Quant2be 🦀

@jorandirkgreef Both correctness and performance are fundamentally architectural in nature. You can change many things about a system but architecture is forever.

Cursor Composer 2.5's is 3–18x cheaper than Opus 4.7 in Claude Code (medium reasoning), and 5–32x cheaper than GPT-5.5 in Codex (medium) based on API pricing This low Cost per Task isn't just driven by relatively low token pricing, it's also driven by low relatively low token usage compared to other leading models. @cursor_ai Composer 2.5 only used 1.6M token to complete our Coding Agent Index benchmarks, while other models used up to 5.7M. This lower token usage also contributes to a low Time per Task. Across the Coding Agent Index configurations shown, average Time per Task was ~12 minutes. Composer 2.5 completed tasks in ~9 minutes on average, making it ~1.3x faster than average, while Composer 2.5 Fast completed tasks in ~7 minutes, making it ~1.8x faster than the average across agents. Link to full benchmark results below

On Nvidia’s Vera CPU – First, let’s revisit the CPU architecture (see attached image) and its bundled memory The referenced SOCAMM is a data centre class modular form factor for LPDDR5X (not to be conflated as two different things) Now, 1 LPDDR5X DRAM die = 9.6Gbps/bit (max). Assuming a 32-bit package, 1 LPDDR5X DRAM package = 9.6 x 32 = 307.2Gbps (~38GB/s) bandwidth 1 SOCAMM is constructed using four LPDDR5X DRAM packages. Total bandwidth per SOCAMM = 38 x 4 = ~154GB/s The Vera CPU setup uses 8 SOCAMM (8-channel), and therefore has 8 x 154 = ~1.2TB/s of bandwidth The 1.5TB refers to the capacity. Assuming 32 (8 x 4) LPDDR5X DRAM packages in 1 Vera CPU, this infers the use of 48GB DRAM packages (not 192GB), i.e., 32 x 48. This could be for thermal/power management reasons In short, each Vera CPU set up = 8 SOCAMMs = 32 LPDDR5X DRAM packages = 1.5TB capacity (32 x 48GB) = up to 1.2TB/s of bandwidth Second, on market opportunity Nvidia guided visibility into ~$20B of CPU revenue this year. What is unknown is how this breaks down into sales configurations (and by extension, the memory modules/density types that may be required) Possible configurations: 1 – as part of Vera Rubin. Assuming NVL72 setup (72 Rubin GPUs, 36 Vera CPUs), total LPDDR5X memory capacity = ~55TB (36 x 8 x 4 x 48), which could potentially be higher if they use higher-capacity memory packages (with liquid cooling) 2 – dedicated Vera CPU racks. Each rack packs 256 Vera CPUs, and therefore up to ~400TB of LPDDR5X memory (256 x 8 x 4 x 48), which could potentially be higher if they use higher-capacity memory packages (with liquid cooling) 3 – individual Vera CPU servers. On the low end will be single and dual socket servers, with one or two Vera CPUs per tray/node respectively. For this configuration, memory capacity = 1.5TB / 3.0TB (assuming use of 48GB modules, with air cooling). Whereas on the high-end, such as the ones being developed by HPE (Cray Supercomputing GX240), this could go up to 640 Vera CPUs = close to ~1000TB. There is also the HGX Rubin NVL8 configuration, with one or two CPUs wired up (with similar memory capacities as a the single/dual socket servers) In short, it’s challenging to accurately size up the total bit demand for LPDDR5X without knowing the exact sales configurations What may be a better play is memory PHY/controllers IP – Each SOCAMM requires 4 x LPDDR5X PHY and memory controllers. For Vera CPU which uses 8 SOCAMMs, each CPU = 32 LPDDR5X PHYs and memory controllers. All that is needed is the average list/sale price for CPU, divide $20B by this number to derive the implied number of Vera CPUs to be sold, that then allows you to derive the dollar content for memory PHY/controllers

This is a unit circle (r = 128px) drawn using my own math library in C. It uses no division, no modulo, and no floating points. It uses a binary angle lookup table for SIN and COS, totaling 256 Bytes plus the few bytes for a couple instructions (accessing it, a bitwise &, and couple shifts.) It scales well by either doubling or halving the table size repeatedly, and if I need more accuracy in a pinch, I can linearly interpolate 2 angles. It also has 512 angles instead of 360. This shines on older hardware, like the PS1, where it can fit in the cache (aka the scratchpad), but its footprint is small enough to be negligible on modern hardware while giving me the added bonus of keeping my compiler simpler, more lightweight, and free of dependencies and potential bloat. If this excites you, congrats. There's not many of us it feels like.

