Roko's Basilisk
May 28, 2026
🧪 Why the AI land grab is a war over distribution.
📰 Altman walks back the apocalypse, governing physical AI, Spotify reads magazines.
🛠️ Three tools worth trying: Claude for Financial Services, Hebbia, and Rogo.
🗳️ Poll: Who wins the mid-market AI distribution war?
Let’s dive in. No floaties needed…
Get the chance to peek inside founders and leaders’ brains and see how they think about going from zero to 1 and beyond.
Join thousands of weekly readers at Google, OpenAI, Stripe, TikTok, Sequoia, and more.
Check out all the tools and more here to outperform the competition.
*This is sponsored content
Private equity as distribution hack: Anthropic and OpenAI raised $1.5B and $4B+ respectively from PE giants who don't just invest in portfolio companies, they direct them, instantly unlocking thousands of mid-market businesses without a single sales call.
Benchmarks are the wrong story: Claude’s 64.4% finance score beat rivals, but even the winner fails 1 in 3 tasks. The real news was Claude plugging directly into Moody’s, FactSet, S&P, and seven other data providers via MCP, making the AI the interface and the terminal the backend.
Data vendors are already spooked: FactSet dropped 8.1%, and Morningstar fell 3%+ on the news. The underlying data didn’t lose value; the customer-facing product did.
The stakes: Whoever owns the mid-market AI layer owns the operational spine of manufacturing, healthcare, retail, and logistics. Which means finance is just the entry point.
In 1994, Jeff Bezos drove cross-country from New York to Seattle, dictating a business plan into a tape recorder. That plan began to take shape when he incorporated a company and launched an online bookstore. But the books were never really the business. They were the wedge: a way to prove that the internet could offer infinite selection, build consumer trust through familiarity, and justify the distribution infrastructure needed for something much larger.
By the time the world understood what Amazon was actually building, the foundations were already in place. The logistics network, the cloud infrastructure, and the merchant relationships had all been assembled in plain sight, disguised as a company selling paperbacks online.
The OpenAI and Anthropic finance deals were never really about finance. They’re part of a larger distribution land grab: whoever becomes the AI layer between businesses and their software controls the workflow itself. Photo Credit: The Balance.
The lessons embedded in Amazon’s rise were not lost on Silicon Valley. Last month, the world’s most powerful AI companies began making moves that felt strikingly similar to Bezos’ original strategy. Anthropic and OpenAI rolled out financial services partnerships, enterprise deployment initiatives, and agent tooling for Wall Street workflows, while Google continued expanding its enterprise AI push.
On the surface, the story appeared to be about which company had built the smarter model or the more capable agent. But much like Amazon’s bookstore years, the products themselves may be beside the point. What these companies are really constructing is the infrastructure layer beneath the next phase of the economy: the distribution channels, enterprise dependencies, and institutional relationships that will determine who sits at the center of AI deployment.
On May 4, OpenAI announced a new venture, The Deployment Company, which raised more than $4B from major private equity firms, including TPG, Brookfield, Advent, and Bain Capital. Within hours, Anthropic revealed a similar $1.5B partnership backed by Goldman Sachs, Blackstone, and Hellman & Friedman, as well as investors including Apollo, General Atlantic, and Sequoia Capital.
These deals were reported as enterprise AI partnerships; however, that framing understates their importance and overlooks the fact that private equity firms do not really sell software to their portfolio companies; they instruct them. This means they can influence the decisions of hundreds, and sometimes thousands, of companies across industries such as healthcare, manufacturing, retail, and finance. When firms like Blackstone or TPG back an AI company, they can strongly influence which tools their portfolio companies adopt. In practice, that means a deal signed in a boardroom can quickly turn into AI deployments across an enormous network of businesses.
According to TechCrunch, OpenAI’s investor group alone provides potential access to more than 2k portfolio companies. Anthropic’s investors control similarly large networks of mid-market firms. Instead of trying to sell AI tools company by company through years of meetings, procurement reviews, and security approvals, these AI firms are effectively gaining built-in distribution channels overnight.
What makes OpenAI’s structure especially revealing is what it offered investors in return—according to Bloomberg, later confirmed by OpenAI itself, the company guaranteed participating firms a 17.5% annual return over five years. That is unusual in the world of startup investing, where returns are normally uncertain and depend entirely on the company succeeding. In this case, the private-equity firms receive something closer to a guaranteed financial floor. At the same time, OpenAI gains something potentially even more valuable: direct access to thousands of businesses that might eventually run its AI systems.
That broader distribution strategy also helps explain why so much of the coverage focused on the wrong thing. The most widely discussed takeaway from Anthropic’s May 5 financial services briefing in New York was a benchmark score. According to Fortune, Claude Opus 4.7 scored 64.4% on the Vals AI Finance Agent benchmark, ahead of OpenAI’s GPT-5.5 at 59.96% and Google’s Gemini 3.1 Pro at 59.72%.
