Yes, little AI agent, you can use my computer
Also: SaaS multiples are rising
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
As a small programming note, your humble scribe is under the weather. Having an infant is magic. And a great way to earn a series of colds. To work, albeit in an abbreviated fashion. — Alex
📈 Trending Up: Bitcoin, now over $100,000 … its leveraged children … crypto policy, with Paul Atkins’ tap … startups built by Mag7 breakaways … AI-powered video games … impacts of the IRA … saber rattling in the South China Sea …
Today in ClimateTech: Direct air capture using crushed lime?
📉 Trending Down: Cybersecurity and privacy … no, really … utterly permissive bigtech M&A? … the French government … China-Europe relations …
Number of the day: $73.8 billion, the US trade deficit in October, down 11.9%.
AI is going to get more personal, more involved, and more powerful
Anthropic, a foundation model company that recently raised $4 billion more from Amazon and intends to use the ecommerce giant company’s upcoming supercluster for its data-crunching needs, is one of the coolest companies on the private markets today.
I don’t say that because I have a particular view of the power of Claude 3.5 Sonnet, though people rave about it. No, I’m most excited by Anthropic’s “computer use” product it announced in late October (emphasis original):
We’re also introducing a groundbreaking new capability in public beta: computer use. Available today on the API, developers can direct Claude to use computers the way people do—by looking at a screen, moving a cursor, clicking buttons, and typing text. Claude 3.5 Sonnet is the first frontier AI model to offer computer use in public beta. At this stage, it is still experimental—at times cumbersome and error-prone. We're releasing computer use early for feedback from developers, and expect the capability to improve rapidly over time.
More recently in an FT interview, Anthropic CEO Dario Amodei said the following, regarding what’s happened since the release of computer use:
It’s not a physical robot, but it’s able to type in . . . automate and control your computer for you. Within a few days of when we released it, people had released versions that control an iPhone screen and Android screen, Linux, Mac.
Improving ability for AI models to interact directly with human-facing digital interfaces means that we’re getting closer to my ability to chat with my computer as I work. By chat I mean give directions as I work so that I can do a lot more, more quickly. I don’t want an AI friend, I don’t think. I have dogs that love me; that’s enough.
Thankfully other companies are working along similar lines. The Browser Company’s upcoming Dia browser does a lot more than sit in a chatbot interface. It wants to know you and help you work.
Dia will use the browser as its canvas. Computer use has a broader landscape to play with. I’m not sure which is the smarter bet. On one hand, my browser is where I work, apart from desktop Slack and Zoom. So, building personal AI agent-ish tools into the browser makes good sense and gets away from the difficulty of working in a far wider digital environment. On the other hand, the computer use-model has a larger canvas!
No matter who wins, we win. AI doesn’t fit into my work life today because it requires me to pause what I’m doing to interact with it. In the future, that just won’t be the case. Good.
But today, corporate AI is a little uneven
There’s a weird dual-screen going on between what some big tech leaders are saying about AI adoption, and some of their customers. Here’s the latest side-by-side.
To start, Salesforce CEO Marc Benioff on his company’s earnings call, discussing sales of his company’s Agentforce product, an AI agent platform:
And with Salesforce Agentforce, we're not just imagining this future, we're already delivering it. And you still know that in the last week of the quarter, Agentforce went into production.
We delivered 200 deals and our pipeline is incredible for future transactions. We can talk about that with you on the call, but we've never seen anything like it. We don't know how to characterize it. This is really a moment where productivity is no longer tied to workforce growth, but through this intelligent technology that can be scaled without limits.
Benioff explains why there is optimism for the product: it will allow companies to get more work from fewer people. That’s always popular, and it is net-good for the economy over a medium-term time horizon (in terms of impacts on individual incomes once productivity gains seep into wages, after jobs that used to exist are munched). So perhaps we should not be that shocked that there is early demand.
And yet, here’s Gavin Baker, also from this week:
We should never expect the goals of the folks who buy software for companies to directly match the needs or wants of the people who have to use that software. If this wasn’t the case, Concur would be used by precisely zero companies for example.
But there cannot be no overlap between complaints that much corporate AI today is a disappointment, and how much employers want to buy it. Right? Right?
A quick earnings recap
Cybersecurity shop SentinelOne is tumbling today after beating expectations, but guiding in a manner that Wall Street found weak-kneed. SentinelOne turned in 29% revenue growth, 200 bips worth of YoY gross margin improvements, and a larger-than-last-year $78.4 million net loss for the period.
The company also boosted its full fiscal year guidance — it just completed the third quarter of its fiscal calendar — from $815 million after Q2 to $818 million. That’s a “revenue growth outlook to 32%” the company reports.
Shares of the company are off 13% in pre-market trading.
nCino is another company losing double-digit points off its market cap this morning after reporting earnings that beat trailing expectations. Why? Its guidance of “for total revenue in the fourth quarter of $139.5 million to $141.5 million, below the $143.8 million expected by analysts polled by FactSet,” MarketWatch reports.
Sprinklr, in contrast, is up 7% in pre-market trading after beating expectations.
Gist: The market is still hungry for both trailing results that beat expectations, and guidance atop estimates. That’s a tough market.
But it’s not all bad news
SaaS multiples are rising, y’all:
