Comfortable Friday. I’m again from trip and nonetheless getting caught up on the whole lot I missed. AI researchers transferring jobs is getting coated like NBA trades now, apparently.
Earlier than I get into this week’s difficulty, I need to ensure you take a look at my interview with Perplexity CEO Aravind Srinivas on Decoder this week. It’s an excellent deep dive on the primary subject of as we speak’s publication. Hold studying for a scoop on Substack and extra from this week in AI information.
From chatbots to browsers
Up to now, when most individuals consider the fashionable AI increase, they consider a chatbot like ChatGPT. Now, it’s changing into more and more clear that the online browser is the place the subsequent part of AI is taking form.
The reason being easy: the chatbots of as we speak don’t have entry to your on-line life like your browser does. That stage of context — learn and write entry to your e mail, your checking account, and many others. — is required if AI goes to turn out to be a instrument that truly goes off and does issues for you.
Two current product releases level to this pattern. The primary is OpenAI’s ChatGPT Agent, which makes use of a primary browser to surf the online in your behalf. The second is Comet, a desktop browser from Perplexity that takes it a step additional by permitting giant language fashions to entry logged-in websites and full duties in your behalf. (OpenAI is rumored to be planning its personal full-fledged browser.)
Neither ChatGPT Agent nor Comet works reliably for the time being, and entry to each is at the moment gated to costly subscription tiers as a result of increased compute prices required to run the reasoning fashions they necessitate. Maybe most frustratingly, each merchandise declare to do issues they will’t, not simply in advertising supplies, however within the precise product expertise.
ChatGPT Agent is a read-only browser expertise — it could’t entry a logged-in web site like Comet — and that severely limits its usefulness. It’s additionally very sluggish. My colleague Hayden Subject asked it to find a particular kind of lamp on Etsy, and ChatGPT Agent took 50 minutes to return again with a response. It additionally failed so as to add gadgets to her Etsy cart, regardless of claiming it had performed so.
Whereas Comet is nowhere close to as sluggish, I’ve had quite a few experiences with it claiming it has accomplished duties it hasn’t, or stating it could do one thing, solely to instantly inform me it could’t after I make a request. Its sidecar interface, which locations the AI assistant to the fitting of a webpage, is superb for read-only duties, comparable to summarizing a webpage or researching one thing particular I’m . However as I informed Perplexity CEO Aravind Srinivas on Decoder this week, the general expertise feels fairly brittle.
It’s simple to be a cynic and suppose the present state of merchandise like Comet is the perfect AI can do at finishing duties on the net. Or, you possibly can have a look at the previous couple of years of progress within the business and make the guess that the identical pattern line will proceed.
Throughout our chat this week, Srinivas informed me he’s “betting on progress in reasoning fashions to get us there.” OpenAI constructed a customized reasoning mannequin particularly for ChatGPT Agent that was educated on extra advanced, multi-step duties. (The mannequin has no public title and isn’t out there by way of an API.)
Even with the various limitations and bugs that exist as we speak, utilizing Comet for only a few days has satisfied me that the mainstream chatbot interface will merge with the browser. It already looks like taking a step again to merely immediate a chatbot versus interacting with a ChatGPT-like expertise that may see no matter web site I’m . Standalone chatbots definitely aren’t going away, particularly on smartphones, however the browser is what is going to unlock AI that truly looks like an agent.
- What might have been for Substack: Earlier than the publication platform raised the $100 million round it introduced this week, two sources inform me that Vice founder Shane Smith approached Substack’s co-founders about buying the corporate. It’s unclear how far the talks progressed, although Smith additionally mentioned the concept with potential monetary backers. Substack’s management rebuffed his takeover curiosity however advised he might spend money on the spherical they simply closed. It’s unclear if he did. Neither Smith nor Substack responded to my request for remark.
- The top of reverse acquihires? Whereas I used to be out on trip, it was fascinating to watch the intense backlash to the Windsurf/Google reverse acquihire. This sample, the place the founders of a buzzy AI startup parachute into the arms of Massive Tech and depart the remainder of their staff to select up the items, is nothing new. It’s an unlucky byproduct of the antitrust scrutiny on Massive Tech, which thus far appears to have found out purchase what it needs by abandoning a husk of a startup and calling its payouts “licensing charges.” However given how Cognition messaged its rescuing of Windsurf’s remaining staff (“each single worker is handled with respect and nicely taken care of on this transaction”), I ponder if the subsequent AI startup founder will suppose twice earlier than leaving their staff behind.
- Mira Murati’s new AI lab may have an enterprise angle. I really feel assured in that prediction after seeing who her financial backers are for her new lab, Considering Machines. ServiceNow and Cisco aren’t investing in a ChatGPT competitor. Given the extent of expertise she has managed to assemble, the business will probably be paying shut consideration to no matter “multimodal AI” product the staff releases within the coming months. Is there room for an additional Anthropic-like rival to OpenAI? We’re about to search out out.
- AI researchers can’t get US visas. NeurlPS, the premier AI analysis convention, has skilled such excessive attendance demand for this 12 months’s occasion in San Diego that they’ve added a second location in Mexico to accommodate roughly 500 extra folks. The convention’s announcement states that there have been “difficulties in acquiring journey visas” for attendees wishing to attend the primary US occasion. Yikes.
Some noteworthy profession strikes
- Zuckerberg’s new Superintelligence lab is getting significantly larger. This week noticed the addition of OpenAI’s Jason Wei and Hyung Gained Chung, which signifies that Meta has now poached 5 of OpenAI’s 21 “foundational contributors” to o1. Augustus Odena and Maxwell Nye, co-founders of the Adept AI startup that Amazon reverse acquihired to kickstart its AGI lab, additionally joined, along with Mark Lee and Tom Gunter from Apple. In the meantime, the whole staff behind the voice AI startup PlayAI has officially joined (some firms are nonetheless sufficiently small for Massive Tech to accumulate outright). And in what ought to be an ominous sign to everybody within the broader AI group at the moment present process DOGE-style interviews with Alexandr Wang’s new staff, VP of Product Connor Hayes has moved over to run Threads.
- Anthropic’s head of engineering, Brian Delahunty, joined Google Cloud to guide AI agent engineering. In the meantime, Boris Cherny and Cat Wu returned to Anthropic after an alarmingly temporary tenure in management roles at Cursor. Paul Smith can be leaving ServiceNow to be Anthropic’s first chief business officer.
- Reddit CMO Roxy Younger is leaving amid what seems to be a broader management reshuffling.
- Extra mind drain at Tesla: This time it’s Troy Jones, head of gross sales for North America.
- Astronomer CEO Andy Byron and HR chief Kristin Cabot (that couple from the Coldplay live performance) have been put on leave pending an inner investigation.
In the event you haven’t already, don’t forget to subscribe to The Verge, which incorporates unlimited access to Command Line and all of our reporting.
As at all times, I welcome your suggestions, particularly if in case you have ideas on this difficulty or a narrative concept to share. You possibly can reply right here or ping me securely on Signal.