Good morning. It's Friday, June 5, and we're covering Anthropic's warning that AI may soon help build its own successors, the debate over AI's impact on jobs, and bots overtaking humans in web traffic.

Know someone who'd love this? Forward it their way. First time reading? Sign up here to keep up with the future of tech.

Brian Chesky is in early-stage funding for a new AI lab focused on developing models with a richer user-interaction and design layer, while remaining Airbnb's CEO and explicitly declining the chief-executive role at the new venture. The bet reflects Chesky's design-school instinct, dating back to Airbnb's founding nearly 20 years ago, that travel and e-commerce AI requires a real UI, not the text-based chatbots that OpenAI and Anthropic have popularized (Airbnb is notably the only major online travel player that has not built a ChatGPT plug-in, while Expedia and Booking have).

The shift arrives sooner than expected, with automated agents now generating more internet traffic than humans for the first time, according to Cloudflare CEO Matthew Prince. Prince says the crossover was previously forecast for late 2027, then early 2027, but accelerating agentic AI activity has dramatically pulled the timeline forward. The milestone reflects rapid growth in AI-powered bots that browse, retrieve, and interact with online services autonomously. The comment adds to broader evidence that AI agents are becoming a significant force in how traffic moves across the web.

While Silicon Valley races toward humanoid robots, Hello Robot is finding traction with a wheeled assistant designed for safety, accessibility, and real-world use. The company’s $30,000 Stretch 4 robot is already deployed in homes, research labs, and enterprises, helping users with disabilities perform everyday tasks while generating valuable operational data. CEO Aaron Edsinger expects to build 200–300 units this year, with the first production run sold out. Researchers say the robot’s cautious design could make it a practical platform for collecting the real-world data needed to advance physical AI.

OpenAI’s biggest internal user now consumes roughly 100 billion tokens per month, yet CEO Sam Altman says someone outside the company spends even more. Speaking at an enterprise event, Altman highlighted how AI usage has exploded, with 100,000 monthly tokens once making someone the world’s top user. OpenAI employees have reportedly logged hundreds of billions of tokens monthly, including one worker who used 210 billion tokens in a week. As usage surges, companies including Amazon and Uber are moving to rein in AI spending amid growing cost concerns.

FRIDAY FACTS

A team of British mathematicians worked out the math behind modern encryption years before the team that got the credit, and the public didn't find out until decades later. The full story is at the bottom of today's newsletter.

Fastino’s Pioneer Agent automates AI fine-tuning, enabling open-source models to self-improve and outperform larger models at lower cost.

Recently, Alexandria Ocasio-Cortez (AOC) held up a jar of brown, dirty water on the floor of Congress and said it was the drinking water of a community living next to a Meta data center in Georgia. She was pressing an EPA official about it during a May 2026 hearing, and the moment went viral. It did something I think we’ve badly needed: it forced a real conversation about what AI and data centers are doing to the environment, and to environmental health, the messier question of how the environment around us affects our bodies.

Her demonstration was powerful. It was also not quite right. → Read the full article here.

Anthropic argues that AI is already speeding up the creation of more capable AI systems and could eventually reach a point where models help design and improve their own successors, a process known as recursive self-improvement. The company says Claude now authors more than 80% of the code merged into Anthropic’s codebase and that engineers are shipping roughly eight times more code than they did in 2024, while AI performance on coding and long-duration task benchmarks continues to improve rapidly.

Internal data also suggests Claude is increasingly capable of handling open-ended engineering and research work, though Anthropic says humans still retain an advantage in setting goals, exercising judgment, and determining which problems matter most. The report outlines several possible futures, ranging from slower progress constrained by technical or economic bottlenecks to AI systems eventually driving much of their own development. → Read the full article here.

Apollo Global Management chief economist Torsten Sløk argues there is “zero evidence” that AI is causing job losses, citing ADP data showing U.S. private employers added nearly 110,000 jobs in April. Sløk says companies are hiring workers to implement AI systems, and that growing demand for AI-related products and services is boosting employment rather than reducing it.

His comments contrast sharply with a growing number of layoffs linked to AI adoption, including recent cuts at Wix, Block, and other major technology firms that have explicitly cited AI-driven efficiency gains. → Read the full article here.

O’Leary Scales Back Utah AI Hub: Kevin O’Leary will cut his proposed 40,000-acre Utah AI data center project by roughly half after backlash.

Quantinuum Jumps in Nasdaq Debut: Honeywell’s quantum computing unit rose 13.3% on its first trading day, reaching a $17.63 billion valuation.

Trump Team Eyes AI Doctors (Paywall): The administration is exploring AI tools that diagnose illnesses and prescribe drugs despite physician concerns.

Data Center Backlash Surges: 71% of Americans oppose nearby data centers, up from 42% in nine months, citing power and water concerns.

Here’s how you voted:

🟨⬜️⬜️⬜️⬜️⬜️ Original articles and essays (12%)
⬜️⬜️⬜️⬜️⬜️⬜️ Our take on the big stories (4%)
🟩🟩🟩🟩🟩🟩 Educational / How-to content (40%)
🟨🟨⬜️⬜️⬜️⬜️ Tool breakdowns (16%)
🟨🟨🟨🟨⬜️⬜️ Don't change it, I just want to stay updated (28%)

FRIDAY FACTS

In 1969, James Ellis at the UK's Government Communications Headquarters (GCHQ) proposed an idea he called "non-secret encryption," a scheme where the encryption key could be published openly while the decryption key stayed private.

In 1973, his colleague Clifford Cocks worked out a concrete way to do it using prime factorization. A year later, GCHQ mathematician Malcolm Williamson independently devised what the world now knows as the Diffie-Hellman key exchange. All of it was stamped Secret.

Four years after Cocks's work, in 1977, MIT researchers Ron Rivest, Adi Shamir, and Leonard Adleman published essentially the same construction in the open. It became RSA, the algorithm that still underpins much of secure communication on the modern web.

GCHQ's role wasn't declassified until December 1997, weeks after James Ellis had died. He never received public recognition for the idea in his lifetime.

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team