Welcome, humans.
Atlassian laid off 1,600 people this week and the CEO went on record saying AI doesn't actually replace people at his company. They're doing it anyway to "reshape the skill mix."
Oracle is reportedly planning to cut up to 30,000 jobs to free up $8-10B for AI data centers. That's 18% of its workforce. The company's stock is down 54% from its September high, and Wall Street expects cash flow to stay negative until 2030.
Let us translate the business-speak for you, in case you missed it: We're firing you so we can afford to build the thing that might fire more of you. Is this what Lion King meant by the Circle of life?? I confused…
Now, in honor of Friday, here’s a fun meme to warm your hearts over the trash can fire that is this week:
ChatGPT literally talks like this; here’s another good one
Here's what happened in AI today:
😺 Businesses are picking sides in the AI platform war, and the data says Anthropic is winning.
📰 Google launched Ask Maps, the biggest Google Maps upgrade since the original.
📰 Lovable hit $400M ARR with vibe-coding, up 33% in a single month.
🍪 MagicPath turns any live website into a fully editable design in seconds.
🎓 How to set up your AI platform in 5 minutes (skills, connectors, and project files).
Want to reach 675,000 AI-hungry readers? Click here to advertise with us.
😺 The AI Platform War Has a Scoreboard Now. Anthropic Is Winning.
You probably have a pretty locked-in preference between ChatGPT, Claude, Gemini, and Grok at this point. You might even feel a little defensive about it.
Turns out businesses feel the same way. And for the first time, we have spending data that shows exactly who's winning.
Ramp's March 2026 AI Index dropped this week with a headline number that stopped people mid-scroll: Anthropic now wins roughly 70% of head-to-head matchups against OpenAI among businesses buying AI for the first time.
That’s a complete reversal from 2025.
Nearly one in four businesses on Ramp now pays for Anthropic (a year ago, it was one in 25).
OpenAI's adoption rate fell 1.5%, its largest single-month decline ever.
The “why” is where it gets interesting: It's not Opus’ benchmarks. Claude Code and OpenAI's Codex are roughly comparable products. Codex is arguably cheaper as Altman “floods the zone” by resetting usage limits. Meanwhile, Anthropic literally can't meet its own demand; every plan still has rate caps because they don't have enough compute (or, put another way, a pretty smart way to control costs imo).
But that’s what makes this meaningful: A company charging more for similar performance, while actively turning away revenue, is growing faster. In most enterprise markets, the cheaper product wins. Not here.
Ramp economist Ara Kharazian thinks the moat might be cultural.
The DoD backlash sharpened the contrast between the companies, and a "certain class of user noticed."
He floated a provocative comparison: choosing between OpenAI and Anthropic might become less like enterprise procurement and more like the green bubble / blue bubble distinction in iMessage. A signal of identity, not just technology.
We posit another theory, which we’ve said before but we’ll say again: No one is innovating in the GENERAL BUSINESS USER EXPERIENCE better than Anthropic right now. Put another way, that’s a bit crass: It’s the UX, stupid.
And yesterday, Anthropic made this case yet again with the launch of inline interactive visuals, letting Claude build charts, diagrams, and interactive visualizations right inside your conversation. It's on by default for all plans.
It sounds like a small feature, but combined with skills, connectors, and plugins, it makes Claude less of a text box and more of a workspace where information is both processed and presented.
Meanwhile, a16z's sixth Top 100 AI Apps report shows the platforms diverging fast. ChatGPT and Claude both have 200+ apps in their connector ecosystems, but only 11% overlap. ChatGPT is going super-app with shopping, travel, and food. Claude is stacking developer tools and financial data terminals. Two platforms, two philosophies, two ecosystems.
Here's the part that actually matters for your workflow: once you've configured your AI with skills, connectors, and project files (more on this below), switching costs compound. Every skill you install, every tool you connect, every memory the system builds... that's an infrastructure lock-in that’s expensive cognitively to unload. As a16z partner Olivia Moore put it: "context and memory compound."
What does all this mean? The model stopped being the bottleneck sometime last year. The system you build around it (what the industry now calls a "harness") is the real product. And to help you take the baby steps to start settings yours up, read below!
FROM OUR PARTNERS
Ready to revolutionize your enterprise AI? With Airia, you gain seamless AI management, from deployment to governance. Overcome adoption barriers and maximize efficiency—without compromising security or scalability. Airia empowers innovation, makes AI actionable, and helps your business achieve more with less.
🎓 AI Skill of the Day: Set up your AI platform in 15 minutes
The biggest unlock in AI right now has nothing to do with which model you use. It's whether you've spent 15 minutes configuring your setup.
Both Claude and ChatGPT now have full customization stacks (skills, connectors, plugins) that turn generic chatbots into personalized workspaces. The people getting dramatically more value configured one thing: a project file that teaches the AI how their specific work gets done.
