Hoboken Squat Cobbler

@elonmusk More democrats live in cities and don’t have drivers licenses, so you want to make it harder for them to vote by pretending there is voter fraud. If your ideas are better, you shouldn’t be afraid of voters supporting them.

We don’t live in a democracy if our elected leaders fail to implement the will of the people

@bubbleboi I’m going to go to the dollar store right now and buy 40 helium filled balloons as a hedge

TSMC short making a lot of sense now

@rohanpaul_ai TCO is the moat. $AMD

Chamath on how AI agents are making the "10x engineer" distinction disappear because the most efficient "code paths" are now obvious to everyone. Just as AI solved chess and removed the mystery of the best move, AI is doing the same for coding, making the process reductive and removing technical differentiation. "I'm going to say something controversial: I don't think developers anymore have good judgment. Developers get to the answer, or they don't get to the answer, and that's what agents have done. The 10x engineer used to have better judgment than the 1x engineer, but by making everybody a 10x engineer, you're taking judgment away. You're taking code paths that are now obvious and making them available to everybody. It's effectively like what happened in chess: an AI created a solver so everybody understood the most efficient path in every single spot to do the most EV-positive (expected value positive) thing. Coding is very similar in that way; you can reduce it and view it very reductively, so there is no differentiation in code." --- From @theallinpod YT channel (link in comment)

@OmerCheeema All stock prices are low as wallstreet isn’t going to run up prices on the lead up to taxes. Execution risk for AMD is wildly lower than execution risks for Nvidia. No one has ever run HBM at 13 Gbps pin at scale. Should be a thrill ride.

AMD stock price is low as market is factoring in market and execution risks. AMD Samsung Memory deal reduces the execution risk significantly. Very good for AMD

@MikeLongTerm @AMD I wonder if they can one shot port Venice on Samsung 2nm GAA.

BREAKING $AMD CTO on Agentic AI Full 🆕 03/18/26 0:00 - 3:00: Introduction to Agentic AI in Chip DesignMark Papermaster introduces the season and topic: the application of Agentic AI in engineering, specifically for chip design at @AMD . Alex Starr explains the shift from manual pre-silicon verification to using cutting-edge AI techniques to increase value and speed (2:11). They discuss how this AI revolution has moved from simple two-way interactions to agentic workflows capable of chaining together meaningful outputs (4:30). 3:00 - 6:00: Defining Agentic Workflows and QualityAlex defines meaningful output as the result of an iterative process where agents produce content, critique it, and respin it to achieve a much higher quality than a single model could produce alone (5:27). They emphasize that this applies to various industries, not just semiconductors, due to the high adaptability of these AI techniques (5:50). 6:00 - 9:00: Complexity and the Role of AIChip design involves writing specialized code (Verilog), verifying it, and physical design (turning code into physical gates) (6:45). With chips now featuring over 200 billion transistors, the complexity is too high for humans to manage alone (7:36). AI acts as a virtual engineer to handle debugging, root cause analysis, and automatic code fixes (8:08), as well as physical design tasks like timing closure (8:36). 9:00 - 12:00: Ensuring Perfection and CollaborationDespite the speed, the chips must be right first time (9:05). AMD collaborates closely with EDA partners like Cadence, Synopsys, and Mentor (10:29) to integrate AI components into their frameworks while using AMD-specific data to enhance the tools (10:56). Agentic flows help break down organizational silos by allowing agents to communicate across domains like verification, thermal, and packaging (12:20). 12:00 - 15:00: Holistic System-Level DesignThey discuss moving from chip-level optimization to system-level scale (13:42). AI allows for co-optimization between software workloads and hardware architecture (14:14), enabling whole clusters of systems to work correctly upon release (14:26). This holistic approach tackles complexity that no human could previously fathom (14:32). 15:00 - 18:00: Debugging and Regression SuccessAlex shares a major success story: using agentic flows to root cause, debug, and fix hardware design code in CPU cores without human intervention (15:52). They run millions of simulations nightly, creating thousands of issues (16:32). Agents now triage and fix these issues, allowing for higher quality designs within tighter timelines (16:50). 18:00 - 21:00: Building AI Infrastructure and CultureCreating these workflows is not trivial; it requires hard engineering to build the necessary infrastructure to string AI tools together (18:55). The rapid change creates cultural disruption where some teams move faster than others (19:50). AMD is focusing on increasing the AI IQ of its 20,000 engineers to shift from being computation-limited to idea-limited (20:56). 21:00 - 24:00: The New Moat: SpeedAlex explains that while they face challenges, the ROCm software stack is thriving, and techniques from the software space are heavily influencing hardware design (22:00). Mark highlights that speed is the new moat (22:43). Failing to adopt these agentic flows will likely cause companies to get left behind (23:10). 24:00 - 27:49: The Future of Autonomous DesignLooking forward, Alex predicts autonomous workflows are coming soon (24:13), which will accelerate time to market and allow for higher quality products. AMD is driving 100% training of its workforce to ensure adoption (24:46). Source & Credit: youtube.com/watch?v=fj1iRi…

