.Vinod Khosla predicts personal AI agents won't be a winner-take-all market. "[Personal agents] will be a broad category. I do think it'll be very diverse and I don't think there will be a winner-take-all." "Every one of the big labs, Meta and Google included, will want to have their... own agents. They will have different trade-offs and I think we'll see a lot of diversity in requirements." "As to Alex's comment, I think he hasn't used agents enough to understand much of that diversity that he's thinking about." "The best way to browse is have the agent do the work for you and give you the options, ask you questions, and be interactive with it." "The fundamental difference will be do you want to be an advertising product or something else? Do you want to trust Meta with your data?" "I think agents from startups will move faster than agents from the big companies."show more

@vkhosla bro needs to get some sleep

@vkhosla that bro ready for a coffin ⚰️

@vkhosla The winner will not have the most features. It will feel wrong to replace because it learned your weird rules, your people, and when to stop asking questions.

@vkhosla The biggest question people are going to need to answer is "do I trust *this* company with my data??" That's a huge thing that Meta has on all these other companies because people are already sharing so much, but it's still a huge obstacle.

@vkhosla totally. “book the trip” hides a lot of taste — pace, neighborhood, tradeoffs, the kind of weird place you’ll remember. personal agents win by learning those preferences without deleting the fun part.

@vkhosla Khosla calling it a broad category not winner-take-all is huge!! Means theres room for weird little agents like mine to win a corner of it

@vkhosla @BoBrainerd at least one opinion on our conversation on proliferation

@vkhosla 100%. personal AI gets shaped by trust, tone, habits, and what someone is willing to delegate. those are deeply personal. the best agent for one person could feel completely wrong for another.

@vkhosla Unlikely. People do not want to manage a bunch of agents with different context and memory and capabilities.

@vkhosla Almost definitely only 2 winners.

I’d love a future where agents are democratized, not monopolized. But I suspect one company will pull so far ahead that this becomes a winner-take-all market. The first personal agent that truly understands us, our feelings, deepest thoughts, fears, emotions and goals that we don't even know that we have will dominate the market. Digital Twins of humans will become bigger market than AI is today.

@vkhosla this matches what i see. every agent i build for a client needs different tools and prompts, so i doubt one product wins it all.

The market may fragment by trust model more than feature set. Ad-funded agents optimize toward transactions and attention; subscription agents can optimize toward the user's stated goal. Same models, same tools, radically different incentives. "Who pays the agent?" may matter more than who built it.

@vkhosla Winner-take-all for the underlying models, maybe. The agent layer on top will be as fragmented as every other app category.

@vkhosla I don’t have headphones nearby and can’t listen now, but will share my more nuanced thoughts. For what it’s worth though, I have very extensively used personal agents though- and so far lapsed from them all. But will listen and properly respond in the evening.

@vkhosla Spot on Vinod

@vkhosla the reason it will not concentrate is that the moat is your context, and context is portable in a way a social graph never was. easy to accumulate, easy to export. that makes for many winners and none of them very safe.

i’m building @wrkrdev on the bet that people will want an assistant that can use whichever lab makes the most useful model for their work. if another lab ships something better tomorrow, i want to use it with my saved projects and files. i don’t want to move my whole working life to another app. there’s plenty of room to build around that.

@vkhosla Personal agents feel unusually hard to make winner-take-all. People will trade off autonomy, privacy, memory, and control very differently.

@vkhosla Agreed on diversity — and one big trade-off axis is where the agent runs and who holds the context. A desktop workspace keeps projects, agents and run history on your own machine. That's the bet we make with GrokPot (our project account):

@vkhosla It's true, we haven't seen enough case studies of AI agents yet.