DeepSeek in the Himalayas

AI is unlike any other technology that has existed.

Everyone has a 'thesis' on how AI would work and change everything.

But the problem is that most folks don't really understand the nature of AI.

I became aware of this dichotomy a while back as I was interacting with the wider industry for work around NonBioS.

Something like AI hasn't happened ever.

The closest proxy that everyone relates AI to is – mobile and internet.

Once mobile was imminent, everyone had time to come up with a thesis on how the internet would change.

Mobile was revolutionary. But it wasn't the kind of shock AI is.

AI represents a fundamental capability that mankind has never observed.

All 'thesis' around AI will be wildly inaccurate because most of these thesis is made based on past observations about similar events.

But there is no past to which to compare AI.

The only way to predict AI is through a first principles approach. But for that you need to be clued into the state of the art.

And for most people, this is literally impossible.

It is only possible if you are building in core AI.

Unless you have that perspective, your thesis is wrong. Even if you have that perspective, your thesis can still be wrong.

The rise of AI, will be the rise of shocks.

DeepSeek was just another one.

For most of us, we cannot play in this game.

We need to hang up our boots and watch from afar.

Here’s why (you can’t afford to miss this real story)...
NonBioS just went into beta.

As part of that the NonBioS team was struggling with alignment issues over the last few months.

Last Tuesday, we pushed out an update.
Actually, it's just one of the two planned updates.

Not only did it work. It worked shockingly well. I thought it was a fluke.

So I gave NonBioS another task. And that worked too. And then another. And then one more.

I wasn't able to work for a day. And then for a week.

Finally, I had to leave everything and head off to the Himalayas for a few days. I came back and now am able to compose this.

You might have heard Meta's CEO say the company will soon deploy AI systems equivalent to mid-level software engineers.

Or Anthropic CEO saying that humans would not need to code by 2027.

We started on this path last year.

Today, we are sure. This is the last year humans will touch code. This is a shock which will hit soon.

If you are reading this, chances are that this shock will impact you too.

Because if you hold mutual funds, chances are that you have invested in TCS and Infosys.

The companies drive India's economy. They might capitulate.

What is happening to NVIDIA, may happen to them, but 10 times worse.

But what can we do? Not much.

It helps me to stay grounded. Smell the roses. Have some coffee. Spend time with family. We had a good run.