In partnership with
Welcome, AI enthusiasts
For decades the standard answer from economists was that every wave of automation ends with more jobs than it destroyed, so there was no reason to panic about this one. That answer just changed. More than 200 economists, sixteen Nobel winners among them, put their names to a warning that AI could reshape the economy faster than any technology in history, and some of the loudest skeptics signed it too. Let's dive in!
In today’s insights:
200 Economists Put the AI's Job Shock on Countdown
You're Paying to Work for AI
Anthropic Found Claude's Personality Varies by Model and Language
Read time: 4 minutes
LATEST DEVELOPMENTS
AI & THE ECONOMY
⏳ 200 Economists Put the AI's Job Shock on Countdown
Evolving AI: Nobel winners and AI lab bosses say the window to adapt runs in years, not decades.
Key Points:
More than 200 economists and AI researchers signed the statement released July 13 through Stanford's Digital Economy Lab, including 16 Nobel laureates.
Signers include OpenAI finance chief Sarah Friar, Google DeepMind's Jeff Dean and Anthropic co-founder Jack Clark.
Nobel laureate Daron Acemoglu, long the profession's most rigorous doubter of fast AI displacement, put his name to it.
Details:
Stanford's Digital Economy Lab pulled the statement together, and organizer Anton Korinek pins the urgency to raw speed. Steam, electricity and computers each gave societies decades to adjust, he argues, while AI may allow only a few years. The document names large-scale job displacement as the risk and major gains in living standards as the prize. It recommends no policy at all, because its authors say nobody yet holds the data to write one.
Why It Matters:
ADP's payroll data offers the closest thing to that missing picture. Its Stanford dashboard tracks 4.6 million workers and finds AI-exposed roles for 22 to 25 year olds shrinking over 4% a year while the wider market looks calm. The losses reach the rung below your own before anyone can see them.
Cut Lead Review From Hours To Minutes
Sign up for a free trial of Attio, the agentic CRM.
Ask Attio to build a daily workflow that surfaces the deals that need your attention today, like anything with a stage change, a recent reply, or a new signal in the last 24 hours.
Review your pipeline in Claude, synced live from Attio via MCP.
That's it.
NADELLA’S WARNING
🧠 You're Paying to Work for AI
Source: Futurism
Key Points:
Nadella coined the term in a Sunday post on X, calling it the Reverse Information Paradox.
He says the gap widens with every use, because the provider keeps learning about the customer while the customer learns nothing back.
His answer is for firms to own their own evals, memory and learning loop, so the intelligence they create while using AI stays theirs.
Details:
Kenneth Arrow won a Nobel for a paradox about information, that to sell it you must reveal it, and revealing it gives it away. Nadella says AI flips that onto the buyer, and the more a firm wants from a model, the more of itself it has to feed in. What escapes is not files or databases but what he calls intelligence exhaust: the prompts people write, the tools agents call, the evals that define good work and the corrections made when a model gets it wrong. Every correction is expertise a rival could never have bought, and it trains the next model the provider sells.
Why It Matters:
ChatGPT and Claude both learn from consumer chats by default, and corrections are the sharpest thing in them. Fixing an assistant feels like cleaning up after it, when it is closer to teaching a student who keeps the notes. The judgment you built over a career is what these models are missing, and every correction you type hands a little more of it to a system that owes you nothing for the lesson.
AlphaSignal's 8-person team automated 18 workflows. No developer.
AlphaSignal runs the most-read technical AI newsletter with an 8-person team. One salesperson installed Viktor for prospect research. Sixty-seven days later it runs 18 workflows across sales, ops, editorial, and finance: proposal builder, deal updates, competitor monitoring, P&L analytics. No developer hired.
Source: Anthropic
Evolving AI: Anthropic read 300,000 chats, found Claude behaving differently in each, and cannot explain why.
Key Points:
The study covered Sonnet 4.6, Opus 4.6 and Opus 4.7 across the 20 languages most used on Claude roughly 5,000 chats per pairing.
Sonnet 4.6 leaned warm and affirming while Opus 4.7 leaned cautious and critical, volunteering hard feedback.
Claude leaned warmest in Hindi and Arabic and most rigorous in English and Russian, where it asked users for evidence.
Details:
Anthropic clustered the 3,307 values Claude expresses down to 339, then into four axes covering deference, warmth, depth and candor. The axes explain 15% of the variation once the task and topic are controlled for. Opus 4.6 landed closest to brevity, getting straight to the point. Anthropic names the consequence directly, that the same business plan reviewed in Hindi and in Russian can return two different verdicts.
Why It Matters:
Sonnet 4.6 will call that plan a strong start where Opus 4.7 finds the hole in it, and the language on the keyboard tilts the answer again. People bring an AI their work and the decisions they are stuck on, and the temperament that answers them was never assigned. Most never see the setting.
👀 Our tip
OpenAI Build Week is running right now and anyone can enter. Ship a project with Codex by July 21 and judges score it on the idea, the build and the impact. Winners get cash, credits, DevDay passes and a spotlight from OpenAI. Beginners are explicitly welcome, and there are daily livestreams and Discord office hours to get unstuck.
Source: OpenAI Devs @X
👀 Click on the image you think is real
QUICK HITS
⚡ Williams signed a $5.34B deal with Blackstone, Apollo, and KKR to fund AI-driven power infrastructure projects.
🤖 Cloudflare launched Precursor, tracking full browser sessions instead of one-time checks to catch AI-driven bots.
🛡️ German defense AI startup Helsing raised $1.8B at an $18B valuation, Europe's biggest defense-tech round ever.
🤝 OpenAI's Deployment Company acquired Northslope, adding embedded engineers for enterprise AI builds.
📈 Trending AI Tools
🛠️ Base44 - Build fully-functional apps in minutes with just your words, no coding needed*.
📊 Databox MCP - Plugs your live business data into Claude and ChatGPT.
🧑💼 Ogment - AI coworker that lives in Slack, just tag @O.
🌐 BrowserAct - Browser layer that lets AI agents click, fill, and extract on real sites.
*partner link