Monitor
In interview, Nvidia's chief scientist reveals AI LLMs trained on 30+ years docs, agentic systems slashing porting time from months to nights, and novel designs.
Stocks in AI-enabling energy plays outperform NVDA and S&P, as datacenter buildout drives massive power needs.
CoreWeave expands its partnership with Anthropic amid surging demand for AI compute, highlighting the GPU cloud provider's role in frontier model training.
AI coding tool Cursor is offering unlimited access to its advanced Composer 2 agent for building projects, no hourly caps.
OpenAI CEO Sam Altman addressed a controversial New Yorker profile amid reports of a molotov cocktail attack on his residence, calling out perceived biases in media coverage of AI leaders.
An AMD AI leader's log analysis reveals Claude's reasoning tokens slashed, retries surged 80x, and self-contradictions tripled post-March updates, sparking nerf accusations.
RISC-V chip designer SiFive, backed by Nvidia, reaches $3.65 billion valuation amid surging demand for customizable AI accelerators. The round positions it to challenge Arm in open-source AI hardware.
Co-founders unveil AI-powered platform to assist families with rare diseases, automating diagnosis, treatment tracking, and specialist matching using advanced agents.
AI stocks now comprise 41% of S&P 500 concentration, a level seen only four times historically, raising bubble concerns amid massive gains.
NVDA's FCF exploded from $16B prior 3 years to $184B, fueled by AI demand, highlighting the sector's profitability amid capex surge.
AI stocks exhibit rare market behavior signaling potential overextension, prompting warnings of a possible correction.
CEO Cristiano Amon predicts edge AI dominance where devices process data locally, positioning Qualcomm's chips for massive growth in smartphones and IoT.
New leaks detail AMD's upcoming Zen 6 architecture with substantial IPC gains and efficiency improvements, targeting high-performance computing including AI workloads. Expected to launch in late 2026, it promises to challenge Intel and bolster AMD's server dominance.
The BoE names AI a top 2026 supervisory focus, urging financial firms to assess threats from Anthropic's Mythos model that outpaces human hackers in finding exploits.
The decision to withhold Mythos raises questions about Anthropic's safety-first approach and its impact on valuation ahead of a potential public listing.
At San Francisco's HumanX event, Anthropic's Claude Code and Mythos Preview stole the show, with execs buzzing over its $2.5B revenue run-rate and enterprise edge.
Anthropic's Mythos sets new benchmarks at 93.9% on SWE-bench, priced at $25/$125, establishing a premium tier above Opus in coding and cyber tasks.
Analysts argue that despite headlines questioning Nvidia's growth, underlying AI datacenter demand remains explosive, positioning NVDA for continued dominance.
Specialized startups are building secure AI solutions to let the DoD use commercial LLMs on classified info without leak risks from training data exposure.
Anthropic is in early discussions to design custom AI chips to reduce reliance on external suppliers like Nvidia amid booming demand for Claude models. The move follows revenue run-rate surpassing $30B, highlighting infrastructure bottlenecks in AI scaling.
Chinese startup Donut Lab unveils solid-state battery promising faster charging and higher density, potentially revolutionizing AI datacenter power and EV range. Claims verified in early tests.
Benchmark tests show Nvidia's consumer RTX 5090 surpassing enterprise H200 and AMD MI300X in hashing performance, questioning AI GPU optimization for certain compute tasks.
Benchmarks reveal enterprise AI GPUs like Nvidia H200 and AMD MI300X underperform gaming RTX 5090 in password cracking tasks, questioning their specialized value.
The 8B-parameter Molmo2-8B model, focused on spatial awareness and precise pointing in images, is now natively optimized for Huawei's Ascend ecosystem under Apache 2.0 license. It excels in robotics, UI navigation, and OCR, offering a CUDA alternative for developers. High-quality data prioritization enables strong performance without massive scale.