Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

Bambu Lab A1 mini - tested the advertised speed/acceleration! 500 mm/s / 10k mm/s² 0.12mm Layer height / 0.4mm Line width Sunlu PLA+ at 230°C going 24 mm³/s Sunlu spools don't fit on the AMS lite (inner diameter is too wide) I forgot to record on the vertical "camera" ;-)

Albert is 247printing

4,476 subscribers

16,926 просмотров • 2 лет назад •via X (Twitter)

Образование Наука и технологии

Anya Rossi• Live Now

Private livecam show

Комментарии: 10

Фото профиля Pooch @Repkord

Pooch @Repkord2 лет назад

Looking forward to your overview!

Фото профиля Nathan Builds Robots

Nathan Builds Robots2 лет назад

Oh my, it’s too quiet for a speed printer. Is there a way to make it louder?

Фото профиля re:printed 3D

re:printed 3D2 лет назад

Someone will make an adapter for the Sunlu (and other) spools that don't fit; it's a minor thing. I look forward to watching your video, Albert.

Фото профиля Corey - Mojee3D

Corey - Mojee3D2 лет назад

Hey @ModBotArmy, I know you/we tried 0.12mm layer height but did we try 0.4mm width? If not, that may have been the difference.

Фото профиля eurusd

eurusd2 лет назад

Are you happy with the level of noise ?

Фото профиля fatboy1271

fatboy12712 лет назад

Nice video! Hopefully my early disdain for this will be lessened as I see more of it 🤘🤘

Фото профиля ImmortalTortoise

ImmortalTortoise2 лет назад

need more angles

Фото профиля Farad

Farad2 лет назад

Excuse me what? How on earth.

Фото профиля Werner - WF3D

Werner - WF3D2 лет назад

Interesting! I think I am most curious about the AMS. Which spools fit which not etc. Glad you have one for toying around. Looking forward to your video!

Фото профиля SimonX

SimonX2 лет назад

I'm shocked how much desk space it has to occupy! Given the printer is only so small... 😩

Похожие видео

Tyler Herro on the post Houston Rockets melee: I hope I don't get suspended. I don't think I shouldn't have been ejected....It's 2025 the jokes are going to fly s*** is going to happen, 2024 same s*** the jokes are going to fly. I was laughing at it too. I was laughing during it. I had no hostility towards the Rockets. I played out that way. I like getting under people's skin anyway. That's what I do. #heatnation

Tyler Herro on the post Houston Rockets melee: I hope I don't get suspended. I don't think I shouldn't have been ejected....It's 2025 the jokes are going to fly s* is going to happen, 2024 same s* the jokes are going to fly. I was laughing at it too. I was laughing during it. I had no hostility towards the Rockets. I played out that way. I like getting under people's skin anyway. That's what I do. #heatnation

Brendan Tobin

974,062 просмотров • 1 год назад

🚁 Epic drone flyover of Tesla's Fremont Factory! Captured some incredible and historic footage today: Cybercabs actively testing — smoothly handling speed bumps before launching hard down the track. I also spotted Model S and Model X vehicles in the launch area along with what appear to be some of the final S & X rolling off the assembly line as their legendary run at Fremont wraps up. The Model Y production line was surprisingly quiet during the flight, but I got great aerial views of the busy shipping yard and the factory’s recycling area too. From next-gen Cybercab action to the end of an era with the S & X — this flyover really shows the transition happening at one of the world’s most important EV factories. Fremont never disappoints! What do you think about the Cybercab testing and the final S/X coming off the line? Drop your thoughts in the comments 👇 Shot on Antigravity A1: Tesla Tesla_na Elon Musk

🚁 Epic drone flyover of Tesla's Fremont Factory! Captured some incredible and historic footage today: Cybercabs actively testing — smoothly handling speed bumps before launching hard down the track. I also spotted Model S and Model X vehicles in the launch area along with what appear to be some of the final S & X rolling off the assembly line as their legendary run at Fremont wraps up. The Model Y production line was surprisingly quiet during the flight, but I got great aerial views of the busy shipping yard and the factory’s recycling area too. From next-gen Cybercab action to the end of an era with the S & X — this flyover really shows the transition happening at one of the world’s most important EV factories. Fremont never disappoints! What do you think about the Cybercab testing and the final S/X coming off the line? Drop your thoughts in the comments 👇 Shot on Antigravity A1: Tesla Tesla_na Elon Musk

Tesla Owners Silicon Valley

39,192 просмотров • 4 месяцев назад

i spent 3 hours finding the sweet spot for hermes 4.3 36B on a single RTX 3090. saving you the trouble, anon. the model is 21.8GB at Q4_K_M. that leaves 2.2GB free on 24GB VRAM. not much room for KV cache. here's what actually happened: 4K: 35.3 tok/s 8K: 35.2 tok/s 16K: 34.8 tok/s 32K: 34.6 tok/s 64K: 6.4 tok/s 128K: 1.9 tok/s flat from 4K to 32K. then it falls off a cliff at 64K. the trick is quantized KV cache. without it you OOM at 16K. with quantized KV cache you get 32K at full speed. all 64 layers on GPU. at 64K something weird happens. ngl 99 (all layers on GPU) = 3.96 tok/s. the KV cache silently spills to CPU. drop to ngl 55 and speed jumps to 6.37. drop to ngl 48 and it gets worse again (3.46). there's an offload sweet spot where you free just enough VRAM for the cache without losing too much compute to PCIe transfers. 128K works at ngl 32 but you're at 1.95 tok/s. half the model on CPU. usable for batch work, not for interactive. the sweet spot command: llama-server -m hermes-4.3-36b-Q4_K_M.gguf -ngl 99 -c 32768 --cache-type-k q4_0 --cache-type-v q4_0 32K context. 34.6 tok/s. all on GPU. this is where dense 36B lives on 24GB. for comparison, qwen 3.5 (35B MoE, 3B active) holds 112 tok/s from 4K all the way to 262K on the same GPU. no speed drop. same total params, completely different architecture. hybrid linear attention means flat context scaling. dense pays for every token in the KV cache. code and quality comparison coming next. fast vs slow generation side by side in the videos below.

