Presented by
Good morning, AI enthusiasts. In 2017, OpenAI's president Greg Brockman wrote in his private diary: "This is the only chance we have to get out from Elon." Nine years later, that diary is now evidence in front of a nine-person jury in Oakland, with $150B on the line.
The trial that could rewrite OpenAI's origin story opened this week, and whether or not Musk wins, the internal emails and journals being read into the record are already changing how the world understands AI's founding moment.
In today's recap:
Musk vs. Altman: the AI trial of the decade
Google signs classified AI deal with the Pentagon
Generate a Complete Project Kit with Google Gemini
Stripe builds the payment layer for AI agents
4 new AI tools, prompts, and more
OPENAI
Elon Musk takes OpenAI to Federal court
Recaply: Elon Musk just took OpenAI to federal trial over its switch from nonprofit to for-profit, seeking $150B in damages as nine jurors took their seats in Oakland and opening statements began Tuesday.
Key details:
Musk says Altman and Brockman promised him a nonprofit and then turned it into a for-profit company in 2019, using his name and $38M in seed money to build what he calls a wealth machine for insiders.
OpenAI is now worth over $850B. Musk put in $38M from 2016 to 2020, when it was still a nonprofit. A potential $1T IPO could be at risk if Musk wins.
Brockman's personal diary is now court evidence. He wrote in fall 2017: "This is the only chance we have to get out from Elon." In the same entry, he asked himself: "Financially, what will take me to $1B?"
Jury selection ended Monday in Oakland. The nine jurors include a nurse and a painting company owner. Musk, Altman, and Microsoft CEO Satya Nadella are all expected to testify.
Why it matters: The private diaries and emails in evidence tell a messy story. Co-founders were asking about personal wealth while pitching Musk on a nonprofit mission. OpenAI says Musk wanted to be CEO and got angry when it didn't happen. Musk says he built it and it was taken from him. Nine jurors now have to decide who's right. Their verdict could reshape a company worth over $850B and derail a $1T IPO.
PRESENTED BY WISPR FLOW
Speak your PR description, bug reproduction, or Cursor prompt. Wispr Flow auto-tags file names, preserves variable names, and formats everything for immediate paste into GitHub, Jira, or your editor.
No re-typing. No context gaps. No mangled syntax. Works natively inside Cursor, Warp, and every IDE at the system level.
4x faster than typing. 89% of messages sent with zero edits. Used by engineering teams at OpenAI, Vercel, and Clay.
Google signs classified AI deal with the Pentagon
Recaply: Google just signed a classified AI deal with the Pentagon, allowing the US military to use its AI models for any lawful government purpose, joining OpenAI and xAI in military AI contracts despite 600+ employees asking the company to walk away.
Key details:
The deal lets the Pentagon run Google's AI on classified networks for mission planning and other national security work. It also requires Google to adjust its AI safety filters at the government's request.
The Pentagon paid up to $200M each to major AI labs in 2025. Google now joins OpenAI and xAI, making three of the biggest AI companies in the world active suppliers to the US military at the same time.
The contract says AI won't be used for mass surveillance or autonomous weapons without human oversight. But it also says Google can't veto lawful government decisions, according to reporting by The Information.
More than 600 Google employees signed a letter to CEO Sundar Pichai on Monday asking him to refuse. Alphabet dropped its ban on AI for weapons last year. Earlier in 2026, Anthropic refused to remove its safety guardrails and was labeled a Pentagon supply-chain risk.
Why it matters: In 2018, Google quit the Pentagon's Project Maven after employee protests. Eight years later, it's signing classified contracts that let the government adjust its safety filters. The reversal is hard to miss. Anthropic refused to remove its guardrails and was labeled a Pentagon supply-chain risk. Google looked at that outcome and chose compliance instead. Six hundred of its own employees are still asking Pichai to change course.
GUIDES
Generate a complete Project Kit with Google Gemini
Recaply: In this tutorial, you will learn how to generate a full project kit, including a brief in Google Docs, a budget tracker in Sheets, and a presentation deck in Slides, from a single Gemini conversation, eliminating the copy-paste cycle between tools.
Step-by-step:
Go to gemini.google.com, open a new chat, and describe your project in 2-3 sentences with key context: name, goal, target audience, and budget. The more specific this first prompt, the less back-and-forth you will need for later refinements.
Ask Gemini to generate the project brief first: "Create a project brief as a Google Doc with an executive summary, goals, timeline, and stakeholder list." Click the file icon in Gemini's response to open it in Drive or download as .docx.
Ask Gemini to build the budget tracker next: "Now create a Google Sheets budget tracker with columns for channel, description, projected spend, and actual spend, pre-filled with line items based on the brief." Export as .xlsx or open directly in Sheets.
