From $0 to $10 Billion in Months: Silicon Valley's AI Assistant Gold Rush Is On (and What It Means for Travel)

I've developed a bit of an obsession with an AI assistant called Instinct.

I don't open an app, write elaborate prompts or set up workflows. I send it a text or email, the same way I'd communicate with an actual assistant.

And that's what makes it so interesting.

It can access my connected accounts, remember previous conversations, manage tasks and work through requests without needing me to explain every step.

I'm still testing what it can handle, but the experience already feels different from the AI tools I've spent the past few years using.

Apparently, I'm in good company.

In September, Instinct raised $1 billion at a $10 billion valuation. A month earlier, investors valued the company at $2.5 billion. The startup had launched its invitation-only assistant in August, employed just 14 people and hadn't introduced a subscription price.

The founder was 23.

If you wanted an indication of where Silicon Valley thinks the next AI opportunity lies, that funding round makes a compelling case.

For the past few years, the technology industry has been obsessed with building smarter models. OpenAI, Anthropic, Google and Meta have competed over benchmarks, reasoning capabilities and increasingly ambitious model releases.

Now they're fighting over something considerably more personal.

They want to become the assistant you trust to run parts of your life.

And with that comes access to your calendar, inbox, preferences, contacts and eventually your money.

Zuckerberg is spending billions. Musk wants in. And the startups aren't backing down.

Meta's entrance into the personal AI assistant market has been difficult to ignore.

In June 2025, Mark Zuckerberg committed approximately $14.3 billion to acquire a 49% stake in Scale AI and recruit its 28-year-old founder, Alexandr Wang, to help lead Meta's superintelligence efforts.

Meta followed with a recruitment campaign targeting researchers from rival AI companies. Former OpenAI researcher Shengjia Zhao became chief scientist.

The company introduced its Muse Spark model in April 2026 and launched its personal AI agent, Muse, in September.

Within its first 13 days, Muse reportedly attracted 2.5 million downloads. By the third week, the figure had reportedly climbed past 3 million.

Those are impressive numbers for a company that had been struggling to establish itself among the leaders in generative AI.

Meta also has something most startups would kill for: distribution.

Billions of people already communicate through WhatsApp, Instagram, Facebook and Messenger. Putting a capable personal assistant inside those environments eliminates much of the work involved in convincing consumers to adopt another application.

Then there's Elon Musk.

His xAI has been developing Grok Bot, which takes a slightly different approach to assistance. Users can create multiple specialized agents and have them work together.

Think of having a researcher, analyst and project coordinator who share information and complete assignments.

I've just built my own small team of expert Grok Bots, and I'm pretty impressed so far.

The latest version supports agents with persistent computers, shared context and the ability to execute workflows across connected tools. They can also participate in workplace conversations through Slack and Microsoft Teams.

It's early days for my own experiment. But the idea of building a team of specialists that I can assign work to is worth exploring.

I certainly wouldn't count Musk out of this competition.

Meanwhile, Anthropic has introduced Claude Duet, a proactive assistant with access to more than 6,000 connected applications, and OpenAI has launched Dots, its approach to persistent agents that can work on ongoing projects.

All of this has happened within months.

The companies are spending enormous amounts of money to build and distribute these products. The question now is which ones people will want to keep using.

The personal assistant is becoming more useful than another chatbot

For years, the most common way to use AI was to ask a question and wait for an answer.

That can be useful. I've used ChatGPT and Claude extensively for research, writing, strategy and countless business tasks.

But the newer assistants can go further.

You can ask one to find a dinner reservation, organize a business trip or follow up on an unanswered email. It can access connected applications, navigate websites and carry out actions with your permission.

Some can monitor ongoing commitments and identify things that need attention.

One early Instinct user described an assistant finding an outstanding document that needed signing without being asked to look for it. Another reported using Instinct to change flights and secure a difficult restaurant reservation through a connected OpenTable account.

The assistant identified when reservations became available, completed the booking and added the details to the calendar.

There's another clever development: agents can coordinate with one another.

Instead of sending messages back and forth to arrange dinner with friends, connected assistants can exchange scheduling information and work out the logistics.

That begins to look a lot like having human administrative support.

Of course, the software still makes mistakes. Websites block automated interactions, instructions get misunderstood and occasional tasks require human intervention.

But enough of these products are working well enough to make people curious about what they can delegate next.

