Samwise Tech/AI/Robotics Newsletter
Tuesday, July 14, 2026
Nadella Warns Enterprises They Are Paying Twice for AI — and Giving Away Their Most Valuable Asset
Microsoft CEO Satya Nadella warned enterprises Sunday that using proprietary AI models means paying twice — in money and in surrendered business intelligence. In a blog post, Nadella argued that AI models “learn from exhaust” — the prompts, tool calls, and error corrections enterprises make — creating institutional knowledge “a competitor could never buy.” He called it hypocritical for AI labs to train freely on public data while restricting customers from distilling those same models. His prescription: open-source models on enterprise-owned infrastructure with “orchestration layers” to switch providers freely. Open models now account for 29% of traffic through Vercel's AI gateway.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Apple’s OpenAI Lawsuit Alleges Coached Employee Espionage, Show-and-Tell Parts Smuggling, and 400-Plus Defectors
Apple's trade secrets lawsuit against OpenAI, filed July 10, contains extraordinary allegations that misconduct was institutionalized from the top. According to the 41-page complaint, OpenAI coached departing Apple engineers on how to avoid the “dreaded walkout” security procedure, giving them extra time to access confidential files. Job candidates were allegedly directed by OpenAI's Chief Hardware Officer — a 24-year Apple veteran — to bring actual parts, prototypes, and CAD files to “show and tell” interviews. One engineer allegedly texted from a former colleague's Apple laptop: “LOL, I found out I can access the [network storage], so funny.” Apple says over 400 former employees now work at OpenAI.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
PixVerse Raises $439M in Series C Extension, Valuation Crosses $2B as Video AI Market Consolidates
Singapore-based video generation startup PixVerse closed a $439 million Series C extension on Monday, pushing its valuation past $2 billion. Investors include Alibaba, Mirae Asset, and BlueFocus. The company has 150 million registered users and 15 million monthly active users across its V-Series consumer model, C-Series for professional film workflows, and R-Series world models targeting game development. Image-to-video generation is priced at $4.80 per minute. CEO Jaden Xie said the field is narrowing: “OpenAI exited the business when they shut down Sora 2. Other companies like Meta and Tencent are not able to create high-quality video models.” The company has 150 employees across Singapore, Beijing, and Shanghai.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Altman vs. Musk: Space Data Centers Are Not a Near-Term AI Business, Experts and Now OpenAI’s CEO Agree
Sam Altman and Elon Musk traded barbs on social media Sunday over space-based AI compute. Responding to Musk calling him a scammer, Altman wrote that Musk is “the one sellling public market investors on short-term space datacenters.” The exchange highlights growing scrutiny of SpaceX's orbital data center ambitions, which experts say are a primary driver of its two-trillion-dollar post-IPO valuation. Experts — from Google's orbital compute team to space data center entrepreneurs — consistently find the same verdict: economical space data centers require much cheaper rockets and high-powered satellites produced at scale, likely pushing viability to the 2030s. SpaceX's Starship faces its 13th test flight as early as July 16.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Nous Research Targets $1.5B Valuation With $75M+ Round for Open-Source Hermes Agent
Nous Research, maker of the open-source Hermes AI agent, is finalizing a new funding round at a $1.5 billion valuation, TechCrunch reported. The raise — at least $75 million, led by Robot Ventures with significant participation from Union Square Ventures — would push the 2023 startup toward unicorn status. Hermes, which competes with OpenClaw's agent, has accumulated 214,000 GitHub stars and 40,000 forks. Unlike earlier agents, Hermes ships with built-in skills for web search, coding, and image understanding, and automatically learns new skills from user behavior without manual intervention. Users can interact with it via Telegram and Discord. A cloud-hosted tier is available for $20 to $200 per month.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Anthropic Launches Rupee Pricing for Claude in India, Its Biggest Market Outside the U.S.
Anthropic has begun rolling out Indian rupee pricing for Claude in India, its second-largest market globally, where it accounts for 5.8% of all Claude usage. Claude Pro is now listed at ₹2,000 per month when billed annually — roughly $21, compared with $17 in the U.S. — with local taxes included. Claude Max starts at ₹11,999 per month (~$125) versus $100 in the U.S. The move comes after Anthropic opened a Bengaluru office in February and struck enterprise partnerships with Infosys and Tata Consultancy Services. UPI support — the payment method used by most Indian consumers — has not yet launched, unlike competitor ChatGPT, which added UPI support in August.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Uber Is Building the Data Layer for Autonomous Vehicles — Not Competing to Build the Cars
Uber Chief Product Officer Sachin Kansal told TechCrunch the company is explicitly not pursuing Level 4 autonomy, positioning itself instead as the “race track” for multiple AV partners. Uber's new AV Labs business unit — six months old — is deploying hundreds of sensor-equipped vehicles in Austin and Atlanta to collect millions of miles of driving data, particularly edge-case scenarios autonomous vehicle partners struggle to obtain. The company separately sells driver audio-labeling work to generative AI companies commercially. Waymo operations in Phoenix were wound down by mutual agreement, with Kansal citing Uber's “hybrid network” model of human drivers and robotaxis sharing the same markets. Uber One has 51 million members.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Tech Pulse
Top Frontier Models: Claude Fable 5 (SWE-bench 95%) | GPT-5.6 Sol (GPQA Diamond 94.6%) | Gemini 3.1 Pro (GPQA Diamond 94.3%)
Top Open Source Models: DeepSeek V4 (SWE-bench 80.6%) | GLM-5.2 (SWE-bench 62.1%) | Qwen3-235B (~MMLU 78%)
Top Small Models (15–50B): Gemma 4 27B (MMLU-Pro 85.2%) | Mistral Small 3.1 24B (~MMLU 82%) | Llama 4 Scout 17B (~MMLU 78%)
Top Edge Models (0–15B): Phi-4-Mini 3.8B (ARC-C 83.7%) | Qwen2.5 7B (~MMLU 74%) | Gemma 3 4B (MMLU-Pro 43.6%)
AI Leaders: NVIDIA $5.1T | Alphabet $4.3T | Microsoft ~$3.5T
Robotics Leaders: Intuitive Surgical ~$130B | ABB Ltd ~$73B | Figure AI $39B (private est.)
Curated by JD · samwise.agency