| |
Four days of testimony just wrapped in Elon Musk's federal lawsuit against Sam Altman and Greg Brockman. The greatest hits: Musk repeatedly insisted "you just can't steal a charity," then admitted xAI partly distilled OpenAI models ("standard industry practice"). He also called his own $38M early donation "I was a fool." Day 4 ended with Musk's family-office manager testifying about a $97.4B Musk-led bid to acquire OpenAI assets. Brockman testifies next; court resumes Monday. | This will make a great Aaron Sorkin movie in a few years. We vote Dane DeHaan for Musk (aged up a bit maybe??) and why not Jesse Eisenberg going two for two as Sam Altman? Come on yāall, you know Hollywood is just drooling to make this one! | Hereās what happened in AI today: | šŗ Starbucks rolled back automation; here's what becomes scarce next š° Anthropic crossed $1T this week, passing OpenAI for the first time š° OpenAI gated GPT-5.5-Cyber after AISI flagged record cyber capability šŖ Best Value AI 2026 ranks 37 LLMs by per-dollar performance š Top stories and tools from this week, all inside
| ā¦and a whole lot more that you can read about here. | Hey: Want to reach 700,000+ AI-hungry readers? Advertise with us!Ā | P.S: Love robots? Weāre starting a new robotics newsletter! Sign up early here. | | šŗ What gets scarce when AI does everything? |  | Read full deep dive here by clicking the image above |
| Starbucks just rolled back its automation push. CEO Brian Niccol told The Guardian that handwritten notes, ceramic mugs, and "the return of great seats" drove customer satisfaction; the company is hiring more baristas and pulling automation out of its busiest cafƩs. | In a must-read essay, University of Chicago economist Alex Imas argues this is the canary for the AI economy. Economics is the study of scarcity. When AI makes everything cheap, the real question is what becomes scarce. His answer: the human element. | Here's what happened: | AI drives the commodity sector toward zero marginal cost; people get richer in real terms. Spending shifts to things money can't easily replicate: relationships, status, exclusivity, provenance. In an Imas experiment, people paid roughly 2x for an identical item when they knew others would be excluded. AI-generated art got less than half the exclusivity premium of human-made art (21% vs 44%). A "relational sector" emerges where the human IS part of the product: teachers, nurses, therapists, craft brewers, hospitality, live performance.
| Why this matters:Ā History backs Imas. Farming was 40% of US jobs in 1900 and is now under 2%; nobody starved. A 2021 Econometrica paper Imas builds on found this income effect (people getting richer and shifting what they buy) drove over 75% of past sectoral shifts in jobs and spending. | His best line: "You don't need to be Picasso. You need to be the person whose involvement makes the product feel like it was made for someone, by someone." | Our take:Ā Imas's frame is the cleanest answer we've seen to "what jobs survive AGI?" But the worry is what wage distribution inside the relational sector looks like. Probably Spotify. According to Spotify's own 2026 Loud & Clear data, 80 top artists each generate over $10M/year while the 100,000th-ranked artist earned $7,300 in 2025; Spotify demonetized roughly 86% of all music last year. A few brilliant teachers and craft brewers will make fortunes; everyone else competes attached to platforms that take their cut. | I proposed a different way out: industrial-scale Data Foundries where your learning activity trains the AI to offset automating the commodity sector and you collectively own the data through co-ops that pay dividends. OpenAIās own principles suggest we need to consider new economic models that āensure that everyone can participate in the value creation in front of us.āĀ Getting paid dividends for your contribution to training data makes sense to me, but would that bankrupt the labs? At a minimum it would require better mechinterp to quantify and value individual contributionsā¦ blockchain could help with this IMO. | The relational sectorās day in the sun is coming either way; the political fight worth watching is over its distribution. | |
|
"How do I start building agents in Claude?" was the most-requested skill in our April reader inbox. And since we just did a livestream on Workspace Agents for OpenAI, itās about time we talk about this. For non-developers, Claude Cowork (the desktop app) is the right starting point; Claude Code is for after you're comfortable in a terminal. Jeff Su just shipped the canonical guide this week, and it boils down to three CLAUDE.md files that stack: | Root CLAUDE.mdĀ (under 300 lines): global voice rules and model defaults (Sonnet 80% of the time; Opus when a task has 3+ dependent steps). Workstation CLAUDE.md files: one per area of life (Email HQ, Newsletter HQ, Personal Finances). Rules stack on top of root automatically. Project subfolders: one-off work like a mortgage refinance or a Q4 trip. Each gets its own CLAUDE.md and MEMORY.md.
| Every folder also gets a MEMORY.md that Cowork writes to whenever you correct it. Corrections compound. Connect Gmail, Drive, and Calendar via Connectors, then type /schedule to make any task recurring. To set up your first workstation, paste this into Cowork: | Create a new workstation called "Email HQ" in /Documents/Cowork OS.
Inside it, create:
1. CLAUDE.md with email-specific rules (default greeting, sign-off, formality level).
2. An empty MEMORY.md (you'll add learnings to this over time).
3. A "Resources" subfolder.
Then use the Gmail connector to search my last 4 weeks of sent emails. Extract the patterns I use repeatedly: my opener, my sign-off, my formality level, common phrases I use with specific people. Save these as Editorial Rules inside CLAUDE.md.
When you're done, write one paragraph summarizing what you learned about my email voice so I can verify before we lock it in.
| Total AI beginner? Start here (goes with this video). | Have a specific skill you want to learn?Ā Request it here.Ā | | |
|
|
|
Did you know we have a podcast (The Neuron: AI Explained) where we talk to fascinating people in the industry who teach us how it actually works? Check it out: |  | Click to view these episodes on YouTube! |
| New episodes air every week on: Spotify | Apple Podcasts | YouTubeĀ | | š° Around the Horn | | SoftBank created Roze AI, a robotics company that deploys autonomous robots to build US AI data centers, eyeing a $100B IPO. Satya Nadella said Microsoft is ready to "exploit" the revised OpenAI deal granting royalty-free access to OpenAI models, IP, and agents through 2032. KKR raised $10B for Adam Selipsky (ex-AWS chief) to launch a new AI infrastructure company partnering with hyperscalers on data centers, power, and connectivity. Ken Underhill, friend of the pod and content lead at our sister publication Cybersecurity Insider sat down with Caleb Kinchlow to talk AIās impact on the cybersecurity profession; itās a fantastic, information dense 2 min watch.
| Want absolutely EVERYTHING that happened in AI this week? Click here! | |
š Sunday Special: Top of the Week | What a week. Monday opened with Microsoft and OpenAI rewriting their partnership; non-exclusive license, no more revenue share, the famous AGI clause quietly deleted. Tuesday, Anthropic crossed $1 trillion and passed OpenAI for the first time, simultaneously launching Claude for Creative Work with Adobe, Blender, Autodesk, and Ableton. Wednesday, Big Tech's Q1 capex hit $130B (Google literally ran out of capacity to sell). By Thursday, Anthropic was reportedly raising at $900B, Mythos was inside the NSA, and OpenAI hit 10GW of compute secured (years ahead of schedule). Oh, and there was a small goblin paper. | Top Stories of the Week | | Top Tools of the Week | | | |
| | | Thatās all for now. | | What'd you think of today's email? | |
|
| P.S: Before you goā¦ have you subscribed to our YouTube Channel? If not, can you? |  | Click the image to subscribe! |
| P.P.S: Love the newsletter, but only want to get it once per week? Donāt unsubscribeāupdate your preferences here. |
|
