Norman built an app that makes you pause before you send the message. He says if he had had it five years ago he would probably still be married. UnSent is not a reply generator. You upload the conversation, it looks for pattern (how long they take to reply when you talk about moving in, when the tone shifts, when you are coming on too strong), and it asks the questions that make you stop. The 10 to 15 seconds it takes to screenshot and upload is often enough to kill the 4.15am four-paragraph message. He has spent five months uploading his own texts. Some of the things it surfaced he already knew were true and had never said out loud. The business angle is the same tool on customer emails. A lot of the complaints that actually lose the customer arrive in writing and get misunderstood. First-line sales teams are already using it to read the emotional driver in the inquiry. He is Australian, the team is here, and he is clear the tax settings make it harder to scale a consumer AI product from this country than from three US states that each have more people than we do.

Japan's debt is more than 200 percent of GDP. If you put every cent of production into paying it down it would still take more than two years. The market has not forced the issue yet because higher yields have kept buyers in the paper. That does not resolve the stock of debt. It manages the symptom. Gold sits in the picture because an unresolvable sovereign load eventually gets answered in the currency or in a crisis, and hard assets are where that answer has shown up before. How long do you think the higher yields keep the buyers?

Yields rolling over is not the all-clear. It is the canary. @realDhruvSuri has been tracking the seven-and-a-half-year bond cycle and the indicators for a major low in long-dated bonds are lining up. That means a high in yields. The market will read the rollover as easy liquidity and buy the final leg. Historically that window is weeks to a few months before stocks peak and the right-translated four-year cycle has to cover a lot of ground in a short time. Leading into the GFC, yields rolled over for about four months while the S&P put in its last higher high. Bond market volatility spiked into that window then pulled back. The pullback is the complacency. Aunties, mechanics and landlords talking about a no-brainer trade is the other signal. The FOMO is higher now than it was when prices were lower. That does not make sense and it is how the last spike works. Gold and silver are coming into their half-cycle lows this year inside an eight-year cycle that started late 2022. The first half of that cycle happened in high yields. The second half, if it arrives with cuts and printing, is the stronger setup for non-yielding assets. Stocks look dicey. The bond market is the clock. How much of the final leg are you still willing to ride?

The FOMO is stronger now than it was a couple of months ago when prices were lower. That does not make sense and it is exactly how the last part of the cycle works. People who should know better are reaching for "this time is different" because of the AI build-out. The conversations Dhruv is having, including with professionals, have shifted from caution to not wanting to miss the last leg. That is the emotion that does the damage.

The next fall in yields is probably the last one before a multi-decade rise. People will treat the dip as a return to low rates and everything being fine. On the cycle read it is the pullback you get in a depression-style recession, and then the new regime. Decades of rising yields, commodities up, gold and silver doing the work they did after the war. Celebrating the dip is how you miss the turn.

Australia is one of the major gold producing countries in the world and almost none of the national reserve is on our soil. We know how to dig it up and we know how to store it. A portion in London for trade is reasonable. Having 99.9 percent of a $16.3 billion reserve in another country's vault is a different decision. Other countries have started treating physical location as part of sovereignty. We have not had that argument properly yet.

Someone has raised serious debt against AI data centre chips that have a six or seven year life. The building is purpose-built for those chips, with the cooling to match. You do not pull them out and drop them in a new hall. It is a dead asset once the generation moves on. Using that as stable collateral for a multi-year loan is a strange way to underwrite anything. Who do you think is actually holding that paper?

Germany brought a large amount of its gold home last year. Belgium and the Netherlands are looking at the same question. The pattern is not subtle. Countries that left bars in London or New York for decades are deciding the physical location matters again. It is not only a rainy-day reserve. It is also the thing that sits behind the currency and the debt. Clients who distrust the inflation print are asking the same thing about ours.

The world uses about 27 to 28 million tonnes of copper a year. We produce about 23 million. The path being priced in needs something closer to 50 million. Every new mine has less copper per tonne. Energy, solar, EVs and data centres are all bidding for the same metal. It will get built. It will not get built at the speed the models assume, because you cannot print ore bodies. The numbers do not close. Where do you think the shortfall shows up first?

Ryza says he is producing about four times the output he was twelve months ago, and he would need one or two extra people to match it. He still wears the CFO, HR and legal hats in a mid-size steel shop. He built agents across those functions and has been testing them for a year. You still check the work. His claim is it is more detailed than the version he used to do by hand, ten times out of ten.