The numbers were widely interpreted as evidence that Anthropic had taken the technical lead in finance AI. But the benchmark reveals something else as well: even the top-performing model still fails roughly one in three finance-specific tasks. In a regulated industry where a hallucinated valuation, an incorrect compliance summary, or a misidentified counterparty can create legal liability, there is a meaningful difference between a system being capable and a system being reliable.
The more consequential announcement from that same event received far less attention. According to Anthropic’s official blog, Claude can now connect directly to major financial-data providers, including Moody’s, FactSet, S&P Global Capital IQ, MSCI, PitchBook, Morningstar, LSEG, and Dun & Bradstreet, through a system called MCP, an open protocol that allows AI models to pull information from external databases in real time. Moody’s built a dedicated integration that gives Claude access to its proprietary credit ratings for more than 600M public and private companies.
That shift may prove far more important than the benchmark rankings themselves. For decades, financial analysts have used products such as Bloomberg terminals, FactSet subscriptions, and Morningstar dashboards to access data. Increasingly, however, those services are being repositioned as the backend infrastructure feeding AI systems like Claude. The analyst no longer manually searches through the data; the AI system retrieves the information, interprets it, and returns a draft analysis.
Investors appeared to recognize the implications immediately. According to Bloomberg, FactSet shares fell as much as 8.1% after the announcement, while Morningstar erased earlier gains to decline more than 3%. The concern was not that the underlying financial data had lost value. It was that the interface layer, the product customers directly interact with, might be shifting away from traditional terminals and toward AI assistants instead.
If that transition continues, it raises an uncomfortable question for financial data companies: if the AI becomes the primary interface, what exactly are data vendors still selling?
That question also helps explain why finance has become such an attractive entry point for AI companies. The industry is data-rich, workflow-standardized, and unusually willing to pay for even modest productivity gains. According to Google Cloud’s official blog, Macquarie Bank saved 130k hours in seven months using Gemini Enterprise, one of the clearest public examples to date of measurable AI-driven efficiency gains within a large corporation. Anthropic has also placed Claude into production at JPMorgan Chase, Goldman Sachs, Citi, AIG, and Visa, according to Fortune.
But the 72-hour wave of announcements was not primarily aimed at Goldman Sachs or JPMorgan. Those relationships already existed before May. The real target was the mid-market: the thousands of companies private-equity firms already control but that lack dedicated AI teams, procurement infrastructure, or even a coherent deployment strategy. Individually, these businesses are too fragmented and operationally slow for any AI lab to reach efficiently through a traditional enterprise sales process. Collectively, however, they represent an enormous distribution network hiding inside private-equity portfolios.
Amazon’s bookstore stopped being about books long before most people realized what the company was actually building. The books were simply the entry point to something larger: a distribution system, an infrastructure layer, and eventually the backbone of cloud computing itself.
That is what makes the recent wave of AI-finance announcements so significant. If Anthropic and OpenAI succeed in embedding their systems across thousands of mid-market companies through private-equity networks, they will not merely have won a competition for Wall Street workflows. They will have built a distribution layer stretching across manufacturing, healthcare, retail, logistics, and much of the broader economy that those portfolio companies already control.
The benchmark scores matter less in that context. The financial products may matter less, too. The more important question is what these companies are quietly using finance to gain access to.
The bookstore was never really about books. And there is a growing possibility that the finance push is not really about finance either.
Altman walks back the apocalypse: Sam Altman says AI is unlikely to cause a jobs apocalypse, softening his earlier warnings as OpenAI courts enterprises and governments worldwide.
Governing physical AI: A new push for autonomous AI governance in physical environments is gaining traction, as robotics and embodied AI raise safety questions that the chatbot rules can't cover.
Spotify reads you the magazine: Spotify now lets users stream narrated magazine articles, deepening its bet that audio is the future of long-form media consumption.
Framer helps teams design, build, and launch their marketing sites lightning fast. With the ability to publish hundreds of CMS pages in a single click, operate at a global scale with seamless localization, and even host unified content across multiple domains, teams have never been able to ship faster. Trusted by companies like Miro, Bilt, and Perplexity.
*This is sponsored content
🗳️ The AI race in finance is really a race for distribution. Who wins the mid-market through private equity? |
Claude for Financial Services: Anthropic's finance-tuned setup with MCP connectors to Moody's, FactSet, and S&P, useful for testing how AI plugs into real financial data.
Hebbia: an AI research platform built for investment analysts and consultants, a strong example of vertical AI winning through workflow depth.
Rogo: AI agent built for investment banking, handles deal screening, memos, and analyst grunt work, the closest thing to an AI junior banker today.
Leonardo AI: AI image and video generator with fine-grained creative controls, built for designers, marketers, and game studios who need consistent style at scale.
Modal: Serverless cloud for running Python and AI workloads, lets you spin up GPUs in seconds without touching infrastructure.
Quillbot: AI writing assistant that paraphrases, summarizes, and rewrites text on demand, useful for tightening drafts or escaping your own voice.
What did you think of today's email? |
.svg)
.svg)
