An OpenAI slowdown

PLUS: Agency and Agents

Together with

howdy, it’s Barsee again.

happy friday, AI family, and welcome back to AI Valley.

here are the biggest things worth knowing today:

  • OpenAI’s chief scientist wants everyone to slow down

  • Agency and Agents

  • Tesla Robotaxi becomes the #1 travel app, surpassing Uber

  • Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

VIKTOR

Courtesy: Viktor

Hampton is a 25-person founders network. Its co-founder was about to hire a data engineer at $200K, with an ops manager and a PM queued up behind that. He installed Viktor on April 12. Forty-four days later the team had 12 web apps live and 26 scheduled tasks running on their own. None of those hires are on the calendar.

Viktor is an AI employee that works in Slack and Microsoft Teams. Describe the tool you need in a message and it builds a real web app, with logins and a database, deployed at a live link. Then it keeps working. It connects to 3,200+ tools and runs the recurring work around the app, so the numbers are fresh when you open it.

Hampton's VP of Core Experience: "I don't have to create from scratch anymore. Just edit what it creates." 60,000+ teams have hired one.

*This is sponsored

THROUGH THE VALLEY

OpenAI’s chief scientist wants everyone to slow down - Jakub Pachocki wants the industry to stop treating max-speed scaling as the default. In an essay called “An Alien Mind,” he argues that today’s systems are grown more than designed: a product of huge optimization runs, not a mind anyone fully understands. That, for him, is the point. You cannot assume this kind of intelligence will inherit human principles on its own.

He dates the shift to mid-2023, when OpenAI’s RLSlow project first showed reasoning models could keep scaling. He now thinks that pace could continue into recursive self-improvement: AI doing more of the work of building better AI. Goal-following has improved. Value alignment (honesty, integrity, “love for humanity” when no one is watching) has not. The Hugging Face incident is his example.

The awkward part is OpenAI also thinks it needs more powerful models to defend against more powerful models. Cybersecurity is the obvious example, and Pachocki says future agents could potentially bargain with, trick or even blackmail people while pursuing their own goals.

And then there’s the timing: GPT-6 Astra just shipped with big jumps in coding, computer use, cyber and science. Pachocki says it’s better aligned than the last generation, and still not enough. His conclusion: no lab has solved alignment and monitoring well enough to keep flooring it. He wants voluntary slowdowns, mandated safety bars and eventually international coordination.

TRENDING TOOLS

  • Kilo - an open-source agentic engineering platform and CLI that works across roughly 500 AI models, giving developers the flexibility to choose their models instead of being locked into a single AI lab

  • Pluto - share your career story in a 10-minute conversation, and Pluto turns it into a living professional profile that recruiters and AI agents can search and trust. When a strong match appears, it makes a warm introduction

  • Gemini Immersive View - Gemini has launched a new Immersive View experience in the Students tab, letting users interactively explore topics ranging from dinosaurs to space and rockets

WHAT I'M CONSUMING

THE VALLEY GEMS

What’s trending on social today:

THAT’S ALL FOR TODAY

Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ New reader? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.