Analysis of current economic conditions and policy
Random Observations
Inversion continues, with 10yr-3mo at same level as on 12/`6. VIX at 45. Kalshi odds on a recession at 68%, highest since trading started in August. EPU at 1203.
Dominic Preston at The Verge noticed a number of people on line using AI tools for forensic analysis of the felon-in-chief’s tariff policy, and decided to join in. He nudged a handful of AI critters to propose tariff policies in service of “fair trade”.
They all came up with formulas that based tariff on some ratio of imports to exports, just like USTR did when concocting the actual tariff rule. Of course, we don’t know what the bots chose as training material; could have been the felon’s own speeches. In any case, the bots apparently decided that “fair” meant to cut bilateral deficits in half or to get rid of them completely. Sound familiar?
Maybe Preston’s discovery gives too little credit to the USTR human(s) driving the bot, since the human(s) wedged a marginal this and a price that into the final formula. Still, it’s notable how far from textbook economics and how close to the internet grazing of a chatbot U.S. trade policy now is. A few clues as to the naive nature of the training material, or perhaps Preston’s wording of instructions, is that bilateral balances are the trade measure used and that U.S. surpluses are ignored.
Macroduck
Reminder: The first 15-minute “circuit breaker” trading halt for the S&P is triggered by a 7% drop, the second by a 13% drop. A 20% drop ends trade for the rest of the day.
baffling
the grind down is going to go slower than the circuit breakers trigger. I will be suprised if a trigger is met. but I will be surprised if the grind down does not continue for the foreseeable future.
Tariff policy, if you ask a chatbot:
https://www.theverge.com/news/642620/trump-tariffs-formula-ai-chatgpt-gemini-claude-grok
Dominic Preston at The Verge noticed a number of people on line using AI tools for forensic analysis of the felon-in-chief’s tariff policy, and decided to join in. He nudged a handful of AI critters to propose tariff policies in service of “fair trade”.
They all came up with formulas that based tariff on some ratio of imports to exports, just like USTR did when concocting the actual tariff rule. Of course, we don’t know what the bots chose as training material; could have been the felon’s own speeches. In any case, the bots apparently decided that “fair” meant to cut bilateral deficits in half or to get rid of them completely. Sound familiar?
Maybe Preston’s discovery gives too little credit to the USTR human(s) driving the bot, since the human(s) wedged a marginal this and a price that into the final formula. Still, it’s notable how far from textbook economics and how close to the internet grazing of a chatbot U.S. trade policy now is. A few clues as to the naive nature of the training material, or perhaps Preston’s wording of instructions, is that bilateral balances are the trade measure used and that U.S. surpluses are ignored.
Reminder: The first 15-minute “circuit breaker” trading halt for the S&P is triggered by a 7% drop, the second by a 13% drop. A 20% drop ends trade for the rest of the day.
the grind down is going to go slower than the circuit breakers trigger. I will be suprised if a trigger is met. but I will be surprised if the grind down does not continue for the foreseeable future.