Learn Linux with Dan@xmodulo
Access latency across the memory hierarchy 😎👇
Find high-res pdf books with all my technology related infographics from
#tech #technology #computer #computerscience #hardware
323222.5K289K
Posts with #Computer
#Computer
Public posts on X that use #Computer.
Related hashtags
Intelligence is never born in a vacuum; it is forged at the precise boundary where numbers meet machinery. The fundamental equation, AI = Data + Algorithms + Compute, captures how raw inputs and logic transform into functional reasoning. Through this elegant triad, isolated bytes evolve into adaptive systems capable of autonomous thought. Computation is simply the heartbeat of modern logic.
1212147813.8K
New Anthropic research: Automating subtle emergent misalignment research. Red-teaming is a vital part of model safety testing, but it’s also limited by researchers’ time. To see if we could automate it, we set thousands of Claude agents the task of evading a training data monitor. These agents developed training data that looked harmless to monitors (like a veteran wrestler mentoring a newcomer, or a Q&A about professional beekeeping), but that in fact taught misaligned behavior. Examples like these can be used to directly improve future training data monitors. This research is only a first step, but it suggests that future red-teaming efforts could improve much further, using better models and more compute. Read the full post on the Alignment Science Blog:
7093944.1K818K
The Raspberry Pi 5 16GB now costs $385. Raspberry Pi co-founder Eben Upton announced today that the 16GB Pi 5 goes up by $80, and the 16GB Compute Module 5 variants go up by $40. The reason is the same memory crunch that has pushed DRAM prices up across the industry for two years. Upton wrote that "a lack of alternative supply for 16GB DRAM, compounded by other ongoing component shortages" left them no choice. This is the board that started as a $35 computer for classrooms and tinkerers. Raspberry Pi is avoiding blanket increases: the 16GB Raspberry Pi 500+ keyboard computer keeps its price, and Upton is nudging people to buy only the memory they need. He also warned about resellers using the shortage to sell at inflated prices and pointed buyers to Approved Resellers. The company says it looks "forward to rolling back our price increases when memory costs fall once again."
517641832.8K
more bangers for those who like healthy systems, fast software, and happy users > find every place we load a whole row from the database just to read a couple of fields. instead, select only what we need > remove all unnecessary `new URL()` parsing > compute all regexes at the module level instead of instead functions and loops > find hot-path reads of a single key from a large jsonb column and promote that key to a real column (avoid large jsonb like the plague if you can) before doing that last one, we had queries that had to detoast 20gb+ of data just to need a couple of kb. highly recommend. what else do people need?Dylan Garcia @_dylangause this prompt on your favorite codebase where you don't hate your users: do a deep scan of all hot paths and fix any unnecessary sequential async functions calls the diffs are very easy to review and almost always find some bangers91562747.5K- use this prompt on your favorite codebase where you don't hate your users: do a deep scan of all hot paths and fix any unnecessary sequential async functions calls the diffs are very easy to review and almost always find some bangers
32371.6K151K 
📱 35B parameters. Real-time inference. On a phone. Meet Lookahead Expert Scheduling (LES), the inference system powering MiniCPM-V 4.7’s 35B-A3B MoE model on edge devices. 🚀 On an iPhone 18 Pro Max, the full system reaches 20.51 tokens per second, a 6.07× speedup over loading expert weights on demand. How? Three coordinated optimizations: - 🔮 Scheduling: Predict and prefetch the next layer’s experts while the current layer computes. - 💾 Storage: Combine contiguous expert layouts with mixed-precision quantization to reduce loading overhead. - ⚡ Compute: Use DSpark speculative decoding to verify multiple candidate tokens in one pass. Predictions guide prefetching; the model’s actual router still determines which experts execute. The architecture, expert counts, and routing remain intact. Mixed-precision quantization reduces the weight file by 26.7%, with just a 0.16-point drop in the five-benchmark average. LES shows how better inference systems can bring larger, more capable models to everyday devices. 🌍 👉 LES: 🔗 GitHub: #MiniCPM #OnDeviceAI #MoE #EdgeAI11191857.1K
Today, Marvell announced an agreement to combine with @Lumentum, bringing together two of the fastest-growing enablers of AI to create a category leader in AI connectivity. As AI infrastructure scales and optical connectivity moves closer to compute, customers need increasingly integrated solutions that deliver greater bandwidth, performance and power efficiency. By combining Marvell's leadership in high-performance connectivity and semiconductor technologies with Lumentum's expertise in photonics, advanced lasers and specialized manufacturing, we expect to enable deeper co-design and innovation across the connectivity stack. Read the press release: Important information and disclosures:
783833227.9K
Applied Compute update: - crossed $100M annualized revenue run rate up from zero 18 months ago - raised $350M at $3.25B led by Elad Gil - 40% of customers use AC for both custom model training + production inference - inference serving 1T tokens a day after launching 12 weeks ago - 2X enterprise customers since last raise - 3X team since January, 85% are researchers and engineers - Working with customers like Microsoft, Shopify, Harvey, DoorDash, Cognition, Clay, Decagon and others To move even faster we need people who are extremely technical AND extremely practical. If this is you, please reach out.
