Marc Rubinstein, who writes the Net Interest newsletter with 101,000 subscribers, has spent the past weeks building a Bloomberg terminal replica using Claude Code. Rubinstein comes with credibility. He spent decades as a senior equity researcher at Credit Suisse and managed a $4 billion financials fund at Lansdowne Partners. He has used Bloomberg for 20 years. If anyone understands what a terminal actually does, it is him.
Claude Code was released in February 2026 and changed the game by allowing non-programmers to build software by describing what they want in plain English. Rubinstein set Claude a task: index his 250-plus archived newsletter issues, then daily skim financial news and research papers and flag anything aligned with his coverage patterns.
The system learned his angles, his topic preferences, his framing style. It now sends him curated morning briefs that are "remarkably well attuned" to his actual editorial needs.
Premium pricing
Bloomberg charges $31,980 per year for a single terminal. That is $2,665 per month on a two-year contract minimum. Prices went up 6.5% in 2025 after a 9.6% jump in 2023.
Rubinstein's Claude setup costs $240 annually. The economics are not close.
The critical question is whether Rubinstein actually needs what Bloomberg sells at those prices. The answer appears to be no. In his own analysis, his actual morning workflow was six discrete tasks: scan top news for overnight moves and dollar positioning, check his watchlist for prices and volume on 25 positions he actually cared about, pull the earnings calendar, glance at macro events, cross-check his book for threats. He estimates using 15% of Bloomberg's functions. The rest—DCM screens, CDS curves, swap pricers—he touched twice a year.
Chat 'moat'
Bloomberg still owns two things Rubinstein needs: IB Chat and distribution to institutional customers who require the terminal's specific workflows. Those are Bloomberg's actual moats. The rest—analytics, data synthesis, news filtering—is now commoditised. It is available for a twentieth of the price.
This matters because Bloomberg depends on the assumption that data processing and financial analysis are scarce. They are not anymore. Those functions have been disaggregated. Rubinstein can string together free APIs, Claude's reasoning, and his own editorial judgment to produce output better than what Bloomberg's standard interface offers.
The terminal's decline is not imminent. Large institutions will keep paying for IB Chat. But the edges are fraying. Solo investors, researchers and newsletter writers no longer have a reason to maintain Bloomberg subscriptions. They can build better tools for less money using Claude Code.
Bloomberg's crisis is that it sells analysis. Analysis is what AI is best at. That was always a risky business.
