Things to think about #10 - Ukraine, the Endgame, and Coding with AI
Glenn Loury and John McWhorter are at their best when they disagree, and I enjoyed their discussion about the disastrous exchange between Trump, Vance, and Zelensky at the White House. Both agree that the U.S. is right to push for a negotiated settlement, which involves pressuring Ukraine to acknowledge its precarious position. However, they diverge on how this pressure was communicated and its potential repercussions. Glenn argues that Trump and the vice president rightly prioritized American interests by applying pressure on Zelensky, while John takes the opposite stance, framing his argument within a broader critique of the U.S. president and his administration.
Since then, the world has moved on—most notably with Volodymyr Zelensky predictably climbing down from his high horse to re-engage with the U.S., effectively apologizing for the brief breakdown in communication in the Oval Office. A friend sent me an episode of the School of War podcast featuring Stephen Kotkin, an American historian. I found Kotkin’s analysis particularly compelling, especially regarding the fundamental choices facing the U.S. as peace negotiations progress and his perspective on how the endgame in Ukraine might unfold. Much of the discussion revolves around a key question: bilateral negotiations between the U.S. and Russia and what leverage, if any, the U.S. is willing to use to bring Russia into line.
Forcing Ukraine to the negotiating table—effectively throwing Zelensky under the bus—was always going to be an inevitable outcome of the initial gambit by the new White House. But will Trump stand firm when facing his counterpart in Moscow? I fear he won’t, and this concern explains why the Europeans are behaving as they are, regardless of their weak starting position. At this point, I would seriously question whether the much-debated mineral deal between the White House and Ukraine gives the U.S. enough incentive to fully support Ukraine—especially if Putin, in closed-door negotiations, pledges to protect, or even enhance, American economic interests in the country on his way to Kyiv.
Now, it is incumbent on the U.S. to show what it is made of. Vance’s Munich tirade and the dressing-down of Zelensky, however inelegant, contained core messages that Europe and Ukraine needed to hear. But they were ultimately empty gestures—momentary ego boosts for the White House. We’ve been told that Trump’s foreign policy is based on “peace through strength.” Such a strategy inevitably requires a tangible display of strength at some point. I agree with Kotkin’s framing of the issues and his proposed solutions, but when it comes to applying real pressure on Russia, I fear he is writing a check that Trump won’t cash. Let’s see.
Coding and AI
Much of the most intense debate about the usefulness of AI—and which model is superior—misses the fundamental point: AI already provides a full suite of virtual assistant capabilities across numerous domains. Why is this the case? First, AI can translate natural language prompts into (Python) code with increasing speed and accuracy. Second, users can input their own data into AI-driven interfaces and request analysis or transformation, making AI a powerful tool for data manipulation, be it numbers, text, images or videos. If you want try out some of my GPTs, build with Open AI, have a look here.
Recently, I took the logical next step and began experimenting with using AI to write code from scratch for tasks I’ve been meaning to tackle for some time. It has proven incredibly powerful. As a quantitative analyst—someone who spends most of my workday analyzing economic data—I have long wanted to integrate coding into my workflow but never fully committed to it. AI is the perfect companion for bridging that gap. Dario Amodei may well be correct in predicting that AI will soon generate most of the code across various domains.
I use OpenAI's GPT (Plus) with Canvas, and it is remarkably effective. It quickly adapts code based on my instructions, identifies errors, and seamlessly adds complexity when needed. I haven't yet tested Claude, Mistral, or DeepSeek, though I’m sure they would perform just as well for my use case. In short, AI is a game-changer for people like me—those who have been trying to break into coding for specific tasks but haven’t had the time or resources to do so properly.
I run my code in Anaconda using Spyder, but in many instances, AI can execute code directly within its native interface. This has significant implications for the future of AI-driven automation and integration directly in existing AI tools or via integration into already existing software and tools. Much of my workflow follows this structure:
Macro/market data (imported and stored in Excel/CSV)
Code processes and analyzes data (transformation of data and statistical analysis)
Outputs are generated (quantitative forecasts, models or simply transformed data in charts, PDFs, or both)
AI-enhanced coding dramatically improves this workflow, particularly in terms of productivity and output per hour. For example, step 1 can be automated via an API, directly importing data into the coding environment. Step 3 can be further optimized with actionable outputs—automatically sending charts and results to a list of recipients, publishing them on a site or blog, or even executing trades. I am increasingly certain that my domain of economic research and analysis will be fundamentally altered by AI. Hopefully, I’ll be able to showcase some examples on my blog soon.