Macro thinking. Politically independent. Searching for the truth—following the path of moral reasoning. It’s never easy. Nothing posted is financial advice.
@JavierBlas Oil fundamentals have been weakening for 2 months….
Brent aug 2026-dec 2026….topped in April. Chart 1
Aug2026-Dec2026 WTI. Chart 2
These spreads were/are indicating no supply issue, plenty of oil was leaking into the market.
Gold verse Bitcoin
Right scale % return since 2022
Gold in red. Now outperforming.
Both assets have been highly correlated as can easily be seen qualitatively.
With Fed tightening cycle in last phase. Both should bottom in next few months.
Uber handed its 5,000 engineers an AI coding assistant in December. By April, the company had blown through its entire AI budget for all of 2026, with two thirds of the year still to go.
Cheap, basic AI has gotten almost free over the past few years. But almost no company builds serious software on the cheap stuff. The tools Uber's engineers reached for sit at the very top of the price list, and the top of the list is where prices have been climbing.
In April, OpenAI released its newest top-tier model at double the price of the one before it. Anthropic left the list price of its top Claude model unchanged, then changed how it counts text, so the same job now adds up to a bigger bill. And the way these tools run makes it worse. Left on their own, they read a company's entire codebase, write new code, test it, and rewrite it, around the clock. A single task can burn through millions of words, and you pay for every one. At Uber, individual engineers were ringing up $500 to $2,000 a month, and the company had even posted leaderboards ranking people by usage, which only pushed them to lean on it harder.
Microsoft hit the same wall from the other direction. When its engineers used Claude Code, Microsoft had to pay Anthropic for every word the AI produced. When those same engineers use Microsoft's own coding tool, the work runs on computers Microsoft already owns and costs the company almost nothing extra. So it told thousands of engineers on its Windows and Office teams to drop Claude Code by the end of June. It still uses Anthropic's models in plenty of other places. The switch came down to one thing: who pays for each word.
For years, a flat monthly fee hid how much AI a team was really using. The new pay-as-you-go meters show every cent, and finance teams are seeing the true number for the first time. The cheapest AI keeps getting cheaper. The AI that companies actually want to run keeps getting more expensive, and the bill just came due.
@ErikSTownsend@CommodMkt The forward curve prices in supply/demand dynamics….
The underlying price is not necessarily bound by the pricing mechanics of the forward curve….hence- the underlying price has multiple dimensions….one being expectations or more directly speculation.