Nous Research has published Lighthouse Attention, a selection-based hierarchical attention mechanism that wraps around standard scaled dot-product attention during pretraining and is removed afterward. Unlike prior methods such as NSA and HISA that pool only keys and values, Lighthouse pools Q, K, and V symmetrically across a multi-resolution pyramid, reducing the attention call from O(N·S·d) to O(S²·d) and running stock FlashAttention on a small dense sub-sequence. Tested on a 530M Llama-3-style model at 98K context, it achieves a 1.40–1.69× end-to-end wall-clock speedup against a cuDNN SDPA baseline with matching or lower final training loss.
The post Nous Research Proposes Lighthouse Attention: A Training-Only Selection-Based Hierarchical Attention That Delivers 1.4–1.7× Pretraining Speedup at Long Context appeared first on MarkTechPost.
Anthropic's strategy may accelerate AI adoption in regulated sectors, influencing market dynamics and boosting decentralized compute solutions.
The post Anthropic prioritizes speed to market over compute costs, says analyst appeared first on Crypto Briefing.
Friday’s Claude for Legal webinar confirmed – without any doubt – what AL has believed since the first plugin arrived, namely that Anthropic intends to ...
The post Anthropic’s Playbook for AI-Native Startups Highlights Claude’s Role appeared on BitcoinEthereumNews.com.
Peter Zhang
May 14, 2026 17:10
Anthropic releases a comprehensive playbook for founders building AI-native startups, showcasing Claude’s capabilities across the startup lifecycle.
On May 14, 2026, Anthropic released a detailed guide, The Founder’s Playbook, aimed at helping entrepreneurs build AI-native startups using its flagship AI model, Claude. The playbook remaps the traditional startup journey—Idea, MVP, Launch, and Scale—through an AI-first lens, offering practical frameworks, exercises, and case studies tailored to startups leveraging Claude’s capabilities. This release comes as Anthropic solidifies Claude’s position as a foundational tool for emerging companies. Claude, introduced in 2021, has evolved into a general-purpose AI system capable of coding, automating workflows, and executing multi-step reasoning tasks. Recent updates, such as memory
The post Lost Bitcoin Just Got an AI Lifeline — and a New Privacy Problem appeared on BitcoinEthereumNews.com.
A Bitcoin holder’s claim to have used Anthropic’s Claude to recover roughly 5 BTC — worth around $395,000 at current prices — has set off a fresh debate about the role AI tools could play in cracking open the multi-million-coin overhang of lost Bitcoin. But the recovery, while real, isn’t quite the cryptographic breakthrough that went viral on X this week. X user @cprkrn posted on May 13 that Claude had located an old wallet backup on his college-era computer after eight weeks of failed brute-force attempts using the btcrecover service on a rented Vast.ai GPU — roughly 3.5 trillion password combinations and around $15 of compute, by his own accounting. The wallet had been dormant since April 2015. “HOLY FUCKING SHIT OMG CLAUDE JUST CRACKED THIS SHIT,” cprkrn posted, vowing to name his next child after Anthropic CEO Dario Amodei, Source: X What Claude actually did was more mund
Anthropic's rapid valuation surge signals a transformative shift in AI investment dynamics, potentially reshaping market expectations and strategies.
The post Anthropic raises over $30B ahead of IPO as AI valuations enter uncharted territory appeared first on Crypto Briefing.
The post Anthropic warns human-level AI could arrive by 2028 and urges US to tighten export controls on China appeared on BitcoinEthereumNews.com.
A top American artificial intelligence corporation issued a strong warning about Chinese AI growth just as Presidents Donald Trump and Xi Jinping concluded technology collaboration talks in Beijing, resulting in an uncommon gap between industry rhetoric and political reality. The AI company Anthropic put out a research paper on Thursday claiming that machines with human-level intelligence could arrive by 2028. The company called on Washington to keep America ahead of China in developing advanced AI systems. Their paper, called “2028: Two Scenarios for Global AI Leadership,” paints a picture of AI systems soon capable of handling complicated work in science, engineering, and cybersecurity at expert human levels. Anthropic paints two futures for AI leadership Anthropic describes a future with what it calls “a country of geniuses in data cente
