New history

📚 PDF ⏳ 6 min 📖 7

Our trend

Our mantra, for years, has been “who will buy the bonds”. In fact, for a variety of reasons lots of people do want to buy them, particularly at the shorter end. Two year bonds are useful, anything longer and it's a bit dangerous. 

The whole bond yield journey reminds me of a podcast we covered back in April 2022 with Russell Napier. US 10 year bond yields were 2.7% then having risen from 1.7% at the start of that year. Now, 4.7%. (The chart above is the 30 year curve). 

Back in 2022:

As far as that went, yields rose a lot and people did buy the bonds but 10 year and 30 year bondholders have been destroyed in the process as the prices of those bonds correspondingly collapsed.

My personal view is that the financial world as we knew it ended on 15 September 2008 with the collapse of Lehman Brothers and subsequent bailout of the entire financial system. Looking at bond returns since then, had you invested $10,000 in 20 year bonds at that point, you would now have $15,000 (assuming full reinvestment), adjusted for CPI over that period you now have $9,677. 

Russell Napier was right in 2022 and he is still right now. Perhaps the most amazing thing is how little narrative there is about the disastrous performance over time of the largest financial market in the world. There is almost never a question to a super fund manager about why they buy them. It is simply assumed that a large chunk of the portfolio will be represented by them.  For example, this is the Fin Review this week. 

It's time to listen! Rightio. 

US Debt hit $40 trillion this week. Far ahead of the Congressional Budget Office prediction of $39.4tn by the end of the fiscal year. It’s interest payments that are accelerating the pain too. It is going to get progressively harder to pull the wool over investors’ eyes about the true return on these assets over time. 

Predictably, the US Treasury has now intervened, buying longer dated bonds with money issued at the shorter end. 

The government's fight with the bond market is ‘our trend’. Like Google tracks clicks, we track this because it is the raison d'être of the fund. The time to listen was 15 September 2008. New history started that day and slowly but surely the chickens will come home to roost.  

Nooooo

Buy Hardware

I can’t actually bring myself to do it though. Currently operating on a MacBook Pro from 2019, the last Intel edition. It’s been, and continues to be, a first class machine. Big keyboard, 16GB of RAM (a lot then, but a standard amount now). So why buy new?

He is referring to a new model known as Qwen 3.8 27b. The 27b part is quite important making it a small 27 billion parameter model when compared to the larger closed source models which are in the trillions. It is at least 50 times smaller, and so capable of running on much less aggressive hardware. Indeed, hardware you can run from home. 

This guy is honestly a bit nuts, but he is worth a follow. So far he has been absolutely right and seems to have a direct line into Chinese model development. 

He has had the same theme since last year. Buy hardware because soon there will be none available. 

But why would anyone bother to buy the hardware? What Jun foresees (because, like I say, he is a bit bonkers) is that the models are going to get so good this year that they will simply not be available. That is in fact already the case with the very best models, which you probably think you are using and are not. Every truly frontier model is now closed within the confines of OpenAI, Anthropic and Google. The only way to get truly frontier performance will be with a Chinese LLM on your own hardware. 

If that is the case there will be real advantages to people that can do it. Like having a staff of geniuses compared to everyone else's ordinary mortals. It’s not certain that this plays out, but that is Jun’s strategy and I believe it is China’s strategy too. What is more, it’s working. 

For Agents and Humans

As we head into the AI age we are going to see digital assets embed themselves into the architecture of finance. For example, if I want to build a product that accepts payments, I can instantly do so with bitcoin, but if I want to accept credit cards it's a whole lot harder. Put another way, the entire banking industry does not have APIs that developers can use. So for all the coding that is going on around the world at the moment, not a single AI agent is wiring itself into the traditional financial system, because they can't. 

Lightning Labs announced their new payment product Wavelength this week. I’m genuinely impressed by this one because it does solve a lot of problems that we have. 

The main issue with the lighting was always that you need a full bitcoin node, a full lightning node and then you have to manage channels within the network. It’s too hard and too technical for most people. Wavelength essentially does all of that bit, but you keep control of the bitcoin. My judgement on this is that I’m going to use it, and that’s all I really need to know. 

It’s very clever, specifically because an AI agent can simply use the Github link build an entire payments platform and the user controls the keys to the kingdom. More than that though, the protocol will support bitcoin, stablecoins and any taproot asset (which will include USD).  

The openness of bitcoin and all the associated protocols are going to make it the default product for AI Agents. They cannot pass KYC, they cannot work with traditional banking because the banks are not open source. Nobody has ever seen the Commonwealth Bank codebase, I’d be surprised if there is one person left at CBA that understands it all.

It’s a huge opportunity. Your business MUST have an API. You must be agent-first and agent-friendly. Twenty years ago “we need a website”; ten years ago “we need an app”. Now, you need an API so the agent can do the work. Bitcoin has them. 

Euro-Trash

The most spectacular thing about the trade balance between Germany and China is how truly balanced it actually was. Right up until 2021 Germany was a huge exporter of cars and machinery to China. Now a net importer of both.

In 2021 Germany turned off its three biggest nuclear power plants and by 2023 they were all gone. Consequently industrial costs have exploded and they cannot compete with Chinese models, ironically on EVs in particular.

Yes, China is dumping cars on Europe because of low domestic demand, but even then, it's not close. 2023 onwards and it absolutely explodes.

The solution for Germany is to leave the EU and drop the deadweight they have been carrying for two decades. Either that or the euro falls apart before they get a chance to dump it. Ten more years. 

Further information

Our July 2026 report to investors can be found here.