All Updates
Page 1733 of 1960
March 6, 2026
Black Sesame Eyes 69% Revenue Jump in 2025
Black Sesame Intelligent forecasts 2025 revenue exceeding 800M RMB, up over 68.7% YoY. Growth driven by high-tier ADAS mass production, L2-L3 commercial vehicles, L4 unmanned logistics shipments, and surging robot business. Net losses narrow to under 1.48B RMB amid AI R&D spends.
AI Fears Ease for IT Giants
Concerns over AI are pressuring IT service companies like Accenture. Profit margins in IT services have been declining for years. Analysts say there's no need to worry.
Anthropic Challenges DOD Supply Chain Label in Court
Anthropic CEO Dario Amodei plans to challenge the Department of Defense's designation of the AI firm as a supply chain risk in court. He claims most Anthropic customers remain unaffected by the label.
Google Workspace CLI for AI Agents
Google Workspace CLI provides a unified command-line interface for apps like Gmail, Docs, Sheets, Drive, and Calendar, designed for humans and AI agents. It offers structured JSON output and eliminates third-party connectors like Zapier for automation. Developers can install it via npm for easy terminal-based workflows.
Chardet License Shift Signals AI's Threat to Software Licensing
Dan Blanchard, maintainer of the Python chardet library, released a new version under a changed license. Bruce Perens argues this dispute illustrates how AI will dismantle traditional software licensing. Alarm bells are ringing in the open source community, with commercial licensing also at risk.
China Targets 12.5% GDP from Digital Economy by 2030
China aims for its digital economy to reach 12.5% of GDP by 2030, up from 10.5% achieved in 2025. This target, announced at the 'two sessions' meetings, exceeds initial goals and emphasizes AI integration with advanced manufacturing. The strategy seeks to build a modern industrial system and narrow the digital economy gap with the US.
MacBook Neo vs Air Winner
MacBook Neo and 13-inch MacBook Air seem similar on surface. Closer look reveals differences; article picks winner with buy reasons.
Anthropic to Fight Pentagon in Court
Anthropic PBC faces blacklisting threats from the US Defense Department over disputes on AI safeguards. The company states it has no choice but to pursue a legal battle. This standoff highlights tensions between AI safety standards and government contracts.
AI Bubble Tied to Mideast War Risks
Canadian scholar Jiang Xueqin predicts US defeat in Iran war, arguing it would disrupt Middle East petrodollar funding for US AI boom via data centers and investments. Tech giants like AWS, Azure, and Google Cloud are building massive AI infrastructure in UAE and Saudi Arabia, fueled by cheap energy and sovereign wealth funds. War could target these assets, crashing AI valuations and supply chains.
Apple iOS 26.4 Beta 3 v2 Bug-Fix Revision
Apple released iOS/iPadOS 26.4 developer Beta 3 v2 (build 23E5223k), replacing the prior 23E5223f to address bugs. No new features added, typical for revisions fixing critical issues. watchOS 26.4 Beta 3 also revised to 23T5226i.
Nanya Advances Custom AI Memory R&D
Nanya Technology revealed custom AI memory R&D progress, with products entering trial production using UltraWIO architecture. It mimics HBM-GPU integration for higher AI data bandwidth. More details expected H2 2024.
Poggio: AI Needs Maxwell-Like Theory
MIT's Tomaso Poggio likens current AI to pre-Maxwell electricity, driven by engineering without deep theory. He advocates sparse compositionality—combining simple low-dim functions—for generalization in intelligent systems. Discussion covers his learning theories from kernel machines to deep nets.
Overseas Power Crunch Spurs Chinese Vendor Exports
Overseas data centers face acute power shortages, with grid queues exceeding 7 years in congested EU/US areas. Investments shift to North/South Europe and emerging markets as self-provided power becomes mandatory in some regions. Chinese power vendors gain a prime window for accelerated overseas expansion.
NVIDIA GTC: Focus on LPU and Supply Chain Leaders
CITIC Securities highlights NVIDIA GTC potential LPU showcase amid LLM inference trends like PD separation. Post-Groq acquisition, NVIDIA may launch LPU-like chips for Decode efficiency alongside Rubin CPX for Prefill savings. Recommends NVIDIA core firms, LPU investors, and domestic super-node/calc leaders.
GPT-5.4 Launches with Native PC Control
OpenAI released GPT-5.4 with native support for controlling computers, enabling direct interaction with desktop environments. Alibaba CEO confirmed executive departure but reaffirmed open-source strategy and increased AI investments. OpenAI is preparing for an IPO with a potential $730B valuation.
Palantir Blocks Ex-Employees' Poaching
Palantir Technologies won a court order blocking three former AI employees from poaching its staff. The injunction also prevents them from using confidential information for their new AI startup. This protects Palantir's talent and IP amid fierce AI competition.
Tech Vet: Best Time for AI Startups
Sudheesh Nair, a Silicon Valley tech veteran and co-founder of enterprise web agent startup TinyFish, declares the current AI boom as one of the best opportunities for founders in years. He emphasizes it's not just another tech cycle but a major opening for new companies.
Lightweight Fusion Beats Heavy Deepfake Detectors
Apple ML proposes LFWS and LFWL, lightweight detectors fusing handcrafted cues with Xception backbone. They add just 292 parameters for higher accuracy than wide/dual-stream models, maintaining 21.9M total params. This enables robust video face forgery detection with smaller models.
GenCtrl: Formal Toolkit for Generative Control
Apple releases GenCtrl, a toolkit addressing controllability in generative models. It frames human-model interaction as a control process and introduces an algorithm to estimate controllable sets in dialogues. The work offers formal guarantees on estimation error.
Apple's Semidiscrete Couplings for Flow Matching
Apple Machine Learning presents Flow Matching with Semidiscrete Couplings, improving generative flow models. These models integrate ODEs via time-dependent velocity fields from noise to data. The method enhances training by carefully matching batches of noise points to target points, beyond independent sampling.