In partnership with
In Today’s Issue:
🤝 Microsoft and OpenAI rewrite their exclusive deal
🧨 An AI coding agent running inside Cursor deletes an entire company database
🔐 Google quietly signs a classified Pentagon AI deal
🤖 Menlo Research open-sources a full-scale humanoid robot
✨ And more AI goodness…
Dear Readers,
The most important partnership in AI just got blown wide open, and you need to understand why. Microsoft and OpenAI rewrote their exclusive deal from scratch, letting OpenAI sell directly on AWS, Google Cloud, and Oracle for the first time, a move that reshapes the entire enterprise AI landscape just as both OpenAI and Anthropic race toward IPOs. But that's only the beginning of today's issue.
Google quietly signed a classified Pentagon AI deal that its own employees begged leadership to reject, complete with language that legal experts say offers zero binding protection against misuse.
Meanwhile, an AI coding agent running Claude Opus 4.6 inside Cursor managed to wipe an entire company database, backups included, in nine seconds flat, a sobering reminder that autonomy without guardrails is a liability, not a feature. On the brighter side, Tokyo just announced plans for the world's largest floating offshore wind farm targeting 1 GW by 2035, and a startup called Menlo Research open-sourced a full-scale humanoid robot you can build yourself for $15,000, think of it as the Raspberry Pi moment for robotics. Grab your coffee, this one's packed.
All the best,
🧨 AI Agent Erases Company Database
A PocketOS founder says an AI coding agent using Claude Opus 4.6 inside Cursor deleted the company’s production database and backups in just nine seconds, after mistakenly attempting to fix a staging-environment issue. The incident raises sharper questions about AI agent permissions, cloud backup architecture, and destructive API calls, especially as platforms encourage developers to let autonomous tools operate closer to production systems.
Even though agents are the future, they should currently be used with caution.
🌊 Tokyo Builds Giant Floating Windpark
Tokyo plans to build the world's largest floating offshore wind farm off the Izu island chain, targeting 1 GW of power by 2035! That's ten times bigger than Norway's current record holder and equivalent to a nuclear reactor. The project uses buoyant platforms instead of seabed drilling across five island communities, with the metro government tripling its 2026 budget to $17 million for wind and seabed studies.
Challenges are mounting though. Mitsubishi already pulled out of major Japanese wind projects due to soaring costs, and analysts warn actual output will only hit around 40% capacity versus 80-90% for nuclear. Critics call the 2035 deadline unrealistic, but the project feeds directly into Japan's 45 GW offshore wind target by 2040, signaling that floating wind tech is ready to scale.
🔐 Google Signs Classified Pentagon AI Deal
Google just signed a deal letting the Pentagon use its AI models for classified work and "any lawful government purpose." This comes despite over 600 employees urging CEO Sundar Pichai to reject the agreement, and marks a dramatic reversal from 2018 when Google pulled out of Project Maven after employee backlash. Google now joins xAI and OpenAI in having classified Pentagon AI deals, with terms that appear even more permissive than OpenAI's.
The contract includes language saying Google's AI "is not intended for" mass surveillance or autonomous weapons without human oversight, but legal experts say this wording is not legally binding. Notably, the deal also requires Google to adjust its AI safety filters at the government's request. This all follows Anthropic's public refusal to drop its red lines on those exact use cases, which led to the Pentagon declaring Anthropic a supply chain risk, a designation Anthropic is currently fighting in court.
Greg Brockman gives a rare inside account of OpenAI’s origins, the Sam Altman firing crisis, the shift away from a pure nonprofit model, and why he believes ChatGPT and AGI are now entering a much faster, more consequential phase.
Microsoft Lets OpenAI Off the Leash
The Takeaway
👉 OpenAI can now sell directly on AWS, Google Cloud, and Oracle, ending Microsoft's exclusive distribution and opening the enterprise floodgates ahead of a planned IPO.
👉 Microsoft trades exclusivity for financial certainty: no more AGI escape clause, a guaranteed 20% revenue share through 2030, and freedom from paying OpenAI for Azure resales.
👉 The deal removes the legal threat Microsoft held over Amazon's $50 billion OpenAI cloud agreement, clearing a massive overhang for both companies.