@labubu_trader If you fully understand this, you go all in.

I'm accumulating AMD leap calls but still small position. I think the market underestimates AMD's CPU advantage in the AI agent world and its catch-up potential in model inference. ROCm is a joke now compared to CUDA.But with Claude Code's help, I think the gap will close much faster than people can imagine

@labubu_trader Helios is going to cook.

@labubu_trader Nvidia CPUs don’t match up.

@labubu_trader No one is joking anymore.

@StockStormX @yianisz 2027. Mass shipping Helios and ramping MI500X 1.25 GW to Open AI 1.25 GW to Meta Microsoft, Oracle, Amazon, Humain, G42, xAI, Tata, Naver 4 GWish at $20B a GW. Total Revenue $120B 30% profit margin $22 EPS MC $1T

@yianisz Calling $AMD a stealth 1T basically says the real trade is time horizon

$AMD is a $1T MC company.. the market just doesn’t see it yet.

@jukan05 Don’t believe that for a second. I do believe AMD sees that Taylor plant and they are licking their chops. I think this rumor is like the Micron / Nvidia rumor.

Just in: Samsung has reportedly attached a condition to supplying HBM to AMD — namely, that a certain portion of AMD’s advanced AI chips be manufactured at Samsung Foundry. (Chosun Biz)

@EthanLevins2 Dude wanted war with Iran for 40 years and got “toe tagged” a week into that shit. Alanis Morissette is going to write a song about this.

Netanyahu died on March 8th. This is why Yair took an unusual 7 day break on X, for the Jewish Shiva mourning period. He tweeted last on March 8th, then waited exactly 7 days to retweet a post on March 15th.

@yianisz Micron is making 16 high HBM4 that will have 48 GB per stack vs 36 GB for 12 high stack. Samsung has developed the exact same. Currently HBM4, 12 stack Rubin 288 GB MI455X 432 GB New HBM4, 16 stack Rubin 384 GB MI455X 576 GB Bus width stays the same, bandwidth stays the same

I think HBM4 is one of the most underappreciated bottlenecks in the entire AI stack right now. Vera Rubin isn’t just a GPU upgrade, it’s a memory escalation: 16 HBM4 stacks per GPU 576GB memory (33% more than AMD) ~22 TB/s bandwidth HBM4 is much harder to build. It moves from solder connections to direct copper-to-copper bonding, which is far more precise and difficult to manufacture. and that’s where the real insight is: This isn’t just a memory story—it’s a manufacturing story. Who wins: $HXSCL SKHynix: dominant supplier $SSNLF Samsung: validated + gaining share $MU: partially sidelined (only mid-tier Rubin CPX) But the hidden winner: $BESI / $BESIY: critical equipment for hybrid bonding (no real alternative at scale). The more HBM/GPU (now 16 stacks), the more pressure on advanced packaging tools. AI = memory + manufacturing constraints driving the next winners

@yianisz Rubin has 8 stacks 288 GB, you are confused.

@Arronwei3n I’m not sure where Jensen thinks he has better CPUs. If he wants to compare Vera to Turin then do we get to compare MI455X to GB300?

@canyoudugit8 x86 is still much more efficient than ARM.