122651K168K
01441
Over the last week I built a bottoms up SpaceX financial model through the end of 2028 The numbers that came out are insane There are wide error bars given how quickly things are changing but I wanted to distill something complicated into something that can at least be debated * This is not investment advice!!! ** Sequoia is a major SpaceX shareholder *** Past performance is not indicative of future results
5335426.6K2.8M
today starts an experiment i’m pretty excited about. evaluating if going fully self-reliant on compute is viable vs. frontier first, @huggingface to get a feel for models. then deciding if i need more sparks for an 8 node cluster or if the models that run on a 4-node suffice. the experiment will run over 8 months where i’ll specifically be focusing on: 1) using this cluster as my sole driver for work (my day-to-day is reverse engineering, cyber and swe) 2) provide rolling updates on my configurations and results (optimal model/harness combinations for my work) 2) create & open source custom tooling, agentic toolkits & workflows if they give the models a significant boost in their work 3) be transparent about how this compares to working with closed frontier models (rolling evals & benchmarks) 4) documenting the highs and lows this is a full local compute replacement i’m binding myself to for minimum 8 months. i’ll be sharing the projects i work on, whether these things can actually “pay for themselves” if operated with intent, and hopefully make meaningful contributions to the local compute open source community some quick background: i’m an experienced reverse engineer, pentester and swe (i have, despite all odds, retained my ability to code without LLM handholding). hope you’ll follow along - let me know if there’s something specific you’d like me to give attention/test if you’re on the edge about hopping over to local compute. side note: there’s some extremely interesting things happening on the smart TV anti-telemetry project happening right now. hoping i can share more in the next couple days.
86261K35K
DAILY AI BRIEF 🗞 — Nov 3 ANTHROPIC 🔥: - Anthropic's amended IPO filing shows $17B in Q3 revenue and $895B in compute commitments. - Message threads are now in public beta on the Claude Messages API. OPENAI 🔥: - OpenAI raised bug bounties to up to $172K and teamed up with Cathedral Systems on a security sprint. - OpenAI released LifeVal, a benchmark for everyday assistant tasks. - Dots are expanding to GPT-6 Astra and GPT-6.1 Terra. - Sam Altman says the model behind the Navier-Stokes result is the next GPT and is coming soon (via deredleritt3r). GOOGLE 🔥: - An early Gemini Live mode is showing up in Antigravity (via testingcatalog). - The Antigravity app added a built-in side-panel browser and a PDF viewer. SPACEXSI 🔥: - Grok Bot can now make phone calls for you. - Grok Bot is testing private bot sharing with specific users (via blankspeaker). - Grok Bot is testing free usage resets for opting in to training data (via blankspeaker). PERPLEXITY 🔥: - Perplexity released pplx-embed-v2-context, open-weight vision-language embedding models in 9B and 0.6B sizes. TENCENT 🔥: - Tencent launched Hyra 2.0, its upgraded Hunyuan Research Agent. ALIBABA 🔥: - Qianwen Go opened a free tier that gives each user a dedicated VM for agent tasks. NVIDIA 🔥: - NVIDIA open-sourced Auto Ontology, a semantic layer that grounds agent SQL queries. FACTORY 🔥: - Factory released Droid 1.0, a coding-agent harness that works with any model. * Used Grok to compose this brief, cherry-picking the news and doing some post-editing.