Alex Sidhu← All posts

3 November 2026

šŸ“© claude is about to go public at $2t!?

Claude is going public, a new way to talk to computers (without actually talking to them), a way to autonomously track context for businesses, and my thoughts on sports and the role they play in society


Howdy y’all,

I'm in Texas this week, following a week in San Francisco.

Apologies for the delay - halloween in USA is one of the craziest events - so it was a very jam packed weekend.

I went to a UT game over the weekend, which was very cool, and it got me thinking about why we care so much about sport. More on that below.

šŸ“ŒĀ TLDR:

  • crazy numbers in AI (again)

  • Anthropic is going public in the biggest IPO ever

  • A new way to talk to your computer (without actually talking)

  • A new product I’ve been mulling over

  • And my thoughts on sports

šŸ’¼Ā 1. In the business

SF → Texas

It's my first time in Texas.

After a week in SF, I've swapped one very particular slice of America for another. The UT football game was a pretty good introduction (see below).

Crazy atmosphere at these games.

SF is super tech-pilled (obviously), there’s an insane amount of work being done. But weirdly the streets are often super quiet (probably because everyone is working)

In terms of the business, I'm working through a question that keeps coming up in our business: how do you capture enough of how a company works to make AI useful?

Context is the absolute king - doesn’t matter if you have Einstein working for you if he doesn’t know how everything works.

And then how do you track the measurable impact from this.

I've put a few thoughts on that in the builders' section.

šŸ¤–Ā 2. New in AI and other cool stuff

Anthropic is about to go public

Anthropic has filed its S-1, the registration document for a proposed public listing.

One figure is geuinely absurd: over US$6.6 million in trailing annual revenue per employee, compared with US$1.4 million a year earlier.

Everyone was kind of freaking out about the loss numbers.

But, the filing shows a US$80.7 billion first-half net loss, including US$72.4 billion of revaluations of convertible notes, derivatives and SAFEs. These financing instruments can produce large accounting charges as the company's valuation changes.

For example, Amazon was an early investor in Anthropic.

Rather than buying ordinary shares immediately, it receives a financial contract that can turn into shares later, under agreed terms.

Anthropic records the cash it received and a corresponding financial liability. Amazon’s contract hasn’t changed but the outstanding amount has - the right to receive those shares may have fone from $1b to $5b. The recorded liability also changes - so it’s now recognised as a $4b loss.

So it isn't the same as spending that amount in cash, and it isn't employees cashing out options. Its first-half operating loss was about US$693 million.

Which on a run rate of about $70b a year is truly incredible.

You can read more about it here:

Twitter tweet

I still find the revenue per employee astounding. It makes me wonder how much more value they could capture with more people, although hiring only helps if those people can solve the actual constraint.

Instinct at US$30 billion?

There’s a claim circulating that Instinct is raising at roughly US$30 billion.

That seems fucking insane to me.

Twitter tweet

Instinct is a personal AI assistant you can reach by message or phone, with the ambition of handling tasks like shopping and bookings. The question I keep coming back to is how it captures enough value to justify that sort of price.

I don't personally know anyone using it to buy things. Yes, I know it’s a good product and it’s growing very very fast (supposedly 10% a day)

Twitter tweet

Agree so much with this take

There are plausible ways to make money. A subscription, a fee for completing something, or a commission from the merchant.

But a subscription feels tough - not sure how many would pay for it, value-based could work, but hard to assign value to each action. Commission from a merchant seems most plausible to be honest.

But even the CEO of Doordash (who was one of the first to adopt agentic commerce) said that only a ā€œcouple hundredā€ orders had come through agents. And they do about 8.4 million orders a day.

So feels like I’m not the only one thinking/seeing this.

They also don’t own the interface that they operate on.

And I feel like people really trust the product. So it feels disingenuous if you started getting suggestions for a new restaurant as an an advertised message from your Instinct.

But idk, the great learning with most of these startups is to make something that people want, and you can always figure the monetisation path later.

And people seem to really want it.

Selling the shovels can still be a tough business

Elon Musk has been warning about failures and consolidation among neoclouds.

Twitter tweet

For reference: A neocloud is a specialist cloud provider built around AI computing infrastructure, particularly GPUs. It supplies the computing capacity used to train or run models. CoreWeave's platform is an example.

So the basic bet is: buy expensive equipment, often with borrowed money, then earn enough renting out the capacity to cover the financing, operations and equipment losing value.

There a lot of companies doing this.

It raises a useful question. What happens if the price customers will pay changes faster than the obligations you took on to build the infrastructure?

More efficient models could reduce the computing needed for a particular task. They could also make AI cheap enough that people use far more of it.

Both are plausible.

But it seems the people providing the debt to these neo-clouds don’t love what they’re seeing at the moment and if your revenue is reliant just on providing training infrastructure to the big labs you could be in trouble.

Fireship's recent bubble video is a related watch.

Basically describes the circular nature of the debt-financing deals that are going on.

The question for these providers is whether they can turn demand into durable returns.

šŸ› ļø 3. Tools and things worth trying

Kino: a proper workspace for AI video editing

I've been looking for an editing platform built around working with agents, and Kino looks really cool.

The launch from Luke Igel brings footage, editing and review into one workspace, with support for Claude Code and Codex.

Twitter tweet

An agent can work on the edit, while the clips remain on a timeline a person can inspect and change. Review notes can feed back into that process.

Plus it seems you can make edits while you’re on the go or from a different device which is great.

I think this is really important, editing requires a visual element. You need to be able to see the output and then go back to it.

Currently, it seems aimed at serious production teams, and not so much smaller creators - I'm going to check it out this week.

Super cool that it’s doing multiplayer work as well.

For a more simple run through of how you should be editing with agents feel free to view this breakdown below - an absolute gem - and the absolute best place to start.

Twitter tweet

Very similar to what I was talking about last week in terms of making templates of how you could edit and produce content.

Typing was too slow

Anyone who's worked from a cafe knows how weird it feels to speak instructions out loud to your computer.

Now you might get to look even weirder by moving your mouth without making a sound!

Subvocal is pitching glasses that turn deliberately mouthed words into text. The company claims more than 200 words a minute.

Twitter tweet

The appeal is obvious. Talking is often faster than typing, but everyone around you doesn't necessarily want to hear the email you're drafting.

Funnily enough though, someone has also released Subvox, an open-source AirPods experiment.

Still, it's a pretty interesting glimpse of people testing new ways to communicate with computers.

Capturing how a business actually works

I've been thinking about what needs to happen before you can properly implement AI in a business.

At a foundational level, you need context.

What work happens? Where does the information live? Who makes the decisions? What happens when the usual process doesn't work?

But how do you capture it all?

The slightly Orwellian answer would be to record everything everyone does, synthesise it and let AI work out what to improve.

And you don't necessarily want to track everyone's every move. May not exactly be great for morale.

I'm trying to work out how to capture the useful context with as little friction as possible.

Right now we go in and audit and interview everyone, map out every process step by step and gain a really deep insight into the business.

Which is honestly great, there’s nothing like being embedded inside of the ops of a business.

But it’d be great to get a more data-driven way to of understanding how the business works. One where we can scientifically track the before and after so to speak at a micro-level.

But I’d be really interested in exploring a product that is able to autonomously capture the intricacies of the business from day one.

I think the vision is a 1-click ontology system and you build out agents from there.

Take invoice handling. A written process might say ā€œcheck and approve the invoiceā€. The useful context is how someone knows the amount is wrong, what they compare it against and who they ask when it doesn't match.

That's what you want the system to understand.

Then you need to check whether the new process actually helps. Less chasing? Fewer corrections? Less time from receiving the invoice to approving it? You can do it manually but having some sort of ontology system would be sick.

Capturing the context feels like one of the hardest parts. Once it's useful and organised, you can test several ways of improving the work.

And from there you can actually measure the fundamental impact AI and technology makes in the business.

Which will be infinitely important as the models get better in the future and AI becomes better and better.

I'm working through what that should look like in practice - will let you know once you do.

šŸ¤”Ā 4. Differentiated take: Why do we love sports so much?

I went to a UT football game this weekend.

It's my first time in Texas, and college football is a pretty good introduction. And everyone goes crazy for sports here.

It got me thinking about the role sports play in our lives.

Why do we care so much about them? And why does every little kid want to be an athlete growing up? (Or at least a large majority of them)

Is it the status? The fame?

I was talking to a friend about this and in some sense perhaps from a biological point of sense it’s the most natural thing to want to aspire to.

People who are athletic, strategic enough to succeed and are able to galvanise together to achieve an outcome - feels like in some sense the epitome of our species.

For aeons we had to work together, be fitter and smarter than the next predator in order to live. Sports is just a microcosm of that.

Plus sports is very linear - work harder, train smarter and generally you will get better than the other person, who does not.

Additionally, the binary nature of them is riveting.

For most of life, things are quite messy and unclosed. Sports provides a binary outcome at the duration of a small sample of time.

But I was looking through the rest of history and I could only find two other notable times when so much glory, status, fame and love was ascribed to sports.

The ancient romans and the greeks, and us as a society over the last 200 years.

The most likely outcome seems that we’re now in a privileged enough position to play sports and watch sports and participate in sports, on a consistent enough basis.

And likely the Romans and Greeks had this luxury (or at least this seems to be the only written record of it).

For most of history we were just trying to get by, making sure we had somewhere to live and something to eat.

I think our affinity for them will only increase into the future. In a world that increasingly feels bounds by averages, constrained to the outputs of LLMs and statistics, sports provides a breathtakingly refreshing way to experience life.

There are no fated outcomes. The player who normally makes the shot at the end of the game, could miss it.

It’s unscripted and unbound by the norm.

And finally I think it is one of the few mediums that is truly uniting regardless of socioeconomic background.

A great example of this is a clip I was shown a little while back.

Twitter tweet

Now to be fair, this could probably be used for fodder for another rant I have (e.g. on the nature of status), but I thought it was a good example of how people even from the highest ranks of life, come together to participate in sport.

In the video is one of the biggest streamers in the world. Then also two of the richest billionaires in Spain.

And surrounding them are a variety of other people from different walks of life who are all there to celebrate their team.

For most of this newsletter I spend time thinking about the world and tech.

Going to a random College Football game in Texas gave me a nice stark reminder that for the average person they don’t care about technology and that sports has one of the most unifying and pivotal forces in our lives.

Would love to know if anyone has thought about this - please reply!

5. Carve outs

  • Cool video on social media marketing

  • Beautiful UI (gone viral)

  • Luka is on a generational run

Until next time, folks

Alex

← Back to all posts
Ā© 2026 Alex Sidhu