Euro notes

📚️ PDF⏳️ 5 min 📖 7

Bail-out

“Bitcoin bails out the UK” was not a headline I expected to see quite yet. Should it go ahead, the sale is likely to go down in history alongside Gordon Brown's memorable decision in 1999 to sell 500 tonnes of Britain's gold for $275/oz. That gold is now worth $40 billion and Gordon sold it for $2.5 billion. 

His argument from the time sounds eerily familiar:

“The Treasury argued that gold, though traditionally held as a safe‑haven asset, was volatile and underperforming, offering no interest or dividends. By reallocating about 60% of proceeds into interest‑bearing currencies (US dollar and yen) and 40% into euros, they aimed to reduce overall reserve volatility by ~20 %”

Volatile? Offers no interest or dividends? Better off allocating 40% to Euros?

Yes, that sounds familiar. 

Employment

Not great stats for June with a sharp rise in unemployment in Australia. Now at 4.3%. As usual, there was a tidal wave of commentary about how this would precipitate a rate cut from the RBA in August. They stand accused of ‘policy errors’, as though a 25 basis point cut this month would have made any difference. 

I remain convinced that the entire conversation is missing the point. Interest rate cuts mean nothing when wages are collapsing. In the US, this trend is becoming even clearer among recent graduates. For men in particular, the unemployment rate has spiked this year because so many of the roles they would have taken are now able to be done by artificial intelligence. Entry-level IT jobs are pretty easy for most AI’s and they are materially better at doing them. 

Graduate women, on the other hand, skew toward healthcare which has been largely unaffected to date. There is a greater degree of regulatory protection in that sector and the AI’s aren’t that good at it yet. 

More importantly, the unemployment rate between college and non-college graduates is now the same for men. It completely erases the degree premium. It’s not a new trend either, the gap has been closing since 2010 and has now slammed closed. Worth noting that artificial intelligence threatens white collar before blue collar, the blue collar risk is from humanoid robots which are some years away.  

None of this is good news for birth-rates either. The mis-match between the number of university educated females and males will grow. Ten-year trend: fewer people, more tech, demand down, money printer on. 

A programmer

‘Psyho’ is an elite programmer. He used to work for OpenAI and went up against their internal programming model in a coding competition. An enormous effort over a gruelling three days, his conclusion following his win was as follows:

key takeaways:

- openAI internal coding model is insane

- this could be the last human win

Later in the week, OpenAI then published that one of their models had won a Gold Medal in the International Math Olympiad (IMO), which happened to be held on our very own Gold Coast. This model is internal and its capabilities will not be released to the public until the end of the year (so it won’t be in GPT 5). They published all the proofs to the problems here (note it could not find an answer to question six, which is perhaps encouraging). 

This was just a raw model doing the work, it did not have tools or access to the internet making its achievement all the more impressive. 

Two days later, Google also announced that their model had received the gold medal. The rate of improvement in these fields is incredible, and it's amusing to watch governments and agencies fiddle with dials. ‘Hopefully the RBA cuts by 50 basis points and that will fix it’. It won’t. 

There will have to be some very large economic adjustments because of this and I feel certain Australia will prescribe precisely the wrong ones. By way of example, this newsletter is now checked by GPT 4.5, most people don’t use that model because it is hidden away in the menu on OpenAI. If you have a paid account it's in the ‘More models’ menu.

I do the following:

1. Copy and Paste the draft

2. I use a simple prompt: review this week’s MoneyBits for grammatical errors and errors in flow. Note the general tone and content which can be found at moneybits.co Please do not make corrections for tone or content. Return your suggestions as a table: Original | Recommended| Reason for change 

The model then makes some good suggestions, the reasons for change are also excellent. If you don’t prompt it like this it returns some anodyne horror-show. Also, using a less capable model just results in a Microsoft Word style spell check. 

This used to be a task for someone in our office and would take about 30 minutes a week. The task itself is nuanced because there is the issue of respecting the tone. One human versus another is never the same, ”I’d say it like this” etc. The AI is much better at this because it doesn’t have an opinion to the extent a human does. 

Then there is the processing of the changes. That is done with Google Gemini and takes two seconds. Prompt: Process changes 1, 5, and 7 from the table, and ignore the rest.

I think it saves us 20 minutes per week and is much higher quality than human review. I estimate the value of this single, minimal task at around $1,500 per year.

Even though the example is simple it explains some of the other issues. Unemployment pressure is rising because most (certainly white collar) roles are getting easier with AI. There is more breathing room and rather than employ someone new to ‘help lift the load’ almost everyone can find simple tasks they can give to AI. Using these tools is the way to increase productivity. The government endlessly bangs on about how ‘we must increase output per worker’. Indeed we are a few weeks away from Jim Chalmers productivity summit, it will be fascinating what comes out of it. Fairly certain the usual tropes of “cutting red tape” and “additional training” will be rolled out but lets see. 

The truth of the matter is all tools are here but we seem more interested in preventing their use than adopting them. DeepSeek, for example, is banned in Australian workplaces because it ‘steals data’. Schools are still reluctant to use the world’s number one teaching tool. We don’t have self-driving cars, our trains still have drivers (why? they are on rails) and we banned a few new data centres too. We don’t need them after all. 

Not one of them…..

Considering that a 12% absolute return is now considered the minimum in order to stand still, only one major Australian Super fund made the grade in 2025. The ETF QQQ (the NASDAQ tracker) did 14.9% over the 12 months. Not one managed to beat that index. 

The funds are chock full of government bonds and commercial real estate too, so their returns at the scale they operate are still impressive. 

My argument remains though. They nearly all went backwards in real terms.

Euro-Trash

Applications for the design of the new Euro bank notes are now open. So far, the submissions have not been entirely in line with ECB guidelines.

This is the clear winner to me. Less melodramatic than the others and 100% accurate.

Further Information

Our June 2025 report to investors can be found here.