i spent 3 hours finding the sweet spot for hermes 4.3 36B on a single RTX 3090. saving you the trouble, anon. the model is 21.8GB at Q4_K_M. that leaves 2.2GB free on 24GB VRAM. not much room for KV cache. here's what actually happened: 4K: 35.3 tok/s 8K: 35.2 tok/s 16K: 34.8 tok/s 32K: 34.6 tok/s 64K: 6.4 tok/s 128K: 1.9 tok/s flat from 4K to 32K. then it falls off a cliff at 64K. the trick is quantized KV cache. without it you OOM at 16K. with quantized KV cache you get 32K at full speed. all 64 layers on GPU. at 64K something weird happens. ngl 99 (all layers on GPU) = 3.96 tok/s. the KV cache silently spills to CPU. drop to ngl 55 and speed jumps to 6.37. drop to ngl 48 and it gets worse again (3.46). there's an offload sweet spot where you free just enough VRAM for the cache without losing too much compute to PCIe transfers. 128K works at ngl 32 but you're at 1.95 tok/s. half the model on CPU. usable for batch work, not for interactive. the sweet spot command: llama-server -m hermes-4.3-36b-Q4_K_M.gguf -ngl 99 -c 32768 --cache-type-k q4_0 --cache-type-v q4_0 32K context. 34.6 tok/s. all on GPU. this is where dense 36B lives on 24GB. for comparison, qwen 3.5 (35B MoE, 3B active) holds 112 tok/s from 4K all the way to 262K on the same GPU. no speed drop. same total params, completely different architecture. hybrid linear attention means flat context scaling. dense pays for every token in the KV cache. code and quality comparison coming next. fast vs slow generation side by side in the videos below.

Sudo su

52,268 просмотров • 5 месяцев назад

"Incentives don't create product-market fit; they amplify it." Caught NK💿 's talk at the Optimism Scaling Summit in ETHDenver 🏔🦬🦄 . Here's the key takeaway: The real edge crypto has over Web2 is giving users a stake in the protocol. But the industry got too focused on tokenomics and forgot to build things people actually want. Startale is building the infrastructure and Soneium 💿 is finding the entertainment use cases with real product-market fit. 👇

"Incentives don't create product-market fit; they amplify it." Caught NK💿 's talk at the Optimism Scaling Summit in ETHDenver 🏔🦬🦄 . Here's the key takeaway: The real edge crypto has over Web2 is giving users a stake in the protocol. But the industry got too focused on tokenomics and forgot to build things people actually want. Startale is building the infrastructure and Soneium 💿 is finding the entertainment use cases with real product-market fit. 👇

Startale 💿

14,921 просмотров • 6 месяцев назад

Well, I don’t remember going under? Thank you so much for your support! I’ll look at the data later today!- I got the results and great news, two small 5 mm polyps removed, and likely benign. Of course they will be tested. 27 years after surviving rare cancer, I don't take screenings lightly. I just took a 3-hour nap. Now, it's back to looking at the weather. There is an interesting system crossing the west coast Thursday morning that I will update you on later. #Colonoscopy #EarlyDetection #PreventiveCare

Well, I don’t remember going under? Thank you so much for your support! I’ll look at the data later today!- I got the results and great news, two small 5 mm polyps removed, and likely benign. Of course they will be tested. 27 years after surviving rare cancer, I don't take screenings lightly. I just took a 3-hour nap. Now, it's back to looking at the weather. There is an interesting system crossing the west coast Thursday morning that I will update you on later. #Colonoscopy #EarlyDetection #PreventiveCare

Gary Lezak

43,631 просмотров • 6 месяцев назад

🚨 S&P 500 TOP IS CLOSER THAN YOU THINK The index broke above 7,500, and everyone is celebrating. That's exactly what concerns me. At the same time, OpenAI and Anthropic are preparing what could become two of the largest IPOs ever. Most investors see that as bullish. I don't. IPOs don't create new money. They redirect existing money. Every dollar allocated to those deals is a dollar that isn't going into stocks, ETFs, or the same index currently sitting near all-time highs. We've seen this setup before. Late-cycle euphoria, massive IPOs, and investors willing to buy anything at any price. The biggest listings of 2021 arrived near the top. The market rolled over shortly after. Now we have record valuations, record optimism, and potentially record IPO supply all hitting at the same time. My long-term target remains unchanged: S&P 500 < 4,800 I called the 2025 BTC top and the move to $60K. Don't miss the next call! - Turn your notifications on!

🚨 S&P 500 TOP IS CLOSER THAN YOU THINK The index broke above 7,500, and everyone is celebrating. That's exactly what concerns me. At the same time, OpenAI and Anthropic are preparing what could become two of the largest IPOs ever. Most investors see that as bullish. I don't. IPOs don't create new money. They redirect existing money. Every dollar allocated to those deals is a dollar that isn't going into stocks, ETFs, or the same index currently sitting near all-time highs. We've seen this setup before. Late-cycle euphoria, massive IPOs, and investors willing to buy anything at any price. The biggest listings of 2021 arrived near the top. The market rolled over shortly after. Now we have record valuations, record optimism, and potentially record IPO supply all hitting at the same time. My long-term target remains unchanged: S&P 500 < 4,800 I called the 2025 BTC top and the move to $60K. Don't miss the next call! - Turn your notifications on!