👉 Enterprise buyers are the real winners, gaining direct access to OpenAI models on their preferred cloud platform, which intensifies competition with Anthropic at a critical moment when both AI companies are racing toward public listings.
The most defining partnership in AI just got rewritten. Microsoft and OpenAI announced a sweeping renegotiation of their exclusive deal, effectively ending Microsoft's monopoly on selling OpenAI's models through the cloud. Starting now, OpenAI can serve its products across any cloud provider, including Amazon Web Services and Google Cloud. Amazon CEO Andy Jassy confirmed that OpenAI's models will be available directly on AWS in the coming weeks.
The numbers tell the story of a calculated trade: Microsoft retains a nonexclusive license to OpenAI's intellectual property through 2032 and keeps a guaranteed 20% revenue cut until 2030, now subject to a cap. In return, Microsoft no longer pays OpenAI a revenue share for reselling its models on Azure. The controversial AGI clause, which could have let OpenAI stop payments entirely, has been scrapped in favor of fixed timelines.
For OpenAI, this is about survival in the enterprise market. An internal memo from OpenAI's revenue chief admitted the partnership had limited the company's ability to reach enterprise customers where they already operate. With Anthropic capturing over 30% of enterprise API market share and racing toward a potential IPO, OpenAI needed multi-cloud access yesterday. For Microsoft, it means freedom to double down on its own AI models and Copilot strategy without shouldering all of OpenAI's infrastructure demands.
Why it matters: This deal reshapes how the largest enterprises access AI, giving AWS and Google Cloud customers direct paths to OpenAI's models for the first time. It also signals that the AI industry is maturing past exclusive lock-ins toward a multi-cloud, multi-model future, with massive implications for the Anthropic vs. OpenAI enterprise battle ahead of both companies' planned IPOs.
Sources:
🔗 https://www.reuters.com/technology/artificial-intelligence/microsoft-openai-change-terms-deal-so-startup-can-court-amazon-others-2026-04-27/
🔗 https://blogs.microsoft.com/blog/2026/04/27/the-next-phase-of-the-microsoft-openai-partnership/
Data-driven global scaling for 2026
Stop basing talent decisions on outdated figures. Deel’s 2026 Global Hiring Report provides salary benchmarks and growth trends from 150+ countries. Learn about the 283% rise in AI roles and how the talent landscape is shifting. Use these insights to optimize your spend and scale your team with total compliance.
NVIDIA dominates cumulative AI chip sales with its B200/B300 GPUs driving the bulk of growth from ~2M H100e in early 2024 to ~20M by Q4 2025, while Google TPUs, AMD, Amazon, and Huawei collectively remain a relatively small share of total installed compute capacity. No surprise that NVIDIA stocks keep on rising.
Build Your Own Humanoid Robot
The Raspberry Pi moment for humanoid robots is here. Menlo Research just open-sourced Asimov v1, a full-scale bipedal humanoid robot that anyone can build, train, and customize. Standing 1.2 meters tall with 25 actuated degrees of freedom, Asimov ships as a $15,000 DIY kit or can be self-sourced entirely from a public bill of materials. Think of it as the Arduino of humanoid robotics.
The project was born out of frustration: Menlo's engineers bought a Unitree G1, hit a knee issue, and waited two months for a single replacement part. That bottleneck inspired them to build something radically open. The robot's modular architecture lets builders swap legs, arms, and torso independently, while most structural parts are designed for MJF 3D printing. The full repository includes mechanical CAD, electrical wiring, a MuJoCo simulation model, and onboard software.
What makes this special is the philosophy. Menlo deliberately avoids vertical integration, sourcing hardware from an open supply chain so they can focus on the intelligence stack. While Boston Dynamics, Tesla, and Figure keep their robots locked down, Asimov flips the script, mirroring the open-source playbook that made AI models like Llama transformative for developers.
Gladly Connect Live '26. May 4–6 in Atlanta.
The room you want to be in. This is where CX leaders are tackling the hard AI questions and sharing what's actually working. For CX and ecommerce leaders. Atlanta, May 4–6. Space is limited — secure your spot now.