Ask Gemini to generate the presentation: "Finally, create a Google Slides deck with 8 slides covering overview, goals, audience, channels, budget, timeline, success metrics, and next steps. Use the brief and budget we just created." Download as .pptx or open in Slides.
Refine all three files from the same chat thread: "Add a risk section to the Doc" or "Update the Slides budget slide with the Sheets numbers." Gemini retains full context, so corrections to one file can reference the others without repeating everything.
Pro tip: Before you start, paste a short style guide or a few sentences from an existing document into your first message. Say "Match the tone and formatting of this example:" followed by your sample. Gemini will apply that style across all three files.
STRIPE
Stripe builds the payment layer for AI agents
Stripe
Recaply: Stripe just announced 288 products at Sessions 2026, including a Machine Payments Protocol for AI agent transactions in stablecoins and fiat, an agent wallet for autonomous purchasing, and an Agentic Commerce Suite with Meta and Google integrations.
Key details:
The Machine Payments Protocol, co-authored with Tempo, lets AI agents buy things using stablecoins, cards, Klarna, or Affirm. Link's new agent wallet lets businesses set spending limits and track every purchase an agent makes.
Stripe made 288 announcements to 9,000+ builders at Sessions 2026. The Agentic Commerce Suite now connects to Meta's Facebook ads and Google's Gemini app, two of the most-used AI surfaces in the world.
Stripe also upgraded Radar to detect fraudulent AI agents trying to abuse payment systems, separate from the legitimate AI agents it wants on its platform. The same wave driving Stripe's agentic features is also creating new fraud, per the Sessions blog.
The Machine Payments Protocol, Checkout studio, and Stripe Database are all live now. The Google Gemini app integration via Universal Commerce Protocol is in preview.
Why it matters: The AI agent economy needs a financial layer, and Stripe just claimed it. When an agent books a service or pays for compute, it needs infrastructure like MPP. That infrastructure is live now, not as a research project. Stripe already connects buyers inside Meta ads and Google's Gemini. Every agentic workflow that handles money now has a Stripe-shaped hole in the middle of it.
TOOLS
Trending AI Tools
🎵 ElevenMusic - ElevenLabs' AI music generator for creating studio-grade tracks in any genre, cleared for nearly all commercial uses
⚙️ Meta Ads CLI - Meta's CLI tool for building and managing ad campaigns from the terminal
🤖 Mistral Medium 3.5 - Mistral's open-weights 128B reasoning and coding model
🍽️ Grader - AI CMO scanner for restaurants that audits your website, Google profile, and local competitors for free
NEWS
What Matters in AI Right Now?
Google rolled out direct file creation in Gemini chat, letting users generate Google Docs, Sheets, Slides, PDFs, Word, Excel, and more from a single prompt without leaving the app. The feature supports 10+ formats and is available globally to all Gemini users starting April 29.
Mistral AI released Medium 3.5, a 128B open-weights model with a 256k context window that merges instruction-following, reasoning, and coding in one set of weights under a modified MIT license. It scores 77.6% on SWE-Bench Verified, runs self-hosted on as few as four GPUs, and replaces Devstral 2 as the default model in Mistral Vibe.
ElevenLabs launched Eleven Music, an AI platform for generating studio-grade music from natural language prompts, with control over genre, style, structure, and lyrics across multiple languages. The platform is cleared for nearly all commercial uses including film, advertising, and gaming, with public API access available today.
Cursor introduced its SDK in public beta, letting developers build and deploy programmatic coding agents using the same runtime, harness, and models that power the Cursor app.
Owner launched Grader, billed as the world's first AI CMO for restaurants, which scans a restaurant's website, Google profile, reviews, and local competitors to surface what's hurting its online visibility. The tool launches as Owner marks $1B in total sales driven for restaurant clients.
Meta released the Ads CLI, a command-line tool that lets developers and AI agents create, edit, and analyze ad campaigns directly from the terminal without writing custom authentication or pagination code. The CLI outputs JSON and tab-separated values, making it compatible with CI/CD pipelines and automated agent workflows.
Taylor Swift filed trademark applications covering two audio clips of her voice and a stage photo, in what a trademark attorney describes as a novel attempt to close the legal gap AI voice mimicry has created. Copyright protects recordings, but AI can now mimic a voice without copying any existing recording, leaving artists with no existing protection against deepfakes.
OpenAI published an internal post-mortem explaining why goblin and gremlin mentions in ChatGPT spiked 175% after GPT-5.1, tracing the source to a Nerdy personality reward signal that accidentally spread creature metaphors across all chat modes through RL transfer. The investigation shows how a single training signal can amplify unintended behaviors across an entire model.
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
