Short pieces on something that was mispriced, misread or not yet decided, one claim each with the numbers behind it. Plus a running set of experiments on where AI actually improves consulting work, each with a method fixed before the work starts.
Castlelake bid 690p, Apollo 715p, my