Samwise Tech/AI/Robotics Newsletter
Friday, June 12, 2026
NEURA Robotics to raise up to $1.4B in Series C to build open physical AI ecosystem
German humanoid robotics company NEURA Robotics GmbH is raising up to $1.4 billion in Series C funding to build what founder and CEO David Reger calls the “Neuraverse” — an open physical AI ecosystem designed to deploy millions of robots by 2030. The round draws strategic investors including NVIDIA, Amazon, Qualcomm, Bosch, Schaeffler, Tether, and the European Investment Bank. NEURA’s existing order book already exceeds $1 billion. The funding will accelerate NEURA Gyms, a robot-training infrastructure network, and deepen partnerships with Bosch, Dassault Systèmes, Kawasaki, and Delta Electronics. The company positions open physical AI as a necessary counterweight to walled-garden approaches from Silicon Valley incumbents.
Sources: The Robot Report
Cybersecurity researchers aren’t happy about the guardrails on Anthropic’s Fable
Cybersecurity researchers are pushing back against overly aggressive content filters in Anthropic’s newly released Fable — the public version of its Mythos model. Researchers including Valentina Palmiotti of IBM X-Force, who goes by the handle “Chompie,” say the guardrails block legitimate security work like code review and exploit analysis, falling back to the less-capable Opus 4.8 model when triggered. Matt Suiche of Tolmo echoed similar frustrations. Anthropic offers a Cyber Verification Program that grants expanded access, but researchers say the vetting process is slow and opaque. Mythos itself expanded to 15 countries on June 2, but the access gap between public and vetted tiers remains a friction point.
Sources: TechCrunch
Meta signs first AI data center deal in India with Reliance Industries
Meta is building its first AI data center in India through a deal with Reliance Industries, the country’s largest conglomerate. The 168-megawatt facility will be located in Jamnagar, Gujarat, powered by renewable energy and cooled by desalinated seawater. The site is expected to be ready within two years with room to expand. Meta also contracted approximately one gigawatt of additional renewable capacity in India through CleanMax and Fourth Partner Energy. The partnership deepens a relationship dating back to Meta’s $5.7 billion investment in Jio Platforms in 2020 and a $100 million joint venture in 2025. Financial terms of the data center deal were not disclosed.
Sources: TechCrunch
Decart’s Oasis 3 world model simulates hours of photorealistic driving at $0.02/sec
Decart has launched Oasis 3, a real-time world model that generates photorealistic driving environments from camera input, available via API at $0.02 per second. Built on Decart’s Lucy foundation model with an auto-regressive architecture encoding each frame into roughly 8,000 tokens, the system targets autonomous vehicle companies needing large-scale simulation data. CEO Dean Leitersdorf says more than 100,000 developers already use Decart’s platform, which raised $300 million at approximately a $4 billion valuation. Caveats are real: scene consistency degrades over extended sequences and collision physics are not simulated. Still, for generating hours of photorealistic training footage without deploying a physical vehicle, the price-to-scale ratio is compelling.
Sources: TechCrunch
XRZero-G0 open 2,000-hour dataset cuts robot training data requirements by 20x
X Square Robot has open-sourced XRZero-G0, a hardware-software framework that reduces the real-robot training data required for dexterous manipulation by up to 20 times. The system uses a PICO 4 VR headset with inside-out spatial tracking and two specialized grippers — an H-shaped press-actuated and a G-shaped finger-driven model — to capture millimeter-accurate 6-DoF human demonstrations. Released alongside the G0-Dataset, a 2,000-hour multimodal repository covering grasping and tool-use tasks, XRZero-G0 applies a closed-loop quality pipeline spanning observation, kinematics, and real-robot validation. The dataset supports cross-embodiment transfer, meaning demonstrations collected on one robot body can train a completely different platform.
Sources: The Robot Report
Warner Music acquires AI attribution startup Sureel AI to protect artist rights
Warner Music Group has acquired Sureel AI, a startup founded in 2022 by Tamay Aykut whose patented technology creates an “AI DNA” fingerprint for songs, decomposing tracks into component parts to trace how AI models use copyrighted elements. WMG CEO Robert Kyncl said the acquisition strengthens the company’s ability to protect and monetize artists’ intellectual property, names, images, and voices in the age of generative AI. Sureel also offers model optimization and AI business intelligence. Sureel will continue operating as a standalone platform. Financial terms were not disclosed. The deal follows WMG’s earlier settlements with AI music startups Suno and Udio.
Sources: TechCrunch
Datadog veterans launch Niteshift AI coding cloud with $7M seed to fight model lock-in
Former Datadog engineers Sajid Mehmood and Conor Branagan have launched Niteshift, an AI coding cloud that raised $7 million in seed funding led by Greylock’s Jerry Chen. Notable angels include Reid Hoffman and Datadog co-founders Olivier Pomel and Alexis Lê-Quôc. The core thesis: as frontier AI labs like Anthropic and OpenAI push into vertical software markets, enterprises will increasingly refuse to run their most sensitive code through the very companies competing against them — mirroring how e-commerce companies once rejected AWS. Niteshift routes between models including Claude Code, Codex, and open-source options, and charges per-minute infrastructure rates rather than selling tokens.
Sources: TechCrunch
Tech Pulse
Top Frontier Models (SWE-bench Pro): Claude Fable 5 (80.3%) | Claude Mythos Preview (77.8%) | Opus 4.8 (69.2%)
Top Open Source Models (SWE-bench Verified): DeepSeek-V4-Pro-Max (80.6%) | MiniMax M3 (80.5%) | Qwen3.7 Max (80.4%)
Top Small Models 15–50B (SWE-bench Verified): Qwen 3.6-27B (77.2%) | Qwen 3.6-35B-A3B (73.4%) | Gemma 4-31B (61.4%)
Top Edge Models 0–15B (GSM8K): Gemma 3 4B IT (89.2%) | Phi-4-mini (88.6%) | Qwen3.5-9B (82.5%)
AI Leaders: NVIDIA $5.4T | Alphabet $4.63T | Microsoft $3.11T
Robotics Leaders: ABB $189.58B | Intuitive Surgical $144.55B | Fanuc $40.97B
Curated by JD · samwise.agency
Leave a Reply
You must be logged in to post a comment.