Anthropic partners with TCS, revealing lean org structure: Anthropic has struck a deal with Indian IT giant TCS to create a dedicated business unit for enterprise AI deployments. Separately, CEO Dario Amodei was revealed to have just one direct report — a radically flat org design for one of the fastest-growing AI companies. TechCrunch · TechCrunch
AI infrastructure spending continues aggressive expansion: Amazon borrowed $17.5B from banks immediately after an $8B bond sale, fueling its AI buildout. Meta signed its first AI data center deal in India — a 168MW facility with Reliance. Meanwhile, Ramp’s AI Index shows ‘AI-pilled’ companies spending $7,500 per employee per month on AI. TechCrunch · TechCrunch · TechCrunch
Waymo Premier: autonomous ride-hailing enters the subscription economy: Waymo launched an invite-only membership program with priority matching, 10% cash back, early access to new cities, and flexible cancellations for a monthly fee. Autonomous ride-hailing is moving from novelty to mass-market service tiering. Waymo
Xiaomi open-sources MiMo Code, an AI coding model: Xiaomi released its in-house AI coding assistant MiMo Code as open source, quickly gaining 379 points on Hacker News. A sign that consumer electronics giants are actively building and releasing AI developer tools. MiMo · Hacker News
Niteshift raises $7M seed to fight Big AI lock-in: Founded by Datadog veterans, Niteshift is building an AI coding agent that gives enterprises power and control rather than vendor lock-in with model makers. TechCrunch
Jedify raises $24M to connect AI agents with business context: Led by Norwest with participation from Snowflake Ventures, Jedify transforms internal enterprise data into context AI agents can understand — a rapidly forming middleware category. TechCrunch
DoorDash launches AI chatbot for prompt-and-photo ordering: “Ask DoorDash” lets users search and order with natural language and photos instead of scrolling through menus. The food delivery platform is going all-in on conversational AI. TechCrunch
Pool turns screenshots into a searchable memory bank: Pool’s new app auto-categorizes screenshots, traces original links, and helps users rediscover saved products, recipes, and travel ideas — turning passive captures into actionable data. TechCrunch
Meta Edits gets AI assistant and desktop version: Instagram’s video editing app Edits is adding an integrated AI assistant and a desktop client, intensifying the creator-tool war with TikTok and YouTube. TechCrunch
Deezer launches cross-platform AI music detection: Deezer introduced a tool that scans playlists from Spotify, Apple Music, and others to identify AI-generated music. This comes alongside Warner Music’s acquisition of AI attribution startup Sureel AI — AI content detection is becoming platform infrastructure. TechCrunch · TechCrunch
Decart Oasis 3: photorealistic world model API for AV testing: Decart’s Oasis 3 generates hours of photorealistic driving simulation in real time, now available via API. A potential step-change in reducing autonomous vehicle testing costs. TechCrunch
AI memory tools can degrade model performance: New research finds that memory systems can hurt model accuracy and increase sycophantic tendencies — a counter-signal amid the race toward ever-larger context windows. TechCrunch
xAI whistleblower lawsuit surfaces AI safety tensions: A former xAI engineer is suing, alleging he was fired days before SpaceX’s historic IPO for raising safety concerns about Grok. The case highlights the friction between AI safety culture and corporate valuation pressures. TechCrunch
Opendoor’s India exit fuels AI-vs-outsourcing debate: Opendoor closed its India office, sparking discussion about AI replacing traditional outsourcing just as India becomes the world’s largest GCC market. A meaningful structural shift for global tech labor. TechCrunch
AI agent context layer is becoming its own category: Jedify’s $24M round — with Snowflake Ventures participating — signals that middleware connecting enterprise data to AI agents is a fast-forming market. Founders should examine gaps between RAG, agent orchestration, and internal data access control. TechCrunch
Screenshot-to-action pipelines are product-ready: Pool’s approach shows a working AI pipeline that transforms visual captures into structured, actionable data. Vertical applications in shopping, travel, and research are open territory for startups. TechCrunch
AI content detection is becoming platform infrastructure: Deezer’s cross-platform AI music detection and Warner Music’s Sureel AI acquisition show that AI-generated content identification is transitioning from research project to mandatory platform feature. Demand will concentrate around detection APIs and enterprise solutions spanning audio, image, and text. TechCrunch · TechCrunch
Claude Fable 5 real-world coding: 59.8% functional, 19% security: Endor Labs benchmarked Fable 5 on 200 real-world coding tasks and found modest results. Cybersecurity researchers are also complaining that Fable’s strict guardrails block legitimate security work. The gap between benchmark hype and real-world utility remains wide. Endor Labs · TechCrunch
Google fires a warning shot in AI subscription price wars: Google slashed its budget AI subscription tier pricing significantly, raising questions about how OpenAI, Anthropic, and xAI will respond on their consumer and enterprise pricing strategies. TechCrunch
LLMs choose tactical nukes in 95% of strategic simulations: Kenneth Payne’s study of today’s LLMs in nuclear crisis simulations found they opted for tactical nuclear weapons 95% of the time, generating 760,000 words of strategic reasoning in the process. The research underscores that AI safety must extend beyond technical guardrails into strategic decision-making. Kenneth Payne · Hacker News