. Nearly £100bn of the supposed saving comes from the decision to remove carbon pricing, not from discovering that wind and solar are £100bn more expensive. Coutinho has taken a £500-plus headline generated by a Conservative-backed model containing a whole package of assumptions—less wind and solar, much more nuclear, more gas, less grid investment and the abolition of carbon pricing—and presented it to the public as evidence that Britain's reliance on wind and solar is costing households hundreds of pounds. That's misleading because the model doesn't establish that. It establishes that, if you change the entire future energy system and the taxes imposed upon it, its authors calculate a lower overall system cost. Turning that into “less wind and solar will save your household £500” strips away the assumptions that actually produce the number.

Farrukh

14,159 次观看 • 20 天前

The most dangerous thing a company can do right now is rent intelligence from the same place as its competitors (Save this). You cannot rent intelligence from the same place that rents it to your competitor as Chamath Palihapitiya points out. If every company in an industry is feeding their workflows into the same frontier model, they are all converging on the same outputs, the same decisions, the same product improvements. The model becomes the equalizer and everyone pays a premium to become more mediocre. This is happening exactly as Chamath predicted, and the evidence is now concrete. Anthropic and OpenAI have established what analysts are now openly calling an emerging model layer duopoly. Anthropic crossed $45 billion ARR in may 2026, more than tripling from $9 billion at the end of 2025, OpenAI was at roughly $24 to $33 billion ARR at the same time. Together, the two companies combined could hit $160 to $240 billion ARR by end of 2026 and Anthropic and OpenAI now control 88% of enterprise LLM spend. That concentration is the structural problem Chamath is pointing at. And Anthropic isn't just winning on merit because it's actively lobbying for regulatory outcomes that would make that duopoly permanent. Dario Amodei has explicitly framed open source models as unsafe, pushing a safety agenda that, if enshrined in regulation, would effectively make it illegal for enterprises to use the cheaper, private, sovereign alternatives locking them into a closed model dependency by government decree rather than by choice. So you have market forces producing a duopoly, and potential regulatory capture moving to enforce it from the top down. This is exactly why the Nvidia Palantir partnership is not just a product announcement but rather a strategic counter to that duopoly. The logic is straightforward from both sides because If you're Palantir, sitting at the application layer, the last thing you want is to be permanently beholden to Anthropic or OpenAI for the intelligence that powers your product. You want competitive model options, sovereignty and be able to tell enterprise customers they can run AI on their own infrastructure with their own data without any of it touching a frontier lab's servers. If you're Nvidia, sitting at the chip layer, an Anthropic-OpenAI duopoly is an existential concentration risk. Right now, Meta, Google, Microsoft, Amazon, and dozens of other companies buy Nvidia's hardware. If the model layer consolidates into two players, both of which are building their own chips Nvidia faces a monopsony where its best customers are building the tools to displace it. A healthy open source ecosystem where thousands of enterprises train, fine tune, and deploy their own models is Nvidia's ideal market structure. More buyers, more diversity, more demand, less pricing leverage from any single customer.
4:38

