AI and I
Claude Code, Bloomberg and the Battle for Data
“For years, I was told Euronext is missing the data revolution. Data is the new oil. You are missing the data boat. As I said earlier, all of us are finding out that maybe we missed the data boat – that maybe this data boat was a Titanic boat, that we missed the Titanic boat.” — Stéphane Boujnah, CEO, Euronext, February 2026.
Before this month, the last time I wrote a piece of software was 40 years ago. It was a school project to design something linked to the history syllabus. Inspired by a popular TV quiz show, I made a game that allowed players to progress around a board by getting questions right – a kind of 1980s Quizlet, though with fewer questions because memory was scarce back then. It marked the pinnacle of my software achievements. After that, my programming skills atrophied and I spent the next few decades deep in Excel instead.
Until last week.
I’d read about Claude Code, Anthropic’s AI-powered coding assistant, but I previously dismissed it as a tool for engineers. As a newsletter writer, most of my software needs are met: Google Docs for drafting, Substack for email management and website hosting, Stripe for payments.
But one part of the stack that I’ve struggled to leverage is idea generation. Indeed, how I come up with a topic each week is the question I most often get asked by readers. So, freshly installed, I set Claude Code a task: fully index my back catalog of Net Interest issues and then, on a daily basis, skim financial news and research papers and other content across a range of sources and flag anything on brand. For the past few days, I’ve received a morning email with a curated list of stories I should consider. It’s been remarkably well attuned.
I have an advantage here: 250-plus issues going back over five years. They serve as a data repository for my bespoke software application to straddle. The model has learned my patterns – what topics I cover, which angles I take, how I frame financial market events. Could I add a module to get Claude to write the whole thing for me? Perhaps – but it isn’t (yet) able to surface information hidden in private repositories like my personal experiences or those of my network of contacts or draw connections using them. Nor can it write in my voice. (Let that sink in.) That doesn’t stop it from providing assistance in other ways.
For a newsletter writer, I may be late to Claude Code. Tech substack has been all over it since the start of the year. But according to Anthropic, the median session on Claude Code is still only 45 seconds, and for only 0.1% of users does a session last longer than around 40 minutes. And that’s among its active users – of which there may be no more than a million. In fact, within the more regulated financial community, I may be early. In a survey of 150 quants, research analysts and data scientists across Europe and North America, Bloomberg found that 54% had yet to start their “generative AI journey” (at least between April and November last year). Another survey finds that among research analysts at banks and brokers in 2025, 58% use AI “a little” with an average rating of 1.77 on a scale of 1 (none at all) to 5 (a great deal).1
There are exceptions. Norges Bank Investment Management – which we’ve profiled here before – worked with Anthropic to establish AI literacy across 600+ employees through comprehensive training programs. It now claims its employees save more than 20% of their time weekly on AI-assisted tasks.
Multi-manager hedge fund Walleye (also profiled) is another early adopter. “As a hedge fund, we should be ashamed to leave money on the table by ignoring tools that make us faster, smarter, and more effective,” CEO and Chief Investment Officer Will England said on a podcast last year. “There’s a non-applicable idea from academia where using AI to do homework or take tests is actually cheating. In the real world, using AI is like taking a magical elixir that makes you 20% smarter instantly, or a lot more. So why wouldn’t you use it? From the very top, we are building a culture around AI.”
Seems that 20% is the efficient frontier in finance AI.
We spoke about the enduring distribution power of Excel two weeks ago and the challenge faced by data providers as AI enables new channels. My next project is to use Claude Code to build my own Bloomberg knock-off. What does that mean for the terminal’s competitive moat – and for the data providers scrambling to position themselves in an AI-first world? To explore with me, read on.