The thing that actually brings bond yields down is a crash in stocks and a pullback in land. Not another intervention. That is the 18 to 20 year land cycle. Rate cuts in that window are a response to a short, sharp deflation in asset prices, with unemployment rising. Everyone asking how we get yields lower is going to get an answer they do not like. The cracks are already widening.

Australia owns about $16.3 billion of gold and 99.9 percent of it sits in the Bank of England! We legally own it. London is the custodian. A portion there for liquidity is fair, and the Bank of England holds something like 400,000 bars for various owners. The open question is why almost none of ours is on Australian soil when we are one of the major producers and we already know how to store it. Germany and others have been bringing bars home. Clients keep asking why the grocery bill and the inflation print do not match, and where the reserve actually is. Does a reserve you cannot touch still count as yours?

Australia owns about $16.3 billion of gold and 99.9 percent of it is stored at the Bank of England. Legal ownership and the vault are not the same thing. A portion in London for liquidity is fair. The open question is why almost none of ours is on Australian soil when we are one of the major producers and we already know how to store it. Germany has been bringing bars home. We have not.

The scarce input for the old railway booms was capital. For AI data centres it is tradespeople, copper and energy. Opposite problem. Railways ran on a gold standard. In 1847 a third of English banks went under and the Bank of England backstopped the system. We cannot print trades or copper. EVs are now the biggest selling cars in Australia and they are copper heavy. AI centres are the other buyer. Non-AI building is back around 2016 levels. World copper use is about 27 to 28 million tonnes a year. We produce about 23 million. The path being priced in needs something closer to 50 million, on poorer ore. Chips with a six or seven year life are being used as collateral for the debt behind these centres. A purpose-built hall is a dead asset once the generation moves on. Which scarcity do you think breaks first?

EVs are now the biggest selling cars in Australia, and every one of them is a copper order. AI centres are the other buyer.

Ryza bought the family steel business when it was on its knees, gave himself five years, and is now supplying Cross River Rail, Gallipoli Barracks and the Fitzroy to Gladstone pipeline. He was 32, living back at home on the coast, earning nothing. Two months of help turned into buying the company. The numbers did not work at five or six people, so he treated it like a project, built out the factory on his parents' land, and stepped into larger work. Revenue per employee was the metric that told him a mid-size shop would be easier to run than a small one. He has been running AI agents across the CFO, HR and legal hats for about a year. His claim is four times the personal output. You still check it. Ten times out of ten it is more detailed than the version he used to do by hand. Would you have made the same call at 32?

The bond market controls the central bank, not the other way around. Interventions in this phase make the eventual yield spike worse. Buybacks and support get read as a reason to speculate. Participants get more excited, the stock bubble gets larger, and when it breaks the yields move harder. The only thing that actually brings yields down is the bubble bursting. We are not far from that test. Still think the interventions help?

Could the RBA switch from raising to cutting?? If Australia starts cutting while US yields are still rising, the Aussie gets too strong for an export economy. We need it somewhere around 0.65 to 0.80. Long-dated US bonds are in the timing window of a seven and a half year cycle for a major low. Sentiment matches. You cannot scroll the bond market without hitting a post about the US government going bankrupt, while private credit and property-heavy banks get a pass. A low in bonds means yields fall for a stretch and rhat gives the Fed room then the RBA follows. This could have been the last rise. Maybe one more. Where do you put the next move?

The number most people watch for global liquidity is still sitting near a record at about US$196.4 trillion, so on the surface it still looks supportive. What is harder to see is that the part of the system that actually creates new money, the Shadow Monetary Base, has fallen to a seventeen-month low of US$107.1 trillion, and on a three-month basis it is now declining at roughly 11 per cent. The collateral multiplier remains stretched at a record 1.83, which is how the headline pool can stay high even while the underlying creation of liquidity keeps shrinking. We've been covering that gap for a fair while now. Bitcoin has pulled back from above $84,000 to around $81,000, and gold and silver are still soft at the same time. That split continues to show up in the way these assets trade. Bitcoin tends to follow the Western side of the liquidity picture, which is losing momentum, while gold is still more of a China story, and the latest weekly PBoC prints have slowed. Bond volatility is also sitting near 100, well above the quieter range it held through the summer, so the conditions that kept things calm earlier in the year are no longer as tight. This week’s letter is really about that seventeen-month low in the base, on top of two consecutive weeks of the headline liquidity number falling. You can still get Bitcoin bounces inside that kind of setup, but it does not mean fresh liquidity is being created again. I am still cautious near-term on Bitcoin and more constructive on gold and silver. Buying the weakness can still make sense BUT treating this as the start of a new cycle does not... tipperanalytics.substack.com…