Get tomorrow's brief in your inbox
May 19, 2026
Today: Anthropic acquired Stainless for over $300M and locked out rivals from critical SDK infrastructure. OpenAI partnered with Dell to bring Codex into on-premises enterprise environments. Elon Musk lost his lawsuit against OpenAI, with jurors taking just two hours to dismiss his claims.
Anthropic announced Monday it has acquired Stainless, the SDK automation platform used by OpenAI, Google, Cloudflare, and dozens of other companies, for over $300 million. The acquisition removes a critical infrastructure supplier from competitors and gives Anthropic exclusive control over tools that automate the creation and maintenance of software development kits across Python, TypeScript, Kotlin, Go, and Java.
Here's what matters for your business. Stainless built software that converts API specifications into production-ready SDKs and automatically updates them as APIs change. This is particularly valuable if you're building AI agents that connect to external software. Those integrations just got harder for everyone except Anthropic. The company confirmed it will wind down all hosted Stainless products, including the SDK generator. Existing customers keep the SDKs they've already generated, but future updates and support are gone.
If you're using tools that depend on Stainless-generated SDKs from OpenAI, Google, or other providers, those SDKs still work today, but the automation pipeline behind them is shutting down. Plan for more manual maintenance work ahead, or evaluate alternatives. If you're an Anthropic customer building agents, you just got an advantage: SDK maintenance stays automated while your competitors scramble. This is a rare example of an acquisition that creates a real competitive moat in AI infrastructure, not just hype.
Read more: TechCrunch
OpenAI partnered with Dell to bring Codex into hybrid and on-premises deployments, connecting it with the Dell AI Data Platform and Dell AI Factory. This matters if you run infrastructure on-premises and want AI coding assistance that can access internal codebases, documentation, and business systems without sending data to the cloud.
Codex is already used by 4 million developers weekly for code review, test coverage, and incident response. The new partnership extends it into non-coding workflows like preparing reports, routing product feedback, qualifying leads, and coordinating work across business systems. If you're in a regulated industry (healthcare, finance, defense) where cloud-based AI tools are blocked by compliance requirements, this gives you a practical path to production AI agents.
The value here is control. You can deploy Codex closer to enterprise data with the governance large organizations demand. If you're already running Dell infrastructure, this is worth evaluating. If you're on AWS, Azure, or Google Cloud, watch for similar on-premises options from those providers.
Read more: OpenAI Blog
Amazon rolled out "Alexa Podcasts" to all U.S. users Monday. Ask Alexa Plus for a podcast on any topic, and it researches the request, generates an outline, and produces an episode with AI-generated host voices. You can adjust length, tone, and focus before generation. Episodes are saved in the Alexa app for replay.
The feature draws from 200 news publications Amazon partnered with, including Reuters, AP, Washington Post, Vox, and Politico. If you're a small business owner who needs to stay current on industry news but doesn't have time to read, this is a practical use case. Ask for a 15-minute podcast on "commercial real estate trends" or "new OSHA regulations" and listen during your commute.
The catch is reliability. AI-generated content can miss nuance or misrepresent complex topics. Use this for overviews and context, not legal or financial advice. It's similar to NotebookLM's podcast feature, which launched earlier, and Microsoft Edge's recent podcast capability. The AI podcast format is becoming table stakes.
Read more: TechCrunch | The Verge
PaddleOCR 3.5 now supports Transformers as an inference backend, making it easier to integrate OCR and document parsing into Hugging Face-centered stacks. This is relevant if you're building RAG (retrieval-augmented generation) systems, document AI workflows, or agent applications that need to extract structured data from PDFs, scanned documents, tables, charts, or screenshots.
The update lets you run PaddleOCR models (PP-OCRv5 and PaddleOCR-VL 1.5) with Transformers, while PaddleOCR continues managing the OCR pipeline behind the scenes. If you're already using Hugging Face tools, this removes integration friction. You can configure backend options like dtype, device placement, and attention implementation through engine_config for your hardware.
If you're processing invoices, contracts, or receipts at scale, OCR accuracy is critical. PaddleOCR is competitive with commercial options and now easier to deploy. Try the live demo on Hugging Face Spaces before committing development time.
Read more: Hugging Face Blog
Zapier published a comprehensive guide on business automation that's worth reading if you're still manually copying form submissions into your CRM or sending the same follow-up email for the 400th time. The piece breaks down task automation (eliminating individual repetitive chores) versus workflow automation (connecting multiple steps across apps) versus process automation (full end-to-end department workflows).
Here's the practical takeaway. Start with task automation for quick wins: automatically create Asana tasks when deals close in HubSpot, copy form submissions to Google Sheets, send Slack notifications when invoices are paid. These deliver immediate time savings without requiring process redesign. Once you've automated 5-10 tasks, look for patterns and connect them into workflows.
The common mistake is trying to automate everything at once. Pick one repetitive task that happens at least weekly, automate it with a tool like Zapier, measure the time saved, then move to the next one. If you're spending more than 5 hours per week on copy-paste work, automation pays for itself in the first month.
Read more: Zapier Blog
Zapier's lifecycle marketing guide explains how to send stage-specific messaging based on what customers just did, not what your campaign calendar scheduled. If you're running a SaaS product, service business, or e-commerce store, this is the difference between a 2% trial-to-paid conversion rate and a 15% conversion rate.