The 10 AI personal assistants I'm using, testing and watching

I've spent time exploring several products and following the developments of others. Some are personal concierges. Others resemble executive assistants or even small teams of AI employees.

Here's where I see the market today.

These are my current interests and assessments, rather than a ranking based on identical hands-on tests.

Instinct: The one I keep coming back to

The best thing about Instinct is how little I have to think about using it.

I send a message. It gets to work.

For someone who spends a substantial part of the day moving between clients, emails, meetings and projects, that simplicity has a lot of appeal.

Instinct can connect with calendars, email and other services. It can handle reservations, purchases and personal coordination, and it can remember context from previous interactions.

The travel element caught my attention.

According to its founder, more than 50% of Instinct's transactions involve travel. The company also reported approaching $1 billion in annualized transaction volume, although that figure comes from its own reporting.

Given my background in hospitality and travel technology, I'm interested in seeing how well it handles bookings, itinerary changes and all the little complications that come with travel.

And the company is letting users experiment for free during its invitation-only beta.

That's an interesting business decision for a startup with a $10 billion valuation and no established subscription model.

For now, I'm enjoying discovering what happens when I give it a request that would normally require several emails or a few browser tabs.

Catch AI: A personal executive assistant for $99 a month

Catch has a narrower focus, which I appreciate.

It handles the administrative work that most executives would love to hand over: coordinating meetings, managing calendars, following up on email and making phone calls.

That last capability gets my attention.

Catch can call businesses, navigate phone systems and report back. Think restaurant reservations, appointment changes or the kind of customer service task that requires sitting on hold.

I've been using Catch for executive administration, and I can see why its $99 monthly price would appeal to consultants and small business owners.

It addresses a familiar business expense with a product that people can understand.

The challenge for Catch will be maintaining that focus while larger AI companies expand into the same activities.

Meta Muse: My growing favorite for personal life

Muse has become one of the more interesting options for my personal life.

It can organize travel, research purchases, make reservations and manage tasks across connected applications.

You can communicate through its app or WhatsApp, and it can continue working after you close the conversation.

Meta also offers a free tier.

I've been drawn to how conversational the experience feels. I haven't explored Muse for professional work yet, but I'm interested in what it can handle outside the office.

Meta has faced questions about privacy and human involvement in some agent-supported telephone calls during testing. Those concerns deserve attention when users are granting access to private accounts.

Still, it's impressive how quickly Meta has turned its AI investment into a consumer product with substantial early adoption.

Claude Duet: The one I suspect could become the leader

Claude Duet may be the most ambitious professional assistant in this group.

Anthropic introduced it in October with access to more than 6,000 applications and tools.

Duet can prepare briefings, monitor work, coordinate information across systems and take on assignments that extend into financial analysis, reporting and software development.

Its ability to recognize outstanding work and act on ongoing responsibilities interests me.

Imagine an assistant that understands your current projects, reviews the relevant documents and prepares the next piece of work without needing a lengthy explanation.

That's the experience Anthropic is trying to deliver.

Duet remains in research preview, and the company has acknowledged limitations in how the assistant interprets priorities and handles certain tasks.

But Anthropic has established a strong reputation for complex professional work. With Duet, it has an opportunity to move from helping people complete assignments to taking responsibility for more of the execution.

I'll be watching this one closely.

Grok Bot: My new team of specialists

I've just started building a small team inside Grok Bot.

The concept is appealing because you can give individual bots specific responsibilities, instructions and tools.

A research bot can investigate a topic and share its findings with another agent. A second bot can develop a deliverable. A third can review the work.

They can coordinate rather than requiring you to manage each handoff.

Grok Bot also supports recurring workflows and persistent computers, so agents can continue working across applications.

I'm interested in how this could support marketing, research and other work I manage across different clients.

My early experience has been encouraging. I still want to see how much oversight the bots need and how consistently they execute complicated requests.

But I like the direction.

Musk is effectively offering users a way to build an AI team around the work they need done. That's a different proposition from hiring one digital assistant.

OpenAI Dots: For people who already live in ChatGPT

Dots could be particularly appealing to professionals who have spent years building context inside ChatGPT.

The agents can retain relevant information, work across connected applications and continue making progress on assignments.

For someone already using ChatGPT to manage research, documents and projects, there's considerable value in an assistant that understands that history.

The challenge is execution.

An agent that remembers everything about a project still needs to complete the assignment without introducing mistakes.

OpenAI has the user base and technical resources to become a major competitor here. How its agents perform over time will determine how far that advantage extends.

