Good morning, {{ first_name | AI enthusiasts }}. The Industrial Revolution gave the world generations to catch up to its economic disruption. A new Stanford-organized statement argues the AI version could land on a much tighter schedule.
“We Must Act Now” puts 16 Nobel winners and 200+ signatories behind a single message: the next decade of AI could displace jobs at historic scale or lift living standards dramatically, and which way it breaks depends on plans that don't exist yet.
In today’s AI rundown:
Economists, researchers put AI’s job shock on the clock
Musk, Altman trade insults after Apple's OpenAI lawsuit
Find winning ad angles with Claude and Meta data
Claude's personality gets lost in translation
4 new AI tools, community workflows, and more
LATEST DEVELOPMENTS
AI & THE ECONOMY
Image source: Stanford Digital Economy Lab
The Rundown: 200+ AI researchers and economists, including 16 Nobel winners, just signed "We Must Act Now," a Stanford-organized statement telling governments to build safety nets and labor policy for an economic shift like the Industrial Revolution.
The details:
The three core points: AI will grow "radically more powerful" within 10 years, the shift could be the biggest and fastest ever, and prep must start now.
Several leaders from top AI labs signed, including Google's Jeff Dean, Anthropic co-founder Jack Clark, and OpenAI's Noam Brown.
UVA Economist Anton Korinek said, "Steam, electricity, and computers each gave societies decades to adapt; AI may give us only a few years."
Why it matters: The letter is light on specifics but heavy on signatories, with one main theme: change is coming much faster than the world’s current structures are built for. But with experts still split on questions like job losses and governments rarely known for moving quickly, getting big names behind the same warning might be the easy part.
TOGETHER WITH UNWRAP
The Rundown: If your team still sorts through customer feedback manually (or with a patchy mix of manual work and AI), there's a better way. Unwrap brings all your feedback across surveys, reviews, tickets, social comments etc. into one view, then uses AI and NLP to surface what actually needs your attention.
With Unwrap, you can:
Automatically categorize all your feedback in one unified view
Ask questions in plain language with Unwrap's AI Assistant or MCP
Get real-time alerts as trends emerge, plus custom reporting on your schedule
Walk into any meeting knowing exactly what your customers care about and what to prioritize
As a Rundown subscriber, you can try Unwrap for free and join teams like Oura, Stripe, Lululemon, Southwest Airlines, and Clay using it at scale!
ELON MUSK VS. SAM ALTMAN
Image source: X / The Rundown
The Rundown: Apple's lawsuit against OpenAI just kicked the Elon Musk vs Sam Altman keyboard battle into a new round, with Musk claiming Altman stole "all of Apple's phone technology," and Altman striking at SpaceX’s space data center pitch.
The details:
Musk spent the weekend on X saying Altman "takes scamming to a whole new level" and "might literally love scamming more than any human alive!"
Altman fired back at Musk, saying “homeboy you're the one selling public market investors on short-term space datacenters.”
Musk replied, "We start flying them next year," adding that Altman can "come see them if your parole officer approves."
Altman then joked that the best proof OpenAI's new GPT-5.6 Sol is the world's top model is "that Elon is obsessed with me again."
Why it matters: It feels like just yesterday that the blockbuster Musk vs. Altman trial had a judge ordering both sides to quit posting about each other. The verdict came and went in May, but the feud clearly never did. Watching two of the most powerful people in the world call each other names is an absurdity that just never gets old.
AI TRAINING
The Rundown: In this guide, you will learn how to use Claude Cowork to audit the angles in your Meta ads, creating an analysis of your ad data alongside your live Ad Library to create a report you can use to decide what to test next.
Step-by-step:
Export your Meta campaign data as a CSV. In Claude, create a meta-angle-review project folder, add the CSV, and copy your brand’s Meta Ad Library link
Prompt Claude to analyze your ad export, match it with active Meta Ad Library ads, flag unmatched rows, and identify the highest-performing creative angles
Ask Claude to turn the findings into a visual report with supporting evidence, key caveats, and three recommended creative tests
Rerun this report with exports from just Instagram or Facebook to isolate which angles work best in each placement
Pro tip: Schedule a rerun of the saved prompt after each new export and keep each report in a dated folder to compare placement-specific winners over time.