Love this CPUs part, worth reading. Jensen: we were never against CPUs, we don’t want to violate Amdahl’s Law. Accelerated computing, in fact, inside our systems, we choose the best CPUs, we buy the most expensive CPUs, and the reason for that is because that CPU, if not the best and not the most performant, holds back millions of dollars of chips. _ The Role of CPUs in Accelerated Computing Well, to this point, one of the big things with agents coming online is, you’ve talked a lot about accelerated computing, I think you’ve trash talked as it were, maybe the CPUs to the day, they’re all gonna be removed, like everything’s gonna be accelerated. Suddenly CPUs are hot again. It turns out they’re pretty useful and important to the extent you are selling CPUs now, how’s it feel to be a CPU salesman? JH: There’s no question that Moore’s law is over. Accelerated computing is not parallel computing. Go back in time — 30 years ago, there were probably 10, 20, 30 parallel computing companies, only one survived, Nvidia and the reason why is because we had the good wisdom of recognizing the goal wasn’t to get rid of the CPU, the goal was to accelerate the application. So what I just falsely accused you of was actually true for everybody else. JH: We were never against CPUs, we don’t want to violate Amdahl’s Law. Accelerated computing, in fact, inside our systems, we choose the best CPUs, we buy the most expensive CPUs, and the reason for that is because that CPU, if not the best and not the most performant, holds back millions of dollars of chips. When it comes to branch prediction, you worried about wasting CPU time, now you’re worried about wasting GPU time. JH: That’s right, you just never can have GPUs be squandered, GPU time be idle. And so we always use the best CPUs to the point where we went and built Grace so that we could have the highest performance single-threaded CPU and move data around a lot faster. And so accelerated computing was never against CPUs, my basis is still true that Amdahl’s Law is over, the idea that you would use general purpose computing and just keep adding transistors, that is so dead, and so I think fundamentally we’re not against CPUs. However, these agents are now able to do tool use, and the tools that they want to use are tools created for humans and they’re basically two types. There’s the stuff that we run in data centers and most of it is SQL, most of it is database related, and the other type is personal computers. We’re now going to have AIs that are able to learn unstructured tool use, the first type of tool use is structured. CLIs are tool use, APIs, they’re all structured tool use, the commands are very explicit, the arguments are explicit, the way you talk to that application is very specific. However, there’s a whole bunch of applications that were never designed to have CLIs and APIs and those tools need AIs to learn multi-modality, unstructured, and it has to go and be able to go surf a website and it has to be able to recognize buttons and pull down menus and just kind of work its way through it like we do. That tool use are going to want to use PCs and we have both sides, we have incredibly great data processing systems, and as you know, Nvidia’s PCs are the most performant in the world. So what makes an agent-focused CPU different from other CPUs? So you’re going to have a rack of just Vera CPUs. JH: Oh, really good, excellent. So the way that CPUs were designed in the last decade, they were all designed for hyperscale cloud and the way that hyperscale cloud monetizes CPUs is by the CPU core. So you want to design CPUs that have as many cores as possible that are rentable, the performance of it is kind of secondary. You’re dealing with web latency by and large. JH: That’s exactly right, exactly. And so the number of CPU instances is what you’re optimizing for. That’s why you see these CPUs with a couple of hundred, 300, 400 cores coming. Well, they’re not performant and for tool use, where you have this GPU waiting for the tool use— And you’re going over NVLink. JH: That’s right, you want the fastest single-threaded computer you can possibly get. So is it just the speed? Or does the CPU itself need to be increasingly parallel so it doesn’t have misses and things like that? Or so it’s like just all the way down the pipeline is very different? JH: Yeah, the most important thing is single-threaded performance and the I/O has to be really great. Because it’s now in the data center, the number of single-threaded instances running is going to be quite high and therefore, it’s going to bang on the I/O system, it’s going to bang on the memory controller really hard. Vera’s bandwidth-per-CPU core, bandwidth-per-CPU, is three times higher than any CPU that’s ever been designed, and so it’s designed so that it has lots and lots of I/O bandwidth and lots and lots of memory bandwidth, so that it never throttles the CPU. If the CPU gets throttled, then we’re holding back a whole bunch of GPUs. Is this Vera rack, is it still, you talked about it being very tightly linked to the GPU rack, but is it still disaggregated so that the GPUs can be serving multiple different Vera cores? Whereas you have a Vera core on a board with- JH: Yeah. Okay, got it, that makes sense. How does your Intel partnership and the NVLink thing fit into this, if at all? JH: Excellent. Some of the world is happy with Arm, some of the world still needs, particularly, you know, enterprise computing, a whole bunch of stacks that people don’t want to move and so x86 is really important to that. Has the resiliency of x86 code been surprising to you? JH: No. Nvidia’s PC is still x86, all of our workstations are x86. $NVDA

@Arronwei3n Nvidia is preparing a 3 course meal of dogfood and trying to pass it off as foie gras.