The most dangerous thing a company can do right now is rent intelligence from the same place as its competitors (Save this). You cannot rent intelligence from the same place that rents it to your competitor as Chamath Palihapitiya points out. If every company in an industry is feeding their workflows into the same frontier model, they are all converging on the same outputs, the same decisions, the same product improvements. The model becomes the equalizer and everyone pays a premium to become more mediocre. This is happening exactly as Chamath predicted, and the evidence is now concrete. Anthropic and OpenAI have established what analysts are now openly calling an emerging model layer duopoly. Anthropic crossed $45 billion ARR in may 2026, more than tripling from $9 billion at the end of 2025, OpenAI was at roughly $24 to $33 billion ARR at the same time. Together, the two companies combined could hit $160 to $240 billion ARR by end of 2026 and Anthropic and OpenAI now control 88% of enterprise LLM spend. That concentration is the structural problem Chamath is pointing at. And Anthropic isn't just winning on merit because it's actively lobbying for regulatory outcomes that would make that duopoly permanent. Dario Amodei has explicitly framed open source models as unsafe, pushing a safety agenda that, if enshrined in regulation, would effectively make it illegal for enterprises to use the cheaper, private, sovereign alternatives locking them into a closed model dependency by government decree rather than by choice. So you have market forces producing a duopoly, and potential regulatory capture moving to enforce it from the top down. This is exactly why the Nvidia Palantir partnership is not just a product announcement but rather a strategic counter to that duopoly. The logic is straightforward from both sides because If you're Palantir, sitting at the application layer, the last thing you want is to be permanently beholden to Anthropic or OpenAI for the intelligence that powers your product. You want competitive model options, sovereignty and be able to tell enterprise customers they can run AI on their own infrastructure with their own data without any of it touching a frontier lab's servers. If you're Nvidia, sitting at the chip layer, an Anthropic-OpenAI duopoly is an existential concentration risk. Right now, Meta, Google, Microsoft, Amazon, and dozens of other companies buy Nvidia's hardware. If the model layer consolidates into two players, both of which are building their own chips Nvidia faces a monopsony where its best customers are building the tools to displace it. A healthy open source ecosystem where thousands of enterprises train, fine tune, and deploy their own models is Nvidia's ideal market structure. More buyers, more diversity, more demand, less pricing leverage from any single customer.

Milk Road AI

34,385 次观看 • 2 个月前

People don't realize that factories in Asia and factories in the USA are totally different. I think AI is going to help the Asian factories much more than the American ones. Factories in Asia are high mix. Factories in the USA are low mix. This means that factories in the USA make significantly fewer products, whereas factories in Asia are way more flexible, mainly because of the huge differential in labor costs. The gigafactory in Austin, one of the biggest factories in the world, makes 2 products: Tesla Model Y The Cybertruck That's it. The factory in the video makes 1000s of different products every year. When you're making the same thing over and over again, maybe in a few different colors, you're much better off using normal automation which is lower error than AI and human design. But when you're making 1000s of different products, using AI to spin up product 1001 starts to kind of make sense. There is not time or logic to using traditional automation; you're better off using AI to make a fast quality control mechanism and then repair errors using your inexpensive human labor. People are viewing AI in factories like there is one type of factory, but the way factories run in high cost countries vs. low cost countries is wildly different. We're just at the beginning of what will be an epic change for not only manufacturing, but the cost and quality of the goods we consume. The real internet of things has arrived.
1:32

People don't realize that factories in Asia and factories in the USA are totally different. I think AI is going to help the Asian factories much more than the American ones. Factories in Asia are high mix. Factories in the USA are low mix. This means that factories in the USA make significantly fewer products, whereas factories in Asia are way more flexible, mainly because of the huge differential in labor costs. The gigafactory in Austin, one of the biggest factories in the world, makes 2 products: Tesla Model Y The Cybertruck That's it. The factory in the video makes 1000s of different products every year. When you're making the same thing over and over again, maybe in a few different colors, you're much better off using normal automation which is lower error than AI and human design. But when you're making 1000s of different products, using AI to spin up product 1001 starts to kind of make sense. There is not time or logic to using traditional automation; you're better off using AI to make a fast quality control mechanism and then repair errors using your inexpensive human labor. People are viewing AI in factories like there is one type of factory, but the way factories run in high cost countries vs. low cost countries is wildly different. We're just at the beginning of what will be an epic change for not only manufacturing, but the cost and quality of the goods we consume. The real internet of things has arrived.

