AI news for Tuesday, July 14, 2026

The Daily AI Espresso — the links the most-followed people in AI actually shared, curated every morning. Edited by Alexis · Live updates →

☕ Daily AI Espresso

July 14 · today the field's most-followed voices converged on one subject: what AI does to work.

Yoshua Bengio, Gillian Hadfield & 5 others
16 Nobel laureates say we have “only a few years” to prepare for AI

More than 200 economists and AI researchers, including 16 Nobel laureates and organizers Erik Brynjolfsson and Anton Korinek, signed an open letter Tuesday urging governments to start building policy for AI-driven labor disruption now. Their case for the urgency: earlier technological shifts gave societies decades to adjust, and this one “may give us only a few years.” wemustactnow.ai

Eileen Clancy & J. Rosenbaum
A Bronx hospital replaced 12 nurses with AI, months after a strike to stop exactly that

Montefiore laid off 12 utilization-review nurses over the weekend and handed their work, confirming that hospital care is medically necessary and covered by insurance, to AI software from Datavant. The union says the move breaks AI-protection language won in a 41-day strike, and warns that pulling licensed nurses out of insurance and discharge reviews puts patients at risk. theguardian.com

Ed Zitron, Karl Bode & 1 other
OpenAI's ad business is tracking to miss its own forecast by 90%

OpenAI reportedly projected $100 billion in advertising revenue by 2030, but eMarketer estimates the entire US chatbot ad market will be worth about $5.4 billion that year. For 2025, OpenAI penciled in $2.5 billion in ad revenue while eMarketer sees standalone chatbots clearing under $1 billion in the US. adweek.com

Sasha Costanza-Chock & 3 others
Palantir won a £330m NHS contract, then most trusts barely used it

The London Review of Books traces how Palantir landed the NHS Federated Data Platform despite no prior healthcare record, leaning on ex-officials and Peter Mandelson's firm to lobby, and finds the software slow and thinly used once installed. The title quotes an NHS data analyst in a 2024 meeting: “This looks absolutely rubbish.” lrb.co.uk

Mike Masnick & Karl Bode
US companies are quietly switching to cheaper Chinese AI models

Futurism reports firms including DoorDash, Airbnb and Siemens are adopting open models from DeepSeek, Z.ai and Moonshot to cut costs, and cites the agent startup Lindy dropping Anthropic for DeepSeek V4. One figure from the piece: the heaviest AI adopters are spending around $7,500 per employee per month. futurism.com

Ted Underwood & 2 others
Arvind Narayanan: the work that's left is judging AI, not out-typing it

In an ICML keynote written up this week, the AI Snake Oil author argues human effort shifts from building to evaluation as models get more capable, with knowledge workers becoming more like “crane operators” steering powerful tools than workers made redundant. His counter to the doom framing: there is no single capability milestone that unlocks the economic disruption, so how societies choose to deploy AI still decides the outcome. normaltech.ai


🤔 Might Have Missed

Two things that flew under the radar today.

Researchers asked computer-science students how AI makes them feel about their careers

A University of Toronto team interviewed 25 CS students and found many already see subfields like software engineering as exposed to AI, and are hedging by piling on AI skills or eyeing other fields entirely. It is a small, qualitative study, but it puts a human face on the day's macro warnings: the anxiety is landing on people picking their majors right now. arxiv.org

How Ai2 stopped its AI agent from making things up: give it a command line

Ai2's writeup of Shippy, a maritime-intelligence agent that tracks vessel behavior, lands on a practical fix: instead of letting the model build raw API calls, it reaches for a purpose-built CLI, so the tools stay predictable even when the model's choices are not. The determinism cut the malformed-query and subtle-data bugs that plagued early prototypes. allenai.org


🌎 In the Wild

What's trending in AI, from the app charts to the community feeds.

People are using AI to redecorate rooms and shoot product photos

A cluster of image apps jumped the US App Store charts this week. AI Home Decor climbed eight spots, Arvin's product-photo maker rose three, and two new AI-art apps cracked the top 20 in Graphics & Design. Looks like a lot of people quietly using AI to mock up a room or fake a clean studio shot. App Store

The Meta smart-glasses backlash is turning into public embarrassment

Futurism reports the reaction to camera glasses has gotten strong enough that some owners now hesitate to wear them in public, worried they will be read as creeps. That is a fast turn for a product Meta hoped would feel normal by now. futurism.com

“Do it for me” agent apps are climbing the consumer charts

Manus, an app that runs multi-step tasks for you, rose two spots on the US Productivity chart this week, and Chat AI (“ask agent anything”) climbed three. The agent idea the labs keep pitching is showing up as things regular people download. App Store


That's the shot. — Alexis · AI Weekly


AI attention this week
Most covered NVIDIA — in 7 of the last 20 issues · 18 tracked stories this week
Fastest riser xAI▲ +100% story volume vs last week
Dominant theme Tools & capabilities — 24 of 149 tracked stories this week
From the AI Weekly Index · numbers frozen at publish time.
Get this in your inbox every morning. The Daily AI Espresso — free, one shot, no refills needed.