Viktor: An AI employee for your existing tools

Viktor focuses on work inside business applications.

It operates through Slack and Microsoft Teams, allowing users to assign tasks conversationally.

The platform can connect with business systems, prepare reports, update records and manage recurring workflows.

I've been evaluating Viktor for potential marketing and business automation, particularly activities that involve several client accounts and applications.

It uses credit-based pricing, so the cost of frequent or complex assignments matters.

I'd want to understand what a typical month of automated work costs before relying on it extensively.

Lindy: For people who know exactly what they want automated

Lindy offers a flexible way to configure AI agents around specific workflows.

It can handle email processing, scheduling, CRM activities and tasks across connected business applications.

I can see its value for teams with repeatable processes and clearly defined requirements.

The trade-off is the work involved in configuring, testing and maintaining those automations.

For businesses that have already documented their workflows, that investment may make sense.

Martin: A conversational approach to everyday assistance

Martin concentrates on personal assistance through voice and messaging.

It helps users manage reminders, scheduling and general coordination.

I like the idea of communicating with an assistant through familiar channels rather than opening another application.

I'd want to test how well it handles complicated requests and unexpected situations before depending on it for important tasks.

Must AI Assistant: A toolkit of specialized agents

Must AI Assistant takes a broader productivity approach.

It offers specialized agents for writing, learning, planning and creative work.

That could appeal to people who want several AI capabilities in one place.

Its current positioning makes it less comparable to assistants focused on executing real-world tasks, but the product is worth following as the category develops.

Who gets to know you best?

Here's something I've been thinking about while testing these products.

A personal AI assistant becomes more useful as it learns how you work.

It may remember which meetings take priority, how you prefer to communicate, what kind of hotels you like, who you work with and which tasks you regularly delegate.

It can also develop a record of previous requests and the decisions you've approved.

That context has value.

If you've spent months teaching one assistant your preferences and connecting your accounts, switching to a competitor becomes less appealing.

You might get a smarter model somewhere else, but you'd have to recreate part of that relationship.

The large AI companies understand this.

Meta has billions of users across its social and messaging platforms. OpenAI has an established audience using ChatGPT for work and personal research. Anthropic is building connections to the software professionals depend on.

The smaller companies have to make their experiences useful enough that people develop lasting habits around them.

For me, that's part of why Instinct has become so interesting. Sending a text feels natural, and the assistant works within a communication habit I already have.

The same familiarity could make these products difficult to give up once they've earned our trust.

There's still the small matter of privacy

Giving an AI assistant access to your calendar is one thing.

Allowing it to read emails, interact with your contacts, make purchases and access financial information is another.

The more responsibility we delegate, the more important permissions and security become.

Companies have introduced dedicated virtual computers, protected credential storage and approval controls for sensitive actions.

Meta, for example, uses a dedicated virtual environment for Muse and requires approval before certain purchases or outgoing communications.

Those safeguards matter, but they don't eliminate the possibility of mistakes.

An assistant can misunderstand a reservation request, misinterpret an email or take an action you didn't intend.

I'm comfortable experimenting with these products, but I'm selective about the access I grant.

I also want to know how developers handle personal information, what they retain and how easy it is to revoke permissions.

The companies asking to manage our lives should be able to explain those details without requiring users to read fifty pages of legal language.

So, which assistant wins?

I don't think we've seen enough to declare a winner.

Instinct has my attention right now because using it feels so natural. I can send a request without thinking about which platform to open or which workflow to configure.

Catch makes sense for executive administration. Muse is becoming a favorite for personal tasks.

I've also enjoyed building my first team of Grok Bots and exploring how specialized agents can work together.

And I suspect Claude Duet could become the strongest professional contender if Anthropic delivers on its ambitions.

What interests me most is how quickly these products are improving.

Six months ago, I wouldn't have expected to be assigning real tasks to several AI assistants across my business and personal life.

Now I'm thinking about which activities I can hand over next.

That's a significant change in behavior, and it helps explain why investors are willing to assign extraordinary valuations to companies that have barely begun charging customers.

Whether those valuations hold up is another question.

For the companies building these products, long-term success will depend on how reliably they perform, how they protect personal information and whether users continue finding reasons to delegate more.

As for me, I still have plenty of testing to do.

But I suspect my collection of AI assistants will look quite different by this time next year. And I'm looking forward to finding out which ones earn their place.