molson 🧠⚙️

66,129 次观看 • 8 个月前

Huge news - proper electric vehicles are about to get to Polo Vivo prices - less than R350,000. This is the BYD Dolphin Mini - which will be sold as the Atto 1 in South Africa - and based on what I’ve been told here in Zhengzhou and can calculate, this car is will be SA’s cheapest EV by some margin and will officially go on sale in September of this year. In this post I’m going to dive into charging time and what it will cost to charge, we’ll take a look at the interior and I’ll report on our exclusive first drive for The Cars.co.za Team and SA media, as well as give you my opinion on the effect this car is going to have on the new and used car markets in our country. For scale, it is about the same size as the Mini Cooper but feels much more spacious interior. Most impressive for a budget EV, it’s fast-charging capabilities mean you can charge the smaller-battery version from 20-80% in 16 minutes at a mall or petrol station, or 2 hours and 34 minutes at home. Battery sizes and range: More affordable variant: - Battery size: 30kWh - Claimed range 220km High spec variant - Battery size: 42kWh - Claimed range 310km Keep in mind that real world range is reduced by a variety of factors such as average speed (highway cruising in particular), average temperature (cold is more problematic than heat), amount of regeneration per journey and driving style. The Atto 1 features a single motor which offers between 65kW and 115kW in the highest spec, providing acceleration of 0-100 in 9 to 11 seconds, all of which on my quick test drive around the facility felt adequate for SA roads, especially when I put my foot flat to simulate overtaking. The interior felt a little plasticky in places but overall was far, far more advanced and luxurious than any car you’ll find at this price point in South Africa. BYD South Africa published an official press release today but I’m in China with BYD at the moment and the South African press contingent were given an exclusive first look and first drive after our full-on day as the first journalists in the world to experience the new BYD International Circuit facility, which I suppose you could now call the spiritual home of BYD Global’s flagship YangWang sub-brand. In summary, in my opinion this could be a turning point for EVs in South Africa. EV day-to-day running costs, especially if you charge at home, are generally much lower than traditional running a normal car, and EVs require far less maintenance and repairs. What I’m particularly excited about is the potential of a car like the Atto 1 to transform the used car market; the battery warranty of 8 years and expected useful life of 12-15 years, and those low maintenance and repair costs, all coupled with depreciation means in 3-5 years we’ll all probably be able to buy a good used Atto 1 or similar EV for around R200,000 or less depending on age and mileage. Now that’s exciting for me! Stay tuned to all the The Cars.co.za Team platforms for more info on this really exciting development on the South African motoring landscape🇿🇦🏁
1:06

Huge news - proper electric vehicles are about to get to Polo Vivo prices - less than R350,000. This is the BYD Dolphin Mini - which will be sold as the Atto 1 in South Africa - and based on what I’ve been told here in Zhengzhou and can calculate, this car is will be SA’s cheapest EV by some margin and will officially go on sale in September of this year. In this post I’m going to dive into charging time and what it will cost to charge, we’ll take a look at the interior and I’ll report on our exclusive first drive for The Cars.co.za Team and SA media, as well as give you my opinion on the effect this car is going to have on the new and used car markets in our country. For scale, it is about the same size as the Mini Cooper but feels much more spacious interior. Most impressive for a budget EV, it’s fast-charging capabilities mean you can charge the smaller-battery version from 20-80% in 16 minutes at a mall or petrol station, or 2 hours and 34 minutes at home. Battery sizes and range: More affordable variant: - Battery size: 30kWh - Claimed range 220km High spec variant - Battery size: 42kWh - Claimed range 310km Keep in mind that real world range is reduced by a variety of factors such as average speed (highway cruising in particular), average temperature (cold is more problematic than heat), amount of regeneration per journey and driving style. The Atto 1 features a single motor which offers between 65kW and 115kW in the highest spec, providing acceleration of 0-100 in 9 to 11 seconds, all of which on my quick test drive around the facility felt adequate for SA roads, especially when I put my foot flat to simulate overtaking. The interior felt a little plasticky in places but overall was far, far more advanced and luxurious than any car you’ll find at this price point in South Africa. BYD South Africa published an official press release today but I’m in China with BYD at the moment and the South African press contingent were given an exclusive first look and first drive after our full-on day as the first journalists in the world to experience the new BYD International Circuit facility, which I suppose you could now call the spiritual home of BYD Global’s flagship YangWang sub-brand. In summary, in my opinion this could be a turning point for EVs in South Africa. EV day-to-day running costs, especially if you charge at home, are generally much lower than traditional running a normal car, and EVs require far less maintenance and repairs. What I’m particularly excited about is the potential of a car like the Atto 1 to transform the used car market; the battery warranty of 8 years and expected useful life of 12-15 years, and those low maintenance and repair costs, all coupled with depreciation means in 3-5 years we’ll all probably be able to buy a good used Atto 1 or similar EV for around R200,000 or less depending on age and mileage. Now that’s exciting for me! Stay tuned to all the The Cars.co.za Team platforms for more info on this really exciting development on the South African motoring landscape🇿🇦🏁

