Charles Gabriel called the Thomson Reuters result before anyone put a price on it!
On Ep13, Charles Gabriel quoted Jamie Dimon; "most AI runs on public data. Ours runs on JP Morgan data" and then added the part people skip: "that data needs to be AI ready. It needs to be clean, it needs to be formatted well, and it needs to be accessible in the right ways… companies that are starting to figure it out are starting to realise that it's the unsexy stuff that they have to do first." His ordering was "pipes then people."
What it actually means
Thomson Reuters is the receipt for that episode.
It didn't get a model of its own by having more data than a frontier lab, it had vastly less. It got one because its data was already clean, structured, labelled and defensibly owned, maintained for a century and a half by editors who had no idea they were building a training set.
That's what $40m buys when the pipes are already laid.
Which changes the question worth asking in the room. Not "should we train our own model" but "is our data the kind anyone could train on." For most companies the honest answer today is no, and the work to change it is slow, unglamorous, and starts before the budget cycle does.
Everyone wants the model. Almost nobody wants the filing that makes it possible.
Source: Brew Bytes Ep13 — Charles Gabriel · Sunday 30 August 2026