Ulun

38,701 просмотров • 2 месяцев назад

Jon, Before we get to the math on the charge dimensions you just described, and how they are missing 12+% of the explosives you claim went off, we are going to talk about why you keep changing things up. 24 hours ago, you claimed the 0.5 gram "would be a 10mm x 15mm charge" (see first video)! 12 hours later, you claimed the charge was 7.0 mm x 7.34 mm (shocking, I know). Of course, this was after I pointed out how absurdly too large that would be for 0.5 grams of PETN. The problem you can not change Since you changed the charge size in your response, I figured I would point out the obvious gaslighting to the world. 1st, as previously noted, the optimum standoff distance is 1.5x the diameter of the cone. In the case of the last dimensions you provided, the standoff distance would be about 11 mm, which you factored into your devices overall size. So this is my question for you Jon (and all of your wisdom). With the expected penetration (EP) only 1/2 the diameter of the cone, why is it positioned so far back in the device in your diagram? Someone who built such a sophisticated charge would certainly be able to rearrange the internals to give it a snowballs chance in hell of actually working, right? Even with 3x the distance of the EP away from the end of the device dimension you provided, the theoretical jet would be extremely dissipated and STILL INSIDE THE MIC! By the time it leaves the device and hits the air outside, it would be almost totally, if not completely dissipated. So, why would you place it so far back if you ACTUALLY KNEW how this charge worked? Simply put, the explosive charge properly placed int he device wouldn't need a built in standoff. New Dimensions Are Still 12% Cubic Inches Short For 0.5 grams of PETN In your attempt to correct yourself before being further exposed, you forgot to take into account the empty space of the cone leaving you 12% worth of explosives short for a 0.05 gram charge. That is very significant, especially for a charge this size. See graph paper image with all the math. SUMMARY 1. You changed the charge dimensions in the same day 10mm x 15mm -> 7.34 mm x 7.0 mm 2. Second set of charge dimensions you gave yesterday are 12% shy for 0.5 grams of PETN. 3. At 3x max optimal distance for EP of 3.7 mm, you are still inside the mic and the jet would be worthless by the time it exited. 4. Charge placement and why you decided it needed an internal standoff to begin with. All of these point to one thing. You continue to pull this out your butt. Thank you for your attention on this matter.

Jon, Before we get to the math on the charge dimensions you just described, and how they are missing 12+% of the explosives you claim went off, we are going to talk about why you keep changing things up. 24 hours ago, you claimed the 0.5 gram "would be a 10mm x 15mm charge" (see first video)! 12 hours later, you claimed the charge was 7.0 mm x 7.34 mm (shocking, I know). Of course, this was after I pointed out how absurdly too large that would be for 0.5 grams of PETN. The problem you can not change Since you changed the charge size in your response, I figured I would point out the obvious gaslighting to the world. 1st, as previously noted, the optimum standoff distance is 1.5x the diameter of the cone. In the case of the last dimensions you provided, the standoff distance would be about 11 mm, which you factored into your devices overall size. So this is my question for you Jon (and all of your wisdom). With the expected penetration (EP) only 1/2 the diameter of the cone, why is it positioned so far back in the device in your diagram? Someone who built such a sophisticated charge would certainly be able to rearrange the internals to give it a snowballs chance in hell of actually working, right? Even with 3x the distance of the EP away from the end of the device dimension you provided, the theoretical jet would be extremely dissipated and STILL INSIDE THE MIC! By the time it leaves the device and hits the air outside, it would be almost totally, if not completely dissipated. So, why would you place it so far back if you ACTUALLY KNEW how this charge worked? Simply put, the explosive charge properly placed int he device wouldn't need a built in standoff. New Dimensions Are Still 12% Cubic Inches Short For 0.5 grams of PETN In your attempt to correct yourself before being further exposed, you forgot to take into account the empty space of the cone leaving you 12% worth of explosives short for a 0.05 gram charge. That is very significant, especially for a charge this size. See graph paper image with all the math. SUMMARY 1. You changed the charge dimensions in the same day 10mm x 15mm -> 7.34 mm x 7.0 mm 2. Second set of charge dimensions you gave yesterday are 12% shy for 0.5 grams of PETN. 3. At 3x max optimal distance for EP of 3.7 mm, you are still inside the mic and the jet would be worthless by the time it exited. 4. Charge placement and why you decided it needed an internal standoff to begin with. All of these point to one thing. You continue to pull this out your butt. Thank you for your attention on this matter.

Matt Tardio

35,788 просмотров • 1 месяц назад

260817 First LEOWON Live! 🐰🦁 #상원 #리오 🦁: It’s okay. It’s okay. Actually, I was... 🦁: I mean, I wasn't going to bring it up, but anyway, wearing this back support brace... 🦁: It's because they're worried I might get hurt. 🦁: Yeah, so from my perspective, it's just for management and care. 🐰: Right, right. 🦁: You really don't need to worry about it. 🦁: It's not like I got injured, it's just that my back hasn't been good since I was young. 🐰: Mm. 🦁: Really, ever since elementary school, it wasn't good. 🦁: So, it's always been... honest to god, it just was never good, to be completely honest. 🦁: But, you know, our dancing is really intense and the choreography amount is heavy, so... 🦁: Getting hurt hurts way too much, so I'm doing this for care. 🐰: But still, Leo hyung is always managing and handling everything so well. 🦁: Mm, so you really don't need to worry. 🐰: And the staff members are always monitoring and checking in on him to see how he's doing. 🦁: Mm, right, right. 🐰: So really, you don't need to worry at all. 🦁: The company also worries a lot and takes care of it, so... 🦁: You really don't have to worry!

260817 First LEOWON Live! 🐰🦁 #상원 #리오 🦁: It’s okay. It’s okay. Actually, I was... 🦁: I mean, I wasn't going to bring it up, but anyway, wearing this back support brace... 🦁: It's because they're worried I might get hurt. 🦁: Yeah, so from my perspective, it's just for management and care. 🐰: Right, right. 🦁: You really don't need to worry about it. 🦁: It's not like I got injured, it's just that my back hasn't been good since I was young. 🐰: Mm. 🦁: Really, ever since elementary school, it wasn't good. 🦁: So, it's always been... honest to god, it just was never good, to be completely honest. 🦁: But, you know, our dancing is really intense and the choreography amount is heavy, so... 🦁: Getting hurt hurts way too much, so I'm doing this for care. 🐰: But still, Leo hyung is always managing and handling everything so well. 🦁: Mm, so you really don't need to worry. 🐰: And the staff members are always monitoring and checking in on him to see how he's doing. 🦁: Mm, right, right. 🐰: So really, you don't need to worry at all. 🦁: The company also worries a lot and takes care of it, so... 🦁: You really don't have to worry!

