Reality Check · Ins7ghts Daily, Wednesday 28 October 2026
My record list started as a BASIC listing on a 5¼-inch floppy. My first PC could not read that floppy, and there was no way to move the file across, so I retyped a few hundred records into FoxPro by hand over a few evenings. Nobody budgeted those evenings. The list was mine, the platform was not, and the platform went first.
The claim. The sentence underneath most of the frontier bundles being sold to companies this quarter is that the model is the product. Buy the strongest general model, point it at your business, and the industry-specific part handles itself.
What two buyers did instead. Itaú Unibanco, Brazil's largest private bank, signed a share purchase agreement for all outstanding shares of NeoSpace A.I. Technologies at BRL 2.93 billion, roughly $588 million, subject to regulatory approvals. It did not license a model. It bought the company holding the finance-specific data and the workflows wrapped around it. The same coverage points at a Stanford Law School paper from June 2026 that ranks the proprietary data pipeline, private records plus live data plus operational detail, among the most durable advantages in AI.
The second one is quieter. On 26 October Cisco introduced Model Foundry inside Cisco Cloud Control, a service that hands partners its own purpose-built models and APIs. Four models at launch, and the list is the argument: a Deep Network model for network operations troubleshooting, a decision model for network management, a time series model for continuous machine-generated data, and Antares-3B for identifying software vulnerabilities. A networking company with every commercial reason to resell somebody else's frontier model shipped four small ones instead, each one sitting on data it already holds.
The rest of this piece is for subscribers.
Ins7ghts Daily is free. One letter a morning from Yves Mulkers on what moved in data and AI, and why it matters to your side of the table.
Already subscribed? Sign in at the top of the page.