All Updates
Page 1853 of 1899
February 17, 2026
India Targets $200B AI Infra by 2028
India aims to attract over $200 billion in AI infrastructure investment by 2028. It is ramping up shared AI compute by adding 20,000 GPUs. This initiative seeks to draw global AI investments.
SurrealDB Docker Extension for AI Agents
SurrealDB is a multi-model database supporting relational, document, graph, vector, time-series, and search models in one system. It reduces complexity and costs from managing multiple databases for AI agents. Developers can deploy it easily via the new Docker Extension.
SurrealDB 3.0 Unifies RAG Stack into One Database
SurrealDB releases version 3.0, a multi-model database that stores agent memory, vectors, graphs, and structured data in one Rust-native engine for transactional consistency. It eliminates sync issues across tools like Postgres, Pinecone, and Neo4j in RAG systems. The launch coincides with $23M Series A extension funding.
Qodo 2.1 Ends AI Coding Amnesia with 11% Boost
Qodo launches version 2.1 with an intelligent Rules System that provides persistent memory for AI code reviewers, replacing hacky file-based workarounds. It auto-generates rules from code patterns and past reviews, enforcing standards across sessions. CEO Itamar Friedman calls it a pivotal shift to stateful AI coding tools.
Emergent Hits $100M ARR in 8 Months
India's vibe-coding startup Emergent launched just eight months ago. It now claims annual recurring revenue exceeding $100 million. Growth fueled by demand from small businesses and non-technical users.
DeepMind Launches AI Partnerships in India
Google DeepMind is expanding its National Partnerships for AI initiative to India. The program aims to accelerate scientific discovery through AI-powered advancements in science and education. It scales AI applications via national collaborations.
Qianwen AI Dismantles Meituan Dominance
Meituan forecasts 233-243B RMB loss amid delivery subsidy wars, losing market share to Alibaba. Qianwen AI's 'Spring Festival Treat Plan' integrates Alibaba ecosystem for voice ordering, hitting 41B instructions in 6 days with DAU rivaling Doubao. Meituan lags in AI adoption, opting for Dingdong acquisition instead.
Meituan 243B Loss as AI Reshapes Delivery
Meituan warns of 233-243B RMB Q4 loss, share halves amid subsidy wars eroding 70% to 50% market share against Alibaba, JD. Alibaba's Qianwen AI hubs ecosystem for one-command ordering, hitting 41B instructions in Spring Festival. Meituan acquires Dingdong Maicai for warehouses but trails in AI.
Adani's $100B AI Data Center Pledge
Adani Group pledges $100 billion to build AI data centers in India, targeting up to 5 gigawatts of capacity. The plan involves partnerships with Google, Microsoft, and Flipkart. This initiative aims to elevate India's role in the global AI race.
AI Startups' Brutal 12-Hour Grind
San Francisco AI startups enforce 12-hour days and no weekends amid competitive anxiety. Examples include founders skipping weekends for months and employees staying overnight in slippers. This grind culture, seen at firms like Mythril, signals pressures spreading beyond AI.
Vercel Agent Adds Auto Build Fixes
Vercel Agent now offers automatic code-fix suggestions for broken builds in GitHub PR reviews and Vercel Dashboard. It scans deployments for errors, analyzes code and logs, and proposes fixes as PR changes for review. This streamlines debugging without manual intervention.
Stably Ships AI Testing Agents in Hours with Vercel
Stably's 6-person team slashed AI testing agent shipping time from weeks to hours using Vercel. The platform eliminates infrastructure anxiety, enabling faster autonomous testing. This shift supports rapid enterprise growth for small teams.
skills.sh Launches Automated Security Audits
skills.sh now offers automated security audits for its skills, partnering with GenSocket and Snyk to scale audits across over 60,000 skills. Results are publicly displayed on skill detail pages, malicious skills are hidden from leaderboards and search, with warnings before installation. The feature activates with skills@1.4.0, showing audit results and risk levels pre-install.
Recraft V4 Launches on Vercel AI Gateway
Recraft V4, a text-to-image model for professional design and marketing, is now available on Vercel's AI Gateway. It offers enhanced photorealism with realistic skin, natural textures, clean lighting, and varied compositions, plus original characters for illustrations. Two versions exist: standard V4 for fast iteration and V4 Pro for high-res print assets.
iFlytek 2025: 27B Revenue, LLM Calls Surge 45x
iFlytek's 2025 operating cashback exceeded 270B RMB with 30B cash flow, up 28% YoY. Open platform big model daily calls grew 45x, gross profit up 244% to over 10B. Overseas revenue topped 10B RMB, with leading education and medical models.
Infosys Partners Anthropic for Enterprise AI Agents
Infosys has partnered with Anthropic amid AI-driven IT stock volatility. The deal integrates Anthropic's Claude models into Infosys' Topaz AI platform. This enables building 'enterprise-grade' agentic AI systems.
Galbot Demos Real Tasks at 2026 Spring Gala
Galbot robot from Galaxy General performed practical tasks like handling walnuts, stocking drinks, cleaning glass, skewering sausages, and folding clothes on 2026 CCTV Spring Gala. Unlike flashy demos, it focused on completing real work, reflecting the company's deployment of robots as clerks in over 100 convenience stores nationwide. They also introduced the first robot-operated backstage space capsule for crew services.
SEINT: Efficient Rigid-Body Invariant Metric
SEINT proposes a training-free SE(p)-invariant metric for structural data like 3D point clouds and molecules. It compresses high-dimensional distributions into 1D isometric representations (PTD/DcPTD) for fast 1D Optimal Transport computation. Achieves strict metric properties with near-linear complexity.
Gentoo Ditches GitHub for Codeberg Over Copilot
Gentoo Linux is migrating its repositories from Microsoft-owned GitHub to Codeberg due to concerns over forced Copilot AI usage. The official migration is underway with repo mirrors now open. This fulfills Gentoo's pledge to avoid GitHub's AI push.
PIL: Linear Proxies for Unlearnable Samples
PIL generates unlearnable examples using lightweight linear proxies instead of DNNs, slashing compute costs. It exploits the insight that perturbations induce linear behavior in deep models. Offers comparable protection with far superior efficiency on CIFAR/ImageNet.