Happy Friday! After a refreshing vacation, I'm diving back into the tech world, where AI research is now being covered like the latest NBA trades!

The Shift: Chatbots to Browsers

When we think of AI, our minds often jump to chatbots like ChatGPT. However, a significant shift is underway as the web browser emerges as the new frontier for AI development. Why the transition? Simply put, chatbots currently lack access to the depth of your online activities, a feature that browsers naturally provide.

For AI to become a true assistant, it needs context—think access to your emails, bank details, and more. Enter two groundbreaking innovations: OpenAI's ChatGPT Agent, which browses the web on your behalf, and Comet, a desktop browser from Perplexity that lets AI interact with logged-in sites directly.

But there’s a catch. Right now, both tools have performance issues and are available mainly through pricey subscriptions due to the demanding computational resources they require.

Meager Performance and Disappointing Results

Unfortunately, both ChatGPT Agent and Comet haven't quite lived up to their promises. ChatGPT Agent is limited to read-only experiences, lacking the capability to engage with user accounts properly. I witnessed it taking an absurd 50 minutes to find a lamp on Etsy—only to fail at adding it to the cart!

Comet, while faster, is guilty of claiming it completed tasks it didn't actually manage. Its interface is great for basic tasks like summarizing content, but it feels fragile and unreliable.

What Lies Ahead for AI?

It can be tempting to dismiss these tools as the pinnacle of AI capabilities, but if history is any indication, we can expect continued advancements in reasoning models. In a recent interview with Perplexity CEO Aravind Srinivas, he expressed optimism about ongoing improvements in AI reasoning capabilities—indicating that we’re just scratching the surface.

Despite today's limitations, my brief experience with Comet convinces me that merging chatbot functionalities with browser capabilities will become essential for AI to truly serve as an intelligent agent.

Other AI Highlights This Week

Meanwhile, Substack recently secured $100 million, but not without some drama. Vice founder Shane Smith reportedly approached them with a takeover proposal that was swiftly rejected.

In a different vein, the recent backlash against the Windsurf/Google reverse acquihire highlights a troubling trend where startup founders abandon their teams for corporate giants, leaving other employees in a lurch.

Mira Murati is making waves with her new AI lab, Thinking Machines, attracting significant backers like ServiceNow and Cisco, indicating a strong enterprise angle in her initiatives.

AI Conference Attendance Issues

In an alarming twist, the prestigious NeurlPS conference seems to be experiencing visa travel difficulties for potential attendees, leading to additional venue arrangements in Mexico.

Career Movements to Watch

On the personnel front, Meta's Superintelligence lab continues to grow with talent from OpenAI, indicating aggressive recruitment trends in the field. Meanwhile, notable departures and shifts in leadership are happening across several tech companies, including Reddit and Tesla.

Stay tuned for more updates, and as always, I invite your feedback and story ideas!