PRESENTED BY VANTA
The Rundown: GRC's reputation is shifting, and GRC Engineering is at the forefront. Ayoub Fandi, GRC Engineer at Lovable, and Justin Pagano, Sr. Director of GRC Engineering at Vanta, have a lot to say about Legacy GRC and what comes next. Consider this your front-row seat to the official mic drop moment.
Join this virtual event to learn:
GRC Engineering vs Legacy GRC
Why GRC Engineering goes beyond compliance automation
Practical ways to start automating, with lessons from Lovable and Vanta
Register to save your spot! Can't make the time? Register anyway, and receive a link to the recording.
AI RESEARCH
Image source: Anthropic
The Rundown: Anthropic published research putting numbers on Claude models' personalities, analyzing 309K user convos into character dials that find each model differing in temperament, and even changing personalities based on users’ language.
The details:
Researchers analyzed chats across three models and 20 languages, on four axes: warmth/rigor, candor/execution, depth/brevity, and deference/caution.
Personalities deviated across models, with Sonnet 4.6 acting warmer and brief, Opus 4.6 getting “straight to the point", and Opus 4.7 candid and cautious.
Language also changed outputs, with Dutch chats bringing more admissions of error, Hindi adding warmth, and English pressing deeper into details.
Anthropic wrote that it doesn't "yet understand why they vary, or whether that's desired," with uneven training data between languages as one suspect.
Why it matters: Anthropic keeps doing some of the most interesting research on AI's inner workings, and it just surfaced yet another weird quirk quietly shaping users' interactions with Claude. If this is true of Claude, it's likely true of every AI model, which potentially leaves personality blind spots the size of entire user bases.
QUICK HITS
🔎 Parallel Search Turbo - Low-latency search API for voice and chat agents
💼 ChatGPT Work - OAI's Codex-powered agent for everyday non-coding work
🎨 Reve 2.1 - Upgraded native-4K image model with element-level editing
🚀 Grok 4.5 - SpaceXAI and Cursor's strongest Grok model
Anthropic expanded Fable 5’s availability in Claude paid plans for a third time, with the model now accessible through July 19 before switching to usage credits.
Microsoft CEO Satya Nadella published an essay titled "The Reverse Information Paradox," arguing firms pay for AI twice, in money and in proprietary knowledge.
Y Combinator partner and entrepreneur Tom Blomfield announced that he is joining Anthropic as part of the company’s compute team.
Turing Award winner Richard Sutton co-founded Oak Lab, a new AI company building “animal-like intelligence that learns purely from its own experience”.
COMMUNITY
Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.
Today’s workflow comes from reader Nathan in Wānaka, New Zealand:
“We run a boutique agency and have always struggled to find the balance between admin and delivery. We get paid based on our results, so any time lost on admin affects our bottom line and our clients.
We've been building layers of our own vibe-coded apps and tools into literally every department — the latest being a workflow to turn meetings into actions. A Zoom meeting occurs, the transcript lands in the task inbox, and Haiku analyzes the recording to find tasks. A team member approves them, which triggers Opus to generate the task in Asana, including deliverables and estimated time based on historical data. We review tasks, check we haven't exceeded client budget, then push.
I'd estimate it's saving one employee 15 hours per week in admin time (at $180/hr) and costing about $12 in tokens.”
How do you use AI? Tell us here.
Read our last AI newsletter: Apple takes OpenAI to court
Read our last Tech newsletter: Netflix's binge model is backfiring
Read our last Robotics newsletter: NEO’s new humanlike hands
Today’s AI tool guide: Find winning ad angles with Claude and Meta data
RSVP to next workshop on July 17: Get knowledge work done with GPT 5.6
That's it for today!
See you soon,
Rowan, Zach, Shubham, and Jennifer — the humans behind The Rundown