This is why RedHat makes like over $6 billion a year by the way. For anyone wondering what the fuck RedHat does, its basically: take the free chips on the side of the road, validate them, and sell the supply chain contract.

@androolloyd @HyperliquidX Try out @AmpCode too? x.com/i/status/20522…

will be cool to watch the $SPCX price action during the flight today only on @HyperliquidX

Fork your dependencies, trim them to only your use case, never update unless it breaks for your users. I’ve been vocal about this for 10+ years. I’ve always said that updating is way riskier than latent bugs (which can be tracked and CVEs monitored). If you are updating a dependency, it’s on you to analyze every single commit in the full transitive set of dependencies. If you dont see anything compelling, dont update! I remember at HashiCorp once in awhile an engineer would try to update a dep or replace a DIY lib with an external one and id always ask “show me the commit we need.” Dont update for the sake of it. Feeling pretty swell about this mentality with all the supply chain attacks happening.

I've got a scary math joke but i'm 2² to say it

The @tursodatabase private beta is going very well. Turso is already faster than SQLite in some operations, and concurrent writes is the major unlock: Imagine SQLite, just without the concurrency issues. Locally, on the Cloud, and anywhere in between.

As someone who has had multiple burn outs, to a terrifying degree, let me tell you If you’re neurodivergent, rest won’t do it You need to refuel with your special interests You need novelty and challenge Sitting in silence and simply “resting” might only torture you, shem.

@mahal0z @34r7hm4n @androolloyd

For years, Rust binaries made reversing a nightmare. Modern decompilers only support C, lacking meaningful types, constructs, and language-specific functions. Led by @34r7hm4n, we're releasing our S&P work Oxidizer, the first deep Rust decompiler, built on angr! Interested? 🧵👇

a general-purpose model solved a major open problem in mathematics. we'll be saying this a lot over the coming years, but this is a kinda big milestone. i'm very excited for AI to greatly extend our understanding of the world, but still, i have complicated feelings today.

@Merridew__ "Only when the tide goes out do you discover who's been swimming naked."

Soon, the concept of a "stock exchange" will not need to exist. The stocks will not be stocks and the exchanges will not be exchanges. Equity in companies will simply be cryptographic claims recorded in smart contract storage on public blockchains The exchanges will be smart contracts against which users swap their equity. The average person is missing the revolution that is taking place. You will hold tokens and interact with a predetermined, immutable set of logic inside of smart contracts. No humans. No broker. No "equity".

Intelligence is a beautiful thing, I will not lie. There is something deeply attractive about a mind that can see patterns others miss, connect distant ideas, and explain the world with startling clarity. But what I have found with many intelligent people is a curious tendency to confuse understanding with action. They become endlessly patient with circumstances while doing absolutely nothing to change them. And so their intelligence, rather than serving as a catalyst for movement, becomes a sophisticated defense mechanism. They can explain why now is not the right time. They can construct elaborate arguments for waiting a little longer. They can intellectualize their inertia so convincingly that even they begin to mistake hesitation for wisdom. But intelligence without ruthless execution is, in many ways, just cosplay. It is the performance of competence without the burden of consequence. The intelligent man continues to defend his position because intellectualizing about an issue is simple; the real work lies in summoning the courage to act. And that, to me, is the greatest limitation of intelligence on its own.

1/ We are sharing additional details regarding our investigation into unauthorized access to GitHub's internal repositories. Yesterday we detected and contained a compromise of an employee device involving a poisoned VS Code extension. We removed the malicious extension version, isolated the endpoint, and began incident response immediately.

We are investigating unauthorized access to GitHub’s internal repositories. While we currently have no evidence of impact to customer information stored outside of GitHub’s internal repositories (such as our customers’ enterprises, organizations, and repositories), we are closely monitoring our infrastructure for follow-on activity.

Quando você não tem pendrive ou rede e precisa transferir os arquivos de qualquer modo 🤯🤯🤯

Most walls you hit are really just staircases with a few missing steps.

Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.