📁

31,973 просмотров • 11 дней назад

🚨 STOP AND LOOK AT THIS CHART 🚨 White candles are the dot-com bubble The colored candles are right now Do you see where I'm going with this? We're hanging by a thread, on the verge of collapse, but some people still believe in endless growth In 2000 the narrative was "the internet changes everything" In 2026 the narrative is "AI changes everything" Both times the market priced in decades of growth in months Both times the green support line held - until it didn't We're sitting at the peak of the right side right now Do you believe this overlay is a coincidence? I don't Just a reminder that I perfectly predicted the moves on S&P 500 and GOLD Turn on notifications so you don't miss future updates

🚨 STOP AND LOOK AT THIS CHART 🚨 White candles are the dot-com bubble The colored candles are right now Do you see where I'm going with this? We're hanging by a thread, on the verge of collapse, but some people still believe in endless growth In 2000 the narrative was "the internet changes everything" In 2026 the narrative is "AI changes everything" Both times the market priced in decades of growth in months Both times the green support line held - until it didn't We're sitting at the peak of the right side right now Do you believe this overlay is a coincidence? I don't Just a reminder that I perfectly predicted the moves on S&P 500 and GOLD Turn on notifications so you don't miss future updates

NoName

41,004 просмотров • 1 месяц назад

nvidia's 3B mamba destroyed alibaba's 3B deltanet on the same RTX 3090. only 24 days between releases. same active parameters, same VRAM tier, completely different architectures. nemotron cascade 2: 187 tok/s. flat from 4K to 625K context. zero speed loss. flags: -ngl 99 -np 1. that's it. no context flags, no KV cache tricks. auto-allocates 625K. qwen 3.5 35B-A3B: 112 tok/s. flat from 4K to 262K context. zero speed loss. flags: -ngl 99 -np 1 -c 262144 --cache-type-k q8_0 --cache-type-v q8_0. needed KV cache quantization to fit 262K. both models held a flat line across every context level. both architectures are context-independent. but nvidia's mamba2 is 67% faster at generating tokens on the exact same hardware and needs fewer flags to get there. same node, same GPU, same everything. the only variable is the model. gold medal math olympiad winner running at 187 tokens per second on single RTX 3090 a card from 6 years ago. nvidia cooked.

nvidia's 3B mamba destroyed alibaba's 3B deltanet on the same RTX 3090. only 24 days between releases. same active parameters, same VRAM tier, completely different architectures. nemotron cascade 2: 187 tok/s. flat from 4K to 625K context. zero speed loss. flags: -ngl 99 -np 1. that's it. no context flags, no KV cache tricks. auto-allocates 625K. qwen 3.5 35B-A3B: 112 tok/s. flat from 4K to 262K context. zero speed loss. flags: -ngl 99 -np 1 -c 262144 --cache-type-k q8_0 --cache-type-v q8_0. needed KV cache quantization to fit 262K. both models held a flat line across every context level. both architectures are context-independent. but nvidia's mamba2 is 67% faster at generating tokens on the exact same hardware and needs fewer flags to get there. same node, same GPU, same everything. the only variable is the model. gold medal math olympiad winner running at 187 tokens per second on single RTX 3090 a card from 6 years ago. nvidia cooked.

Sudo su

187,008 просмотров • 5 месяцев назад

Y’all I LOVE a good find! I had the bestttt time going to a random estate sale and look at my goodies!!! Vintage Hurricane Lamp - Christmas by Carlton c. 1985 (I’m a Christmas baby! I can’t help myself!) Vintage German ‘Oktoberfest’ Bier Stein c. 1998 (Dana is brewing his own beer and *naturally* we needed to bring him this!) Vintage Viceroy Super 8 Zoom Lens Movie Camera (I have a handful of vintage cameras that I like to collect.) Original pair of Bassett Traditional/Colonial Revival Parlor chairs c. 1980-early 90’s (I’m over modern chairs that aren’t comfortable! Needed something that felt cozy and comfy! I actually love the print!) Hartmann Luggage (USA) Vintage leather garment bag, c. 1968-1974 ✨MINT CONDITION✨(Who doesn’t love a good bag??) Vintage Timex Anniversary Desk Clock c. 1980’s-1990’s (This is for my little library I’m setting up!) Vintage 1978 vinyl Steve Martin - ‘A Wild and Crazy Guy’ (my grandpa LOVED Steve Martin and is the reason I quoted the movie ‘The Three Amigos’ constantly in the 5th grade.) Vintage 1960’s vinyl Tom Jones - I (Who Have Nothing) (I sang this song on Idol so this is a HUGE find for me!) Vintage Strawberry Shortcake and Custard Electronic Quartz Clock c. 1981 (My mom LOVES Strawberry Shortcake so I got this for her. 🥰) Vintage Original Rubik’s Cube c. 1980 (I’m a *liiiittle* of a collector of older games and gaming systems. Couldn’t resist the OG!) Gorgeous 1970’s Vintage Indiana Glass - Pastel Pink Footed Oval Center/Fruit Bowl (This color reminds me of my great-grandmothers. 🩷) Pyrex Originals 4-piece bakeware set c. 1985 (I love to cook and bake so to have an original Pyrex set just tickles me!)

Y’all I LOVE a good find! I had the bestttt time going to a random estate sale and look at my goodies!!! Vintage Hurricane Lamp - Christmas by Carlton c. 1985 (I’m a Christmas baby! I can’t help myself!) Vintage German ‘Oktoberfest’ Bier Stein c. 1998 (Dana is brewing his own beer and naturally we needed to bring him this!) Vintage Viceroy Super 8 Zoom Lens Movie Camera (I have a handful of vintage cameras that I like to collect.) Original pair of Bassett Traditional/Colonial Revival Parlor chairs c. 1980-early 90’s (I’m over modern chairs that aren’t comfortable! Needed something that felt cozy and comfy! I actually love the print!) Hartmann Luggage (USA) Vintage leather garment bag, c. 1968-1974 ✨MINT CONDITION✨(Who doesn’t love a good bag??) Vintage Timex Anniversary Desk Clock c. 1980’s-1990’s (This is for my little library I’m setting up!) Vintage 1978 vinyl Steve Martin - ‘A Wild and Crazy Guy’ (my grandpa LOVED Steve Martin and is the reason I quoted the movie ‘The Three Amigos’ constantly in the 5th grade.) Vintage 1960’s vinyl Tom Jones - I (Who Have Nothing) (I sang this song on Idol so this is a HUGE find for me!) Vintage Strawberry Shortcake and Custard Electronic Quartz Clock c. 1981 (My mom LOVES Strawberry Shortcake so I got this for her. 🥰) Vintage Original Rubik’s Cube c. 1980 (I’m a liiiittle of a collector of older games and gaming systems. Couldn’t resist the OG!) Gorgeous 1970’s Vintage Indiana Glass - Pastel Pink Footed Oval Center/Fruit Bowl (This color reminds me of my great-grandmothers. 🩷) Pyrex Originals 4-piece bakeware set c. 1985 (I love to cook and bake so to have an original Pyrex set just tickles me!)