Here's the fastest way to start:
Open your AI tool's settings menu. Connect ONE tool you use daily
(Google Calendar, Slack, or your CRM).
Then install ONE plugin that matches your job function (marketing, legal, finance, and data analysis are all covered).
Finally, write a short project file (CLAUDE.md or AGENTS.md) with three things: your role, your preferred output format, and one process you repeat weekly. If you only use the web version, the equivalent would be setting custom instructions on a project folder.
Phil Schmid (ex-Hugging Face) calls the project file "the most important concept in 2026." The model is the engine, but the file is the steering wheel. Most people are driving without one.
Want more tips like this? Check out our AI Skill of the Day Digest for this month.
Have a specific AI skill you want to learn? Request it here.
🍪 Treats to Try
MagicPath turns any website URL into a fully editable app design in seconds; paste the link, the AI imports the live page, and you can visually edit, iterate via chat, or extract components—free to try.
OpenJarvis is a personal on-device AI with five composable primitives (intelligence, inference engine, agents, tools/memory across 26+ channels, and self-improving loops) for CLI/browser/desktop use with all data staying on your machine; Stanford collab with John Hennessy (GitHub).
AgentMail gives your AI agents their own email address so they can sign up for services, verify via OTP, and manage inboxes autonomously—free to try.
DenchClaw wraps OpenClaw into a fully managed framework for CRM automation, outreach agents, and local productivity workflows. Free and open source.
Citecat semantically searches 10M+ papers, chats with any paper for methodology insights, annotates PDFs with AI explanations, and compiles LaTeX with automatic citations—free to try.
📰 Around the Horn
Google launched Ask Maps, a Gemini-powered conversational feature for complex real-world questions with personalized recommendations, plus upgraded Immersive Navigation with richer 3D views.
Bumble will launch an AI dating assistant called "Bee" that helps users write profiles, generate conversation starters, suggest date ideas, and plan outings; meanwhile, Tinder is trying to reverse its apparent user decline by launching an Events tab for IRL curated events, video speed-dating, AI Chemistry and Learning Mode, and a "Does This Bother You?" language model to keep users safe.
Lovable hit $400M annual recurring revenue, up 33% in a single month, with its platform that turns natural language descriptions into production-ready applications.
Perplexity launched a full-stack API platform for building agents with one key, including an Agent API for multi-step orchestration, real-time Search API (SOTA on SimpleQA/SEAL), and Embeddings API (docs).
Cursor shipped 30+ new Marketplace plugins (Atlassian, Datadog, GitLab, Glean, HuggingFace) plus MCP Apps bringing interactive UIs directly into agent chats. Revenue reportedly surpassed $2B annually; check out CursorBench too, which is how they eval agent performance in Cursor (they found GPT 5.4 performs the best of those tested using less than 16K tokens (less than Opus).
Ukraine's Ministry of Defense opened millions of annotated combat frames for partners to train AI for autonomous drones, a world first.
Writer Julia Angwin sued Grammarly for allegedly using her copyrighted work to train its AI without consent (case).
FROM OUR PARTNERS
Privacy-first email. Built for real protection.
End-to-end encrypted, ad-free, and open-source. Proton Mail protects your inbox with zero data tracking.
💡 Intelligent Insights
@Kimmonismus on X shared some insights from a NYT deep dive on Anthropic this week which they say contained several buried revelations that are worth reading between the lines:
Model releases are now separated by weeks, not months. Some 70-90% of the code used in developing future models is written by Claude itself. Staff internally believe 2026-2030 is where "all the most important things happen."
And Dario Amodei warned AI could displace half of entry-level white collar jobs in 1-5 years, urging the industry to stop "sugar-coating" the workforce impact.
Perhaps the most striking detail: some employees have begun to question whether Anthropic has crept to the cusp of recursive self-improvement, the moment where AI models meaningfully accelerate their own development. Some external experts believe fully automated AI research could be as little as a year away.
Dwarkesh Patel argues the Department of War is making a huge mistake threatening to destroy Anthropic over its red lines on mass surveillance and autonomous weapons. His point: AI structurally favors authoritarian applications (cheap ubiquitous camera monitoring is trivial to scale), so the US should be preserving the independence of companies that set norms against misuse, not compelling them to serve without limits like China would.
Todd Saunders makes the case that the token cost to build a production feature is now literally lower than the cost of a 30-minute meeting to discuss whether to build it. His conclusion: stop planning anything that can be empirically tested. Build it in 2 hours, measure with real customers, kill or keep. The "planning industrial complex" is dead.
François Chollet argues the core bottleneck in current AI is that every technique still relies on pattern memorization and retrieval, requiring humans to decide which patterns to memorize and what goals to pursue. Today's models remain a reflection of human cognition, not an autonomous thing. (This is the guy who created the ARC benchmark, so when he says models haven't crossed the threshold, it carries weight.)
A Cat’s Commentary
 | That’s all for now.
|
