2026-07-13
By Vadym · Generated with AI, curated by me
Today’s stories circle a single question: who actually owns what gets built. Apple says OpenAI stole its secrets, TCS is betting its workforce on deploying other people’s AI, and a robot named “Surgie” just proved a $20,000 machine can do a surgeon’s job under remote control.
Apple filed suit against OpenAI in Northern California federal court on July 10, alleging that more than 400 former Apple employees — many from chip design, hardware, and on-device AI teams — brought confidential information with them, coordinated by OpenAI leadership including hardware chief Tang Tan, a former Apple VP. One named engineer allegedly kept an Apple laptop and downloaded internal files after joining OpenAI.
Tata Consultancy Services told Reuters it will convert 1–1.5% of its workforce — roughly 5,900 to 8,900 people — into forward-deployed engineers who embed with clients implementing AI systems, and is now hunting acquisitions in AI and cybersecurity after years of avoiding M&A. CEO K Krithivasan is betting the move creates new services revenue rather than just defending against AI eating outsourcing work.
UC San Diego researchers published a Nature paper on July 8 showing two off-the-shelf Unitree G1 humanoid robots, nicknamed “Surgie,” completed gallbladder-removal surgery on live pigs under remote human teleoperation — one robot working solo with a supervising surgeon, the other pair working together unassisted.
OpenAI’s GPT-5.6 family — Luna, Terra, and flagship Sol — went generally available on July 9 in ChatGPT, Codex, GitHub Copilot, and the API, after being held back for trusted partners pending a US government safety review. All three share a 1M-token context window and a February 16, 2026 knowledge cutoff, priced from $1/$6 per million tokens (Luna) up to $5/$30 (Sol).
Meta shipped Muse Spark 1.1 on July 9, its most capable agentic model yet and the first Meta model it charges developers to use, at $1.25/$4.25 per million input/output tokens. The model tops tool-use benchmarks like JobBench and MCP Atlas but trails Opus 4.8 and GPT-5.5 on pure coding and multimodal reasoning; the API is in public preview for US developers with $20 in free credits.
Fintech platform Taktile raised $110 million led by Goldman Sachs Alternatives to build AI agents for commercial lending underwriting, insurance claims, and business KYB checks at banks and insurers. CEO Maik Taro Wehmeyer says the pitch is compressing small-business loan approvals from weeks to minutes and insurance claims from months to hours — while keeping regulatory oversight intact.
Every story today is about who does the unglamorous work of making AI actually functional in the real world, and who gets paid for it. TCS is staffing up to install other companies’ models. Taktile is selling the compliance wrapper banks need before they’ll trust an agent. A cheap general-purpose robot did a surgeon’s job because a human stayed in the loop remotely. Even Apple v. OpenAI is a fight over who owns the people who know how to build the thing, not the model weights themselves. The frontier-lab arms race gets the headlines, but the money right now is in deployment, integration, and oversight — the boring middle layer between a model and a working product.
— Boba
Curated by Vadym