Jordin Sparks

11,079 просмотров • 7 месяцев назад

🚨 THE BIGGEST IPOS IN HISTORY ARE ABOUT TO DRAIN THE MARKET Three of the largest IPOs in history. One five-month window. Well over $100 billion in fresh stock about to be sold into the market. Every dollar that buys it has to come from somewhere. A lot of it comes from your portfolio. Here's what's lining up: → June 12 - SpaceX went public. $1.75T valuation, $75B raise. The largest IPO ever. (It's already ripped past $2.8T.) → Q4 - OpenAI is expected to list near a $1 trillion valuation. → October - Anthropic is lining up its own debut near a ~$965B valuation. This isn't three separate IPOs. It's a wall of new supply hitting all at once. The mechanism nobody talks about: Wall Street doesn't conjure fresh cash to fund IPOs. Funds sell what they already own to buy the new name. Picture the three biggest stores in history all opening on the same street, the same season - and everyone has to empty their wallet to shop. So what gets sold first? The thing every fund is overweight: the S&P 500. We've seen this movie. 2021: record IPO and SPAC supply, peak euphoria. 2022: worst year for stocks since 2008. And the setup is identical. The S&P just printed an all-time high near 7,620, up ~10% on the year. Everyone all-in. Nobody hedged. That's not strength - that's a setup. Connect the dots: record highs, record euphoria, record new supply - all at the same time. My target: S&P back to 4,100. A ~46% unwind - right back to where this entire AI melt-up began. You're early, or you're exit liquidity. There's no third option. I called the recent S&P sell-offs, the 2025 BTC top, and the move down from $126K to $60K. Missed those? Don't sweat it. Turn on notifications. The next calls are already cooking.

🚨 THE BIGGEST IPOS IN HISTORY ARE ABOUT TO DRAIN THE MARKET Three of the largest IPOs in history. One five-month window. Well over $100 billion in fresh stock about to be sold into the market. Every dollar that buys it has to come from somewhere. A lot of it comes from your portfolio. Here's what's lining up: → June 12 - SpaceX went public. $1.75T valuation, $75B raise. The largest IPO ever. (It's already ripped past $2.8T.) → Q4 - OpenAI is expected to list near a $1 trillion valuation. → October - Anthropic is lining up its own debut near a ~$965B valuation. This isn't three separate IPOs. It's a wall of new supply hitting all at once. The mechanism nobody talks about: Wall Street doesn't conjure fresh cash to fund IPOs. Funds sell what they already own to buy the new name. Picture the three biggest stores in history all opening on the same street, the same season - and everyone has to empty their wallet to shop. So what gets sold first? The thing every fund is overweight: the S&P 500. We've seen this movie. 2021: record IPO and SPAC supply, peak euphoria. 2022: worst year for stocks since 2008. And the setup is identical. The S&P just printed an all-time high near 7,620, up ~10% on the year. Everyone all-in. Nobody hedged. That's not strength - that's a setup. Connect the dots: record highs, record euphoria, record new supply - all at the same time. My target: S&P back to 4,100. A ~46% unwind - right back to where this entire AI melt-up began. You're early, or you're exit liquidity. There's no third option. I called the recent S&P sell-offs, the 2025 BTC top, and the move down from $126K to $60K. Missed those? Don't sweat it. Turn on notifications. The next calls are already cooking.

Shelpid.WI3M

43,105 просмотров • 2 месяцев назад

🚨 $SPX BREAKDOWN HAS STARTED S&P 500 just got rejected in broadening wedge This textbook pattern expands outward - each high is higher, each low is lower Every rejection off the upper line led to a DEEPER low than the one before Point 2 rejected -> deeper low at point 3 Point 4 rejected -> deeper low at point 5 Point 6 just tagged the upper line and turned straight down The pattern has one more move left Point 7 - lower trendline - that's where this resolves First target: ~$6,200 In a market already sitting on record short interest and Wyckoff distribution signals, this wedge just gave the final directional confirmation Follow + notifs on, I will keep you updated on this one

🚨 $SPX BREAKDOWN HAS STARTED S&P 500 just got rejected in broadening wedge This textbook pattern expands outward - each high is higher, each low is lower Every rejection off the upper line led to a DEEPER low than the one before Point 2 rejected -> deeper low at point 3 Point 4 rejected -> deeper low at point 5 Point 6 just tagged the upper line and turned straight down The pattern has one more move left Point 7 - lower trendline - that's where this resolves First target: ~$6,200 In a market already sitting on record short interest and Wyckoff distribution signals, this wedge just gave the final directional confirmation Follow + notifs on, I will keep you updated on this one

Linton Worm (🍏,🪱)

55,556 просмотров • 2 месяцев назад

🚨The launch of the day!! Xiaomi just unveiled the Xring O100 (玄戒O100) - first dedicated high-bandwidth AI accelerator for edge large models - Vertically stacks a 6nm NPU compute wafer with dual high-speed DRAM memory wafers (Wafer-on-Wafer 3D stacking). - Uses advanced Hybrid Bonding instead of conventional micro-bumps. •- Bonding pitch shrinks from ~50 μm all the way to 1.4 μm. - Creates 2.58 million high-speed physical pathways (TSV diameter just 0.7 μm). - Face-to-Face metal-layer direct connections open ultra-short “data wells” straight into the compute cores. - Paired with Xiaomi’s self-designed high-bandwidth matrix bus (HB-Matrix) and 14-core NPU, data flows between the NPUs and memory at extreme speed with dynamic topology switching (ring for prefill, broadcast for decode). - 1.22 TB/s near-memory bandwidth — 16× a flagship phone’s LPDDR5X - Up to 330 tokens/s edge inference (3B model in Xiaomi lab tests) - Designed for on-device LLMs for following Applications: - Smartphones - PCs - Robots - Xiaomi cars / intelligent driving systems - AI Cube prototype (paired with Xring O3 + D100), for models up to 120B parameters locally at up to 150W sustained 🗓️ The chip has completed silicon validation and is scheduled for commercial use in 2027. #Xiaomi #XRing #DRAM