The framework breaks customer relationships into stages: awareness, engagement, activation, retention, and winback. Each stage gets different messaging. Someone who just signed up for a free trial should get onboarding tips, not a re-engagement discount. A customer who's been with you for three years should get loyalty perks, not beginner tutorials.
The article walks through automation setups for lifecycle campaigns. If you're using a CRM or email platform with automation capabilities (HubSpot, ActiveCampaign, Mailchimp), this is the playbook to follow. The ROI is measurable: lifecycle marketing increases customer lifetime value by 20-30% compared to one-size-fits-all campaigns.
Read more: Zapier Blog
A California jury unanimously dismissed Elon Musk's lawsuit against Sam Altman, Greg Brockman, OpenAI, and Microsoft on Monday, ruling that his claims were filed too late under the statute of limitations. Musk accused the defendants of "stealing a charity" by creating a for-profit affiliate of OpenAI. The jury found that any harms Musk suffered occurred before the legal deadline for filing.
The trial revealed a parade of executives and co-founders testifying under oath that Altman was, in various ways, unreliable. Former CTO Mira Murati testified Altman told her OpenAI's legal team had approved skipping a safety review for a model, which turned out to be false. Co-founder Ilya Sutskever spent over a year assembling a 52-page memo alleging a pattern of lying and pitting executives against each other, which led to Altman's brief removal as CEO in November 2023.
But Musk didn't come off better. Testimony showed he was obsessive about controlling OpenAI and took an "obviously unsafe and reckless" approach to achieving AGI, according to OpenAI's chief futurist Joshua Achiam. The verdict removes a major threat to OpenAI ahead of its reported IPO, but the broader takeaway is less encouraging: the most powerful people in AI seem incapable of dealing with each other honestly. If you're building AI tools into your business, bet on the technology, not the personalities running the labs.
Read more: TechCrunch | The Verge
Anthropic's lawsuit over its Department of Defense blacklisting moved to a federal appeals court in Washington, D.C., on Tuesday. The DOD declared Anthropic a supply chain risk in March after negotiations collapsed over the Pentagon's demand for unfettered access to Claude models across all lawful purposes. Anthropic wanted assurance its technology wouldn't be used for fully autonomous weapons or domestic mass surveillance.
Defense contractors must now certify they won't use Claude in DOD work. The appeals court denied Anthropic's request to temporarily block the designation in April but agreed to expedite the case since Anthropic "will likely suffer some irreparable harm" during litigation. The government argues Anthropic could "encode limitations" into its model, presenting an "untenable national-security risk." Anthropic counters that this notion is unsupported and violates constitutional procedures.
If you're a defense contractor or work with federal agencies, this affects your AI tool options. Claude was competitive with GPT-4 and Gemini for many enterprise use cases, but the DOD designation creates compliance risk. A separate case in San Francisco granted Anthropic a preliminary injunction, allowing non-DOD agencies to use Claude while litigation unfolds. Watch for the D.C. court's written opinion in the coming weeks.
Read more: CNBC
Global AI glasses shipments hit 8.7 million units in 2025, up more than 300% from the prior year, and analysts project 15 million units this year. Meta's Ray-Ban AI glasses launched in 2023, Google is building Android XR, Apple is expected to enter the market, and Samsung is reportedly unveiling AI-capable smart glasses in July. China's Huawei, Alibaba, and Xiaomi are also moving.
South Korean startup LetinAR, which builds the optical modules that make AI glasses wearable, just raised $18.5 million ahead of a planned 2027 IPO. The company's PinTILT technology delivers brighter images with better battery efficiency than existing waveguide or birdbath lens approaches. If you're in field service, logistics, or manufacturing, AI glasses that overlay information in your field of vision (turn-by-turn directions, inventory data, repair instructions) are becoming practical tools, not sci-fi concepts.
Don't buy yet. The market is fragmented, and standards are still forming. But if your business involves hands-free work environments, start evaluating use cases now so you're ready when enterprise-grade options ship in 2027-2028.
Read more: TechCrunch
LLM agent checklist. Before deploying an AI agent in your business, ask three questions: (1) Can I clearly define the goal and success criteria? (2) What tools does the agent need access to, and are those integrations reliable? (3) How will I monitor and correct failures? If you can't answer all three, you're not ready. AI agents work best for well-defined, repetitive tasks with clear verification steps, like lead enrichment, data entry, or report generation. Start small, measure results, and scale only what proves reliable. (Zapier Blog)
Agentic RAG for better AI answers. Standard RAG (retrieval-augmented generation) pulls information once and hands it to the AI. Agentic RAG lets the AI decide when to retrieve more information, cross-check sources, and iterate until the answer is complete. If you're building internal AI assistants that answer questions from scattered documentation (support tickets, knowledge bases, CRM notes), agentic RAG reduces hallucinations and improves accuracy. Tools like LangChain and LlamaIndex support this pattern. The tradeoff is higher API costs and slower responses, so use it for complex queries where accuracy matters more than speed. (Zapier Blog)
The Anthropic-Stainless acquisition is the first major AI infrastructure consolidation move that creates a real competitive advantage. Expect more acquisitions targeting the connective tissue between AI models and enterprise software. The OpenAI-Dell partnership signals on-premises AI deployment is becoming a priority for enterprises that can't or won't send data to the cloud. If you're in a regulated industry, this is your opening. Finally, the Musk-Altman trial revealed what many already suspected: the personalities running AI labs are unreliable. Build your AI strategy around the technology and your business needs, not the charisma or promises of lab founders.
Word count: 2,346