In partnership with |  |
|
|
|
Anthropic ran Code with Claude SF yesterday, the biggest developer event of its year. The keynote announced a SpaceX partnership for 220,000+ NVIDIA GPUs, doubled Claude Code rate limits, and put GitHub, Datadog, Netflix, Vercel, Cursor, and Replit on stage to defend the platform thesis. |
And about an hour into all of that, Claude itself briefly went down for thousands of users (per DownDetector, spiking around 11:16AM ET). |
Truly nothing tells the audience "we're spending billions to fix our compute problem" like the compute problem briefly breaking the app. |
By the time the keynote pitched how the SpaceX deal buys 300+ MW of new capacity within the month, the service was back up. There's a reason they need that much! |
Here’s what happened in AI today: |
😺 Anthropic doubled Claude Code limits and signed a SpaceX compute deal 📰 Apple is talking to Intel and Samsung for US-made device chips 🍪 Adobe unveiled a productivity agent that turns PDFs into interactive AI experiences 🧩 One of these images is real, one is AI; vote below
|
A wild Neuron appearance! Our own Corey Noles joined NVIDIA's Nemotron 3 Nano Omni livestream this week with Wendell Wilson of Level1Techs for a working dev's take on the new multimodal open model: which model do you reach for and when, how to actually architect coordinated sub-agents, and where multimodal open weights unlock things text-only models can't. |
|
😺 Anthropic doubled Claude Code limits and locked up Elon's Memphis data center. |
As established, yesterday was Code with Claude SF, Anthropic's biggest developer day of the year. The opening keynote did what every Anthropic keynote does (announce capacity, demo a feature, parade big customers). |
But the thing that mattered to anyone who's actually used Claude Code in the past six weeks was buried in a single sentence: the five-hour rate limits just doubled. |
Here's what happened: |
Anthropic signed a partnership with SpaceX for full access to Colossus 1, SpaceX's 300+ MW Memphis data center (yes, the former xAI campus). 220,000+ NVIDIA GPUs come online "within the month." Claude Code's five-hour rate limits doubled for Pro, Max, Team, and seat-based Enterprise plans. The peak-hours Claude Code limit reduction is gone for Pro and Max. Opus API rate limits went up "considerably." Anthropic also said it's "expressed interest" in partnering with SpaceX on multi-gigawatt orbital AI compute (yes, datacenters in space).
|
 | Oh no, this couldn’t possibly have anything to do with the very unrelated legal case involving our mutual frenemy ALSO happening at the same time about now…” |
|
On the product side, three new Claude Managed Agents features dropped at the same keynote (per Simon Willison's live blog): |
Multi-agent orchestration (research preview) lets you spawn fleets of agents that hand off subtasks to each other. The on-stage demo built a hypothetical lunar drone-landing system out of three coordinating agents (a Commander, a Detector, and a Navigator). Outcomes (research preview) lets you define what success actually looks like, then let Claude iterate until the agent hits that bar instead of just stopping when the prompt ends. Dreaming (research preview) lets you run an agent overnight to inspect its prior sessions, find what it missed, and write itself new memory. The keynote demo: dreaming on a landing task spat out a descent-playbook.md ready to use next time.
|
For more on these, watch Claire Vo’s video where she breaks it all down. |
Why this matters: The last six weeks of Claude Code have been rough. There were public performance complaints, brief plan-tier confusion (Claude Code briefly removed from Pro before being restored within 24 hours), and a Fortune piece where Anthropic admitted to engineering missteps. |
The diagnosis everyone landed on: Anthropic ran out of compute. The reason? Dario said they grew by 80x last quarter. Daaaamn, Daniela! (that’s millennial for “i’m impressed). The doubled rate limits and SpaceX deal are the answer in two parts. |
Now, stack today's announcement against the 5GW Amazon deal, the 5GW Google/Broadcom agreement, the $30B Microsoft/NVIDIA partnership, the $50B Fluidstack investment, and the $200B five-year Google Cloud commitment The Information surfaced yesterday (Anthropic alone now reportedly accounts for more than 40% of Google Cloud's revenue backlog). |
Anthropic is basically signing every cloud and chip deal it can find, and now Elon's data center too. The same Elon whose Pentagon allies designated Anthropic a "supply chain risk" last quarter. Politics is downstream of gigawatts I guess? |
Well, If you’re AGI-pilled, meaning you believe “artificial general intelligence” is imminent and going to lead to a “fast take-off” scenario for technology, then everything is downstream of gigawatts… |
Want to watch it yourself? The keynote and every breakout session are on YouTube. Latent Space also has a nice recap of some other takeaways from Dario and Daniela’s talk and the other sessions. |
|
|
GTM Atlas, by Attio |
|
Your GTM motion is creative. The thinking behind it should be too. |
GTM Atlas is the ultimate resource on AI GTM for early-stage builders, providing foundational knowledge for teams navigating growth from scratch. Curated by Attio, the AI CRM, Atlas gives you: |
Systems thinking for every stage of the customer journey Frameworks and templates that scale with you Conversations with GTM operators at Clay, Lovable, and Vercel.