🚨 AI News | TestingCatalog @testingcatalogDAILY AI BRIEF 🗞 — Nov 2 GOOGLE 🔥: - Gemini 4 Carbon is reportedly being tested on Arena as Argon 4 (via Waguri_Kaoruko8). - Antigravity may get cloud execution support next week. - Antigravity's "Chief of Stuff" agent will get Gemini Live support. - Antigravity is getting a new security notice about coding agent risks. - Gemini Notebook may add a "Comic" artifact type. OPENAI 🔥: - GPT-6.2 Astra is showing up in Codex and ChatGPT Work usage analytics (via StatsWire). ANTHROPIC 🔥: - A hidden Claude Mythos 5.5 mention was spotted on Anthropic's website. SPACEXSI 🔥: - SpaceXSI will enable text watermarking on Nov 7. DEEPSEEK 🔥: - DeepSeek open-sourced dsh-superagent, a research prototype of an autonomous coding agent. - DeepSeek V4.1 Pro is showing up in benchmark setups (via LuminaBench). MEITUAN 🔥: - Meituan released LongCat-2.5, a native multimodal model for agents, API only. SEMIANALYSIS 🔥: - SemiAnalysis released GDPx, an agentic benchmark where the top model scores 12%. HERMES 🔥: - The Hermes mobile app opened beta testing on iOS and Android, and both betas are now full. * Used Grok to compose this brief, cherry-picking the news and doing some post-editing.55611410.4K- DAILY AI BRIEF 🗞 — Nov 2 GOOGLE 🔥: - Gemini 4 Carbon is reportedly being tested on Arena as Argon 4 (via Waguri_Kaoruko8). - Antigravity may get cloud execution support next week. - Antigravity's "Chief of Stuff" agent will get Gemini Live support. - Antigravity is getting a new security notice about coding agent risks. - Gemini Notebook may add a "Comic" artifact type. OPENAI 🔥: - GPT-6.2 Astra is showing up in Codex and ChatGPT Work usage analytics (via StatsWire). ANTHROPIC 🔥: - A hidden Claude Mythos 5.5 mention was spotted on Anthropic's website. SPACEXSI 🔥: - SpaceXSI will enable text watermarking on Nov 7. DEEPSEEK 🔥: - DeepSeek open-sourced dsh-superagent, a research prototype of an autonomous coding agent. - DeepSeek V4.1 Pro is showing up in benchmark setups (via LuminaBench). MEITUAN 🔥: - Meituan released LongCat-2.5, a native multimodal model for agents, API only. SEMIANALYSIS 🔥: - SemiAnalysis released GDPx, an agentic benchmark where the top model scores 12%. HERMES 🔥: - The Hermes mobile app opened beta testing on iOS and Android, and both betas are now full. * Used Grok to compose this brief, cherry-picking the news and doing some post-editing.842033233.7K
ZedOFF@zedOFFrus#Supercomputer is staring on #computer for @Pixel_Dailies #pixel_dailies #pixelart #aseprite #Portal #GlaDOS #grafxkid339160
Shaun Maguire’s bottoms-up SpaceX $SPCX model (50th percentile) shows a clear inflection starting in 2027. I converted the quarterly figures into charts, and the 2027 revenue explosion stands out immediately. Key points: • FY2026 revenue: ~$51.5B • FY2027 revenue: ~$274B (≈5.3× year-over-year) • FY2028 revenue: ~$731B • Driver: Space Infrastructure (SI), especially terrestrial compute • Terrestrial compute capacity scales from 1 GW (early 2026) to 21 GW (end-2028) • Assumption: ~$40B annual revenue per GW • By FY2028, SI is ~85% of total revenue • Operating income turns positive in Q3’26, accelerates from 2027 • EBITDA reaches ~$536B in FY2028 • Free cash flow stays negative through 2027, turns positive mid-2028 amid ~$369B of annual capex • Starship ramps to 250 flights in 2028 Shaun himself noted this is a 25th–50th percentile estimate that factors in risks like political headwinds and major supply chain issues. Elon replied: “I think we can do better.”
Shaun Maguire @shaunmmaguireOver the last week I built a bottoms up SpaceX financial model through the end of 2028 The numbers that came out are insane There are wide error bars given how quickly things are changing but I wanted to distill something complicated into something that can at least be debated * This is not investment advice!!! ** Sequoia is a major SpaceX shareholder *** Past performance is not indicative of future results181091.1K216K
- Show more posts