🚨The launch of the day!! Xiaomi just unveiled the Xring O100 (玄戒O100) - first dedicated high-bandwidth AI accelerator for edge large models - Vertically stacks a 6nm NPU compute wafer with dual high-speed DRAM memory wafers (Wafer-on-Wafer 3D stacking). - Uses advanced Hybrid Bonding instead of conventional micro-bumps. •- Bonding pitch shrinks from ~50 μm all the way to 1.4 μm. - Creates 2.58 million high-speed physical pathways (TSV diameter just 0.7 μm). - Face-to-Face metal-layer direct connections open ultra-short “data wells” straight into the compute cores. - Paired with Xiaomi’s self-designed high-bandwidth matrix bus (HB-Matrix) and 14-core NPU, data flows between the NPUs and memory at extreme speed with dynamic topology switching (ring for prefill, broadcast for decode). - 1.22 TB/s near-memory bandwidth — 16× a flagship phone’s LPDDR5X - Up to 330 tokens/s edge inference (3B model in Xiaomi lab tests) - Designed for on-device LLMs for following Applications: - Smartphones - PCs - Robots - Xiaomi cars / intelligent driving systems - AI Cube prototype (paired with Xring O3 + D100), for models up to 120B parameters locally at up to 150W sustained 🗓️ The chip has completed silicon validation and is scheduled for commercial use in 2027. #Xiaomi #XRing #DRAM

Neil Shah

67,195 просмотров • 3 дней назад

I'm not as optimistic that jawboning the Fed is going to get rates reduced anytime soon. I mean this is an unusual situation obviously, but I don't think you can count on rate cuts while Powell's there. This tariff situation is getting really interesting. Tariffs themselves on commodities or goods that you don't create yourself is inflationary. There's no way around it. Get rid of those tariffs so they can bring down those increased costs on that affordability issue. The reason the S&P maintains new highs, in my opinion, is the productivity factor brought in by AI tools that have just started to be used in the last 24 months is really effective. We're starting to see how good this can be across all 11 sectors. Here's our problem and why I have a little caution to these statements. We have no power. The Chinese have built 500 gigawatts in the last 24 months. We have built zero. We have no power on the grid. This is a big problem.

I'm not as optimistic that jawboning the Fed is going to get rates reduced anytime soon. I mean this is an unusual situation obviously, but I don't think you can count on rate cuts while Powell's there. This tariff situation is getting really interesting. Tariffs themselves on commodities or goods that you don't create yourself is inflationary. There's no way around it. Get rid of those tariffs so they can bring down those increased costs on that affordability issue. The reason the S&P maintains new highs, in my opinion, is the productivity factor brought in by AI tools that have just started to be used in the last 24 months is really effective. We're starting to see how good this can be across all 11 sectors. Here's our problem and why I have a little caution to these statements. We have no power. The Chinese have built 500 gigawatts in the last 24 months. We have built zero. We have no power on the grid. This is a big problem.

Kevin O'Leary aka Mr. Wonderful

46,788 просмотров • 7 месяцев назад

Many people still don't grasp just how quickly India is moving 🇮🇳 The new Delhi-Mumbai Industrial Corridor (DMIC) will be completed in the next few years. The $100 billion project started in 2007 and has seen significant Japanese investment since the start. Stretching approximately 1,500 kilometers between the capital city Delhi (metro pop. of 34M) and India's financial capital Mumbai (metro pop. of 21M, this initiative will enhance connectivity and stimulate economic growth by developing smart cities, industrial parks along a new Delhi–Mumbai Expressway and the new Western Dedicated Freight Corridor (railway). The 1350 km 8-lane wide (expandable to 12-lane) expressway connecting the two mega-cities will be completed next year and will reduce the travel time from 24h to 11h. The 1500 km Western Dedicated Freight Corridor (WDFC) will also be completed in 2025. The high-speed line will support freight trains reaching 1,500 m (4,900 ft) length pulled by electric locomotives running at speeds greater than 100 km/h (62 mph). The tracks will be entirely grade-separated and have a generous loading gauge of 3,660 mm width and 7,100 mm maximum height allowing for double-stacked shipping containers to be transported on flatcars. This allows for single trains to have a 400+-container capacity. The corridor will reduce the travel time for containers from 50 h, by the existing freight train, to 17 h while significantly increasing the transport capacity on the route as well as decreasing transport costs. The DMIC megaproject also incorporates 9 Mega Industrial zones of about 200-250 sq. km, 3 ports, 6 airports and a 4000 MW power plant. India is rapidly outpacing China in both demographic and economic growth. The 21st century won't be the Chinese century. It will be the Indian century.

Many people still don't grasp just how quickly India is moving 🇮🇳 The new Delhi-Mumbai Industrial Corridor (DMIC) will be completed in the next few years. The $100 billion project started in 2007 and has seen significant Japanese investment since the start. Stretching approximately 1,500 kilometers between the capital city Delhi (metro pop. of 34M) and India's financial capital Mumbai (metro pop. of 21M, this initiative will enhance connectivity and stimulate economic growth by developing smart cities, industrial parks along a new Delhi–Mumbai Expressway and the new Western Dedicated Freight Corridor (railway). The 1350 km 8-lane wide (expandable to 12-lane) expressway connecting the two mega-cities will be completed next year and will reduce the travel time from 24h to 11h. The 1500 km Western Dedicated Freight Corridor (WDFC) will also be completed in 2025. The high-speed line will support freight trains reaching 1,500 m (4,900 ft) length pulled by electric locomotives running at speeds greater than 100 km/h (62 mph). The tracks will be entirely grade-separated and have a generous loading gauge of 3,660 mm width and 7,100 mm maximum height allowing for double-stacked shipping containers to be transported on flatcars. This allows for single trains to have a 400+-container capacity. The corridor will reduce the travel time for containers from 50 h, by the existing freight train, to 17 h while significantly increasing the transport capacity on the route as well as decreasing transport costs. The DMIC megaproject also incorporates 9 Mega Industrial zones of about 200-250 sq. km, 3 ports, 6 airports and a 4000 MW power plant. India is rapidly outpacing China in both demographic and economic growth. The 21st century won't be the Chinese century. It will be the Indian century.