|
Mapped by operators. Curated by Attio. |
Explore now |
|
|
You probably use Codex or Claude Code for code. Austin Tedesco, head of growth at Every, spends roughly 80% of his working day inside Codex doing go-to-market plans, recruiting outreach, KPI tracking, and emails. |
His Codex Camp livestream with Dan Shipper yesterday was a 60-minute walkthrough of the exact setup; here's the short version. |
The trick is treating the desktop app (not the CLI) as an "agent management interface" for everything you do. Austin's setup: |
Make a folder per work area (his is called "every growth OS"). Folders give you persistent named chats per project, so you can ship a PR in one chat and draft a strategy doc in another without leaving the app. Connect the plug-ins for everything you live in: Gmail, Slack, Notion, Stripe, your data sources. Then drop in a markdown project file that explains what your business is, your goals, and how you like to work. Add reviewer agents. Austin forked compound engineering into "compound knowledge" so reviews check for strategic alignment and data accuracy, not security or front-end design. Run Austin's recommended starter prompt in a fresh chat inside that folder. Always do the final human review in the external app (Slack drafts, Gmail drafts), never inside Codex; the context switch is what keeps you honest before something goes to a real person.
|
Austin's verbatim starter prompt (the @ signs in the below refer to apps, so they require you to connect those apps as Plugins; you’d just replace those with w/e tools you use): |
Go take a look at the things I use the most (@Notion, @Slack, and @Gmail) and think of some automations that would help me with my work. For each one, explain what it does, when it should run, and which tool it touches. Ask me which ones look good before you build any of them.
|
Other plays Austin demoed live: synthesize meeting transcripts and Slack threads into a draft go-to-market plan, build a live KPI tracker in Notion that other agents can read, and find alums of a specific company who later got into AI (he used this exact play to surface a perfect L&D hire candidate in under a minute). |
Total AI beginner? Start here (goes with this video). |
Have a specific skill you want to learn? Request it here. |
|
|
|
|
|
📰 Around the Horn |
 | Put this to use in the trivia below! |
|
Apple is in talks with both Intel and Samsung to make its devices' main processors in the US, sending Intel up 6.4% pre-market on the Bloomberg report. The White House is reportedly drafting an executive order to pre-vet frontier AI models before release; we wrote up the case against it yesterday (TL;DR: the cure may very well be worse than Mythos). Boris Cherny, the head of Claude Code at Anthropic, livestreamed his actual everyday Claude Code workflow yesterday, the first time the creator of an AI coding tool has done a public, unfiltered "watch me drive" session.
|
|
|
One brand built 30+ landing pages through Viktor without a single developer. |
|
Each page mapped to a specific ad group. All deployed within hours. Viktor wrote the code and shipped every one from a Slack message. |
That same team has Viktor monitoring ad accounts across the portfolio and posting performance briefs before the day starts. One colleague. Always on. Across every account. |
5,700+ teams. 3,000+ integrations. |
Start free. $100 in credits → |
|
🧩 Thursday Trivia: |
You know the drill: one is AI, and one is real. Which is which? |
A. |
|
B. |
|
Which is AI, and which is real? Which is AI, and which is real? The answer is below, but place your vote to see how your guess everyone else (no cheating now!) |
|
|
|
|
|
Trivia answer: A is real (although my partner literally doesn’t believe me, it is a ACTUAL cat crying while its owner cuts onions), and B is AI. |
| That’s all for now. | | What'd you think of today's email? | |
|
|
P.P.S: Love the newsletter, but only want to get it once per week? Don’t unsubscribe—update your preferences here. |
