The AI Newsletter Built for People Who Make Decisions
If you're still getting your AI news from X and random headlines, you're already behind.
Get expert analysis, market trends, founder insights, and signal over noise - so you can make better decisions without spending hours researching.
Today In Ai
3 things that happened while you were busy
1 SpaceX acquires Cursor for $60B, days after its record IPO.
SpaceX exercised its option to buy the AI coding tool in an all-stock deal first floated back in April. The two companies have been jointly training a new model that will ship inside Cursor and Grok Build soon. The acquisition closes as SpaceX stock rides a post-IPO wave - read all the details from the offering.
2 Databricks ships an agentic coworker built on your company's own data.
Genie One automates workplace tasks across apps, documents, and chats using Genie Ontology, a self-improving context layer that plugs into your organization's data. The goal is more accurate answers at lower cost, without the usual setup overhead. Try Genie One or see how it works.
3 Microsoft's Copilot Cowork goes generally available, at 30-40% lower cost per prompt than Claude Cowork.
Microsoft is rolling out its agentic workplace tool to all Microsoft 365 users globally, with model choice, usage-based billing, and cost controls baked in. For any org already inside the Microsoft stack, this is a real consideration. Watch Copilot Cowork in action or learn effective prompting strategies to get started.
Bonus reads: Anthropic's dire marketing may have worked too well, while DeepSeek just became China's most valuable AI startup. (Both unlocked via WSJ.)
From The Frontier
OpenAI lost $38.5 billion last year. It still plans to go public.
The numbers. Leaked audited financials, independently verified by the Financial Times, show OpenAI posted a $38.5 billion net loss in 2025. That is more than seven times the $5 billion it lost in 2024. Revenue grew from $3.7 billion to $13 billion, which sounds impressive until you notice the losses grew far faster.
Awkward timing. On the same day the financials leaked, a separate report found that ChatGPT's market share had dipped below 50% for the first time, squeezed by Gemini, Claude, and Grok. The pressure from below is real, and it is arriving just as OpenAI filed confidentially to go public.
The SpaceX precedent. None of this is automatically fatal, and investors have already shown they can overlook ugly financials for a sufficiently large story. SpaceX went public this week disclosing a $4.28 billion loss in Q1 2026 alone and roughly $41 billion in cumulative losses since founding. It still landed the largest IPO on record. The market bought the trajectory, not the balance sheet.
Sam's challenge. Elon Musk reframed SpaceX's story as a bet on the future of human civilization. Sam Altman needs a comparable narrative shift. OpenAI still runs one of the fastest-growing consumer platforms ever built, and its enterprise pipeline is substantial. But the question IPO investors will ask is simple: when does the bleeding stop? That answer is not yet in the filing.
What people are actually watching and sharing
Model cost map. Artificial Analysis released a cost vs. performance index for leading frontier models that has been making the rounds. It is a clean way to see which models are built for heavy lifting and which ones make more sense when you are watching your API bill.
Brilliant boring ad. Bland AI leaned into its own name and cast a notoriously boring office worker in its Series C announcement video. It cleared 1M views, which suggests that deadpan self-awareness is an underused marketing strategy in the AI space.
Sundar on career pressure. A two-minute clip from Sundar Pichai's Stanford commencement speech picked up 5,000 bookmarks. The framing is for graduates, but the message applies broadly to anyone rethinking how they work in an AI-accelerated world.
New coding king. A model called GLM-5.2 just dethroned Claude Fable 5 at the top of the Code Categories Arena leaderboard. It is now one of the strongest options available for coding and website building. Try it here.
Agents in the wild. Anthropic's Applied AI team published a blog post on building with Claude Managed Agents that has been widely shared. If you want a clear picture of how real companies are moving agents from prototype to production, this is worth fifteen minutes.
Prompt Station
This Claude prompt turns any product into a full objection map: 15 common buyer hesitations, the exact internal monologue behind each one, the sharpest one-line counter for your copy, and a ranked list of which objections are costing you the most conversions. It also surfaces the trust gaps and proof elements you might be missing entirely. Paste it into Claude or ChatGPT before your next landing page rewrite.
Decode exactly why customers don't buy
You are a world-class conversion copywriter and customer psychologist. Analyze [PRODUCT] and identify the 15 most common objections prospects have before buying. For each objection, explain the type of objection (price, trust, timing, competition, complexity, risk, need, etc.), write the exact thought the customer is likely having in their own words, and create a concise one-line counter-argument that can be used in sales copy. Also recommend the strongest form of proof to overcome the objection, such as a testimonial, case study, guarantee, statistic, comparison, or demo, and suggest the best place to address it on a landing page or within the funnel. Rank all objections by their likely impact on conversion, highlight the 5 most important objections, and rewrite those into persuasive landing-page-ready copy. Finally, identify any missing trust builders, guarantees, proof elements, or messaging opportunities that could eliminate these objections before they arise. Focus on uncovering both obvious and hidden objections, and think like a skeptical buyer actively looking for reasons not to purchase.Replace [PRODUCT] with whatever you are selling or promoting. Try: "a $297 online course on freelance writing", "a B2B SaaS tool for HR teams", or "a monthly subscription box for home bakers". For a sharper output, add one sentence after the main prompt describing your target customer, for example: "My buyer is a 35-year-old marketing manager at a mid-size e-commerce brand."
Not just another AI newsletter. MavSource aggregates updates from all major AI newsletters, podcasts, company news, AI labs, and hundreds of other sources — then summarizes what matters, analyzes emerging trends, and adds founder commentary. One 5-minute daily email. Free.