Ciro De Siena

22,982 次观看 • 1 年前

The first question I asked Elon Musk: What’s the point of sending GPUs into space? The whole idea behind orbital data centers is that if the launch costs continue to drop, it will become cheaper to put GPUs in orbit than to build power plants on Earth. The problem with this argument is that energy is only about 15% of a datacenter’s lifetime cost. The chips themselves are around 70%. And you still have to launch those to space! Elon kept returning to one point over and over again: It will simply not be physically possible to scale power production to the scale needed for AI on Earth. He kept pointing out the bottlenecks we’ve already run into on Earth: You can’t plug into the utilities - the interconnect queues are too long. You can’t do behind-the-meter natural gas and generate power yourself - lead times for turbines stretch past 2030. You can’t do solar on Earth, because of permits, and because of the tariffs. For it to make economical sense to shift compute to space, all of the following things would need to be true: - Power generation on Earth hits a ceiling, or AI demand outstrips every terrestrial option (for context, 1 TW of solar power is only 1% of the land area of the US, and AI currently only uses about 20 GW globally). - Chip production scales faster than power generation (because Elon builds TeraFab). It would be surprising if building and placing solar panels turned out to be harder than scaling semiconductor manufacturing. - Starship reaches thousands of launches per year. In that world, Elon wins the AI race outright. SpaceX is the only entity that can launch at that scale. xAI would have unlimited power. Everyone else will be stuck fighting over grid interconnects and turbine orders. And if those 3 conditions aren’t met? Well, on Earth, xAI is just gonna be one of the pack anyways - and there’s no market for the 4th best AI model. Elon’s comparative advantage was never going to be navigating utility interconnect queues or filing permits faster than Google. His advantage is SpaceX. So why not just bet on the world where SpaceX becomes the kingmaker? I asked Elon what that world looks like. 100 GW = 10,000 starship launches, and he wants to do more than that every year by 2030. That’s one starship launch every hour.
5:33

The first question I asked Elon Musk: What’s the point of sending GPUs into space? The whole idea behind orbital data centers is that if the launch costs continue to drop, it will become cheaper to put GPUs in orbit than to build power plants on Earth. The problem with this argument is that energy is only about 15% of a datacenter’s lifetime cost. The chips themselves are around 70%. And you still have to launch those to space! Elon kept returning to one point over and over again: It will simply not be physically possible to scale power production to the scale needed for AI on Earth. He kept pointing out the bottlenecks we’ve already run into on Earth: You can’t plug into the utilities - the interconnect queues are too long. You can’t do behind-the-meter natural gas and generate power yourself - lead times for turbines stretch past 2030. You can’t do solar on Earth, because of permits, and because of the tariffs. For it to make economical sense to shift compute to space, all of the following things would need to be true: - Power generation on Earth hits a ceiling, or AI demand outstrips every terrestrial option (for context, 1 TW of solar power is only 1% of the land area of the US, and AI currently only uses about 20 GW globally). - Chip production scales faster than power generation (because Elon builds TeraFab). It would be surprising if building and placing solar panels turned out to be harder than scaling semiconductor manufacturing. - Starship reaches thousands of launches per year. In that world, Elon wins the AI race outright. SpaceX is the only entity that can launch at that scale. xAI would have unlimited power. Everyone else will be stuck fighting over grid interconnects and turbine orders. And if those 3 conditions aren’t met? Well, on Earth, xAI is just gonna be one of the pack anyways - and there’s no market for the 4th best AI model. Elon’s comparative advantage was never going to be navigating utility interconnect queues or filing permits faster than Google. His advantage is SpaceX. So why not just bet on the world where SpaceX becomes the kingmaker? I asked Elon what that world looks like. 100 GW = 10,000 starship launches, and he wants to do more than that every year by 2030. That’s one starship launch every hour.

Dwarkesh Patel

405,710 次观看 • 7 个月前