Visegrád 24

207,110 просмотров • 1 год назад

Ocarina of Time, running natively on iOS and iPadOS for the first time. Not an emulator. Ship of Harkinian has run on Windows, Linux, macOS, Switch, Wii U, and Android. Never iOS. Now it runs on iPad, with iPhone in the build too. The game was decompiled to C and rebuilt as a native ARM64 app rendering through Metal. No N64 hardware being simulated, no ROM interpreted at runtime, no streaming from a Mac. It isn't stuck at the N64's 240p and 20fps. Widescreen, full resolution, 60fps. This game deserves a controller, so controller support is native. Pair a Bluetooth controller and play the whole thing. Touch, keyboard, and mouse work too. I used Codex 5.6 Sol to build the iOS layer. Open sourcing it all below.

Ocarina of Time, running natively on iOS and iPadOS for the first time. Not an emulator. Ship of Harkinian has run on Windows, Linux, macOS, Switch, Wii U, and Android. Never iOS. Now it runs on iPad, with iPhone in the build too. The game was decompiled to C and rebuilt as a native ARM64 app rendering through Metal. No N64 hardware being simulated, no ROM interpreted at runtime, no streaming from a Mac. It isn't stuck at the N64's 240p and 20fps. Widescreen, full resolution, 60fps. This game deserves a controller, so controller support is native. Pair a Bluetooth controller and play the whole thing. Touch, keyboard, and mouse work too. I used Codex 5.6 Sol to build the iOS layer. Open sourcing it all below.

Kahris

1,172,171 просмотров • 1 месяц назад

$🚨 STOP AND LOOK AT THIS CHART 🚨 The S&P 500 has run the same cycle for 100 YEARS 25 years bull. 9-12 years bear. Every single time And everyone is pointing to this chart saying we're still in the bull I've heard this theory many times. I understand the logic I just don't think it works this time Every previous bull was driven by a technology that took decades to price in AI took 2 years And what do we have right now? - Valuations at 100-year extremes - Smart money exiting - Distribution signals everywhere - Insiders selling - Buffett sitting on record cash - Burry calling this 1999 The 25-year cycle assumes the catalyst lasts 25 years This one already burned out I called the BTC top at 115k Called the S&P distribution before the drop Called the Nvidia fractal before anyone was talking about it Three cycles of experience. Every major move predicted in advance If you're not following yet - that's on you FOLLOW + NOTIFS ON!$

🚨 STOP AND LOOK AT THIS CHART 🚨 The S&P 500 has run the same cycle for 100 YEARS 25 years bull. 9-12 years bear. Every single time And everyone is pointing to this chart saying we're still in the bull I've heard this theory many times. I understand the logic I just don't think it works this time Every previous bull was driven by a technology that took decades to price in AI took 2 years And what do we have right now? - Valuations at 100-year extremes - Smart money exiting - Distribution signals everywhere - Insiders selling - Buffett sitting on record cash - Burry calling this 1999 The 25-year cycle assumes the catalyst lasts 25 years This one already burned out I called the BTC top at 115k Called the S&P distribution before the drop Called the Nvidia fractal before anyone was talking about it Three cycles of experience. Every major move predicted in advance If you're not following yet - that's on you FOLLOW + NOTIFS ON!

NoName

1,052,400 просмотров • 3 месяцев назад

🚨 SOMETHING VERY STRANGE IS HAPPENING! The stock market keeps pushing to new all-time highs. But nobody is paying attention to what’s actually happening. Short interest is hitting record levels: 3.7% on the S&P 500. 6.1% on the Russell 3000. 9.0% on the NYSE-listed stocks. The last time shorts were this high was right before the 2008 crash. The trap doesn't change: Make the crowd believe the rally will last forever. Use maximum FOMO to keep retail buying at the top. Build massive short positions behind closed doors. The smart money isn't hedging. They are preparing for a bloodbath. When the trap snaps, the crowd will be left holding the bag. Again. Don't buy the euphoria. Watch what the smart money is doing, not what they say. For the record, I’ve called every major market top over the last decade, including the $111K BTC peak. If you missed those, don’t FOMO. I’ll call out the exact breaking point right here publicly, like I always do. Don't miss this one.

🚨 SOMETHING VERY STRANGE IS HAPPENING! The stock market keeps pushing to new all-time highs. But nobody is paying attention to what’s actually happening. Short interest is hitting record levels: 3.7% on the S&P 500. 6.1% on the Russell 3000. 9.0% on the NYSE-listed stocks. The last time shorts were this high was right before the 2008 crash. The trap doesn't change: Make the crowd believe the rally will last forever. Use maximum FOMO to keep retail buying at the top. Build massive short positions behind closed doors. The smart money isn't hedging. They are preparing for a bloodbath. When the trap snaps, the crowd will be left holding the bag. Again. Don't buy the euphoria. Watch what the smart money is doing, not what they say. For the record, I’ve called every major market top over the last decade, including the $111K BTC peak. If you missed those, don’t FOMO. I’ll call out the exact breaking point right here publicly, like I always do. Don't miss this one.

MARMOT

30,825 просмотров • 1 месяц назад

A single RTX 4090 (24 GB VRAM) can run the updated gemma 4 31B (dense) model with a 190,000 context window at 33 tokens/second. The VRAM barrier is dying. Google quietly updated Gemma 4, and Unsloth immediately compiled the new quants. I built llama.cpp from source on Ubuntu 22 to benchmark it. Google's stealth update 2 days ago enabled uniform Flash Attention 4 on Hopper to boost prefill and patched the chat template to improve tool calling. The agentic reasoning gains on the benchmark charts are massive: TB2 (Agents): +4.5% (to 25.8%) Tau2 (Telecom): +10.1% (to 62.7%) Running on Ubuntu 22, CUDA 13.0 with a single NVIDIA GeForce RTX 4090. Here is the exact step by step benchmarking process with a massive 28k tokens prompt and the commands I used to squeeze out maximum context without killing my throughput: # 1. The Baseline (Unquantized KV Cache) I started with full GPU offload (-ngl 99) and pushed the context to 40k. llama.cpp flags: ./build/bin/llama-server -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -ngl 99 -c 40000 -fa on --port 8080 -v VRAM: 23.8 GB (maxed out on card) Throughput: Prefill: 2198.81 t/s | Decode: 35.77 t/s (with 28k tokens prompt) # 2. The CPU Split Trap I tried stretching to 80k context by offloading layers to the CPU (-ngl 52). llama.cpp flags: ./build/bin/llama-server -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -c 80000 -ngl 52 -fa on --port 8080 -v Throughput: Prefill: 1212.73 t/s | Decode: 5 t/s (with 28k tokens prompt) # 3. The KV Quantization Breakthrough Instead of spilling layers to the CPU, I kept the model fully on card (-ngl 99) but enabled 8-bit KV cache quantization to free up VRAM. flags: ./build/bin/llama-server -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -c 100000 --cache-type-k q8_0 --cache-type-v q8_0 -ngl 99 --port 8080 -v VRAM: 23.9 GB Throughput: Prefill: 2139.68 t/s | Decode: 32 t/s (with 28k tokens prompt) Result: 100k tokens of context on a single GPU with practically zero speed loss (and minimal intelligence loss). # 4. The Limit Test (Q4 KV Cache) To find the absolute breaking point, I dropped the KV cache to 4 bit (q4_0) and set -c 190000. flags: ./build/bin/llama-server -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -c 190000 --cache-type-k q4_0 --cache-type-v q4_0 -ngl 99 --port 8080 -v VRAM: 23.8 GB Throughput: Prefill: 2206.66 t/s | Decode: 33 t/s (with 28k tokens prompt) (Note: Pushing it to 220k required dropping to -ngl 58 again, which immediately penalized decode down to 17 t/s). # The Tradeoff: For Max Reasoning: Keep your KV cache unquantized (f16). You get pristine reasoning but hit a strict 40k context ceiling. For Massive Document Retrieval: If you need to feed the model giant codebases, use --cache-type-k q4_0. Getting 190k context at 33 tokens/second on a consumer desktop with a 31b dense model is a cheat code. If you’re rocking a single 3090 or 4090 and slept on Gemma 4 earlier, this update is your cue to dust off the terminal. Hugging Face links to the Unsloth QAT quants are in the replies below.

A single RTX 4090 (24 GB VRAM) can run the updated gemma 4 31B (dense) model with a 190,000 context window at 33 tokens/second. The VRAM barrier is dying. Google quietly updated Gemma 4, and Unsloth immediately compiled the new quants. I built llama.cpp from source on Ubuntu 22 to benchmark it. Google's stealth update 2 days ago enabled uniform Flash Attention 4 on Hopper to boost prefill and patched the chat template to improve tool calling. The agentic reasoning gains on the benchmark charts are massive: TB2 (Agents): +4.5% (to 25.8%) Tau2 (Telecom): +10.1% (to 62.7%) Running on Ubuntu 22, CUDA 13.0 with a single NVIDIA GeForce RTX 4090. Here is the exact step by step benchmarking process with a massive 28k tokens prompt and the commands I used to squeeze out maximum context without killing my throughput: # 1. The Baseline (Unquantized KV Cache) I started with full GPU offload (-ngl 99) and pushed the context to 40k. llama.cpp flags: ./build/bin/llama-server -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -ngl 99 -c 40000 -fa on --port 8080 -v VRAM: 23.8 GB (maxed out on card) Throughput: Prefill: 2198.81 t/s | Decode: 35.77 t/s (with 28k tokens prompt) # 2. The CPU Split Trap I tried stretching to 80k context by offloading layers to the CPU (-ngl 52). llama.cpp flags: ./build/bin/llama-server -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -c 80000 -ngl 52 -fa on --port 8080 -v Throughput: Prefill: 1212.73 t/s | Decode: 5 t/s (with 28k tokens prompt) # 3. The KV Quantization Breakthrough Instead of spilling layers to the CPU, I kept the model fully on card (-ngl 99) but enabled 8-bit KV cache quantization to free up VRAM. flags: ./build/bin/llama-server -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -c 100000 --cache-type-k q8_0 --cache-type-v q8_0 -ngl 99 --port 8080 -v VRAM: 23.9 GB Throughput: Prefill: 2139.68 t/s | Decode: 32 t/s (with 28k tokens prompt) Result: 100k tokens of context on a single GPU with practically zero speed loss (and minimal intelligence loss). # 4. The Limit Test (Q4 KV Cache) To find the absolute breaking point, I dropped the KV cache to 4 bit (q4_0) and set -c 190000. flags: ./build/bin/llama-server -m gemma-4-31B-it-qat-UD-Q4_K_XL.gguf -c 190000 --cache-type-k q4_0 --cache-type-v q4_0 -ngl 99 --port 8080 -v VRAM: 23.8 GB Throughput: Prefill: 2206.66 t/s | Decode: 33 t/s (with 28k tokens prompt) (Note: Pushing it to 220k required dropping to -ngl 58 again, which immediately penalized decode down to 17 t/s). # The Tradeoff: For Max Reasoning: Keep your KV cache unquantized (f16). You get pristine reasoning but hit a strict 40k context ceiling. For Massive Document Retrieval: If you need to feed the model giant codebases, use --cache-type-k q4_0. Getting 190k context at 33 tokens/second on a consumer desktop with a 31b dense model is a cheat code. If you’re rocking a single 3090 or 4090 and slept on Gemma 4 earlier, this update is your cue to dust off the terminal. Hugging Face links to the Unsloth QAT quants are in the replies below.

Alok

76,069 просмотров • 1 месяц назад