{{acive_subscriber_count}} active subscribers.

Intrepid's co-founder and CEO don't do corporate gloss. Their opening letter in the Integrated Annual Report gets into what 2025 actually required: the hard calls, the strategy reset, and how a nearly 30% growth year still came with real challenges.

Read their story

It has been a long time since I have written a letter.

The simple reason is I have no clue what is going in the world right now. Bad news is good news, and good news is great news, and I am clueless.

At times like this it is often better to observe and do and say very little.

I recently spotted this chart on the 25th of April; it was the 18th straight-up day of the Philadelphia Semiconductor Index, otherwise known as the SOXX Index.

I put in my notebook a great short opportunity.

I carried on my business and then thought about this trade some more as I heard the financial media carry on about the streak. I realised that most mean reversion traders would have started shorting this a while ago. So I did a little exercise.

I wanted to understand what this extended streak would have done to a reasonable person trading mean reversion with a martingale strategy. For those who don’t know, a martingale strategy is when you double the position as it goes against you, so 1, 2, 4, 8…. Martingale on its own is a bad thing, but its responsible use is a discussion for another time.

So this was my approach. I said, 'When the streak gets 5 standard deviations from the mean, go short, and each day thereafter, add another short double the size.’ Very reasonable?

The average streak-up is 2.08 days; the standard deviation is 1.51, so a 5 sigma event is, call it, 10 days.

I then run a sequence of different starting trading amounts, the smallest being 5% and the largest 10%. The threshold of a 6.5% first bet size resulted in the first blow-up. The 5% initial wager was down -50% the day the streak ended. Not a blow-up but a very deep loss.

My point is simple.

Strange things happen when trading financial markets. Black swans are by definition unpredictable. There is a lesson here. One needs to have rules in place to avoid the inevitable black swan; it will find you. We witnessed a 10-sigma event that normal statistics suggest will never happen in our history or history for millions of years.

By all means trade mean reversion. However, there has to be a line in the sand where you move on or get run over.

I have a lot of thoughts and observations in the AI space. There is so much happening here, both good and bad; it is really difficult to decide how this will play out in the short to medium term.

If you really want to learn from people plugged in at the highest levels of the technology and investment game, you have to listen to or watch the All-In podcast. Do yourself a favour; this is top drawer.

I started this section with the joke because I recently had a meeting with an AI expert and was pissed off by a comment he made that I knew was coming. He asked me how much I was spending a month on tokens, and when I told him the ridiculously low number, he said, ‘your aren’t burning enough tokens.’

That is precisely the attitude that is going to get us into trouble and lead investors to pour trillions of dollars into infrastructure spending and overcapitalise. I am not naive; I understand the need to go for the land grab, but spending for the sake of looking busy is what interns do to impress their boss.

The Jevons paradox is where, due to the increased productivity from technology, costs come down, but instead of using the same or less of the product, you end up using more. The radiologist example is what is preached by the high priests of the industry. Everyone thought that technology would eliminate the need for radiologists. But the opposite happened; many, many more people started getting tested because of the lower costs, and the demand for radiologists went up significantly.

One observation that I made and cannot seem to shake is that I am noticing a plateauing of productivity coming from AI coding. What do I mean?

Claude Opus 4.7 is so good that it literally writes all the code for you near perfectly. What I mean by ‘perfectly’ is the language syntax and the surface output. However, there is a massive difference in writing good code and good architecture and good process management. This is where there is still a massive gap, and the sheer pace of the code being written is making the management of it very difficult in many instances for managers to stay on top of. Let’s call that my mini insight. The bigger insight is of a financial nature and should be keeping many investors up at night.

The cost of the AI tools is increasing a lot. The inflation in your frontier LLM tokens is way above the Fed's inflation target it doesn’t seem to care about. The open source LLM competitors are now only 4-6 months behind these models at tiny fractions of the costs.

Some of the agents I am developing could easily run on an open-source LLM at almost no cost, which begs the question: how the hell will investors ever see a return on their dollar if, through inference and deception, as Anthropic claim, Chinese copycats are tuning their AI models on the back of Anthropic's and OpenAI’s research spend?

It makes sense that these frontier AI companies are encroaching on the software tooling space to grab more of the ecosystem. This was termed the "SaaSpocalypse" last month and has the industry feeling fragile.

Yesterday Bill Ackman IPO’d his stock-picking fund. According to the Wall Street Journal.

I have no idea why I threw that in the mix. I have been a follower of Ackman for decades and have always been impressed by him, even if he likes the limelight a little too much.

As far as IPOs go, all eyes or I’s are on the $2 trillion SpaceX and $1 trillion OpenAI.

The Euro Stoxx 50 Index is down 8 days in a row. It has only happened 13x in 27 years.

If someone forwarded you this email, you can subscribe for free.

Please forward it to friends if you think they will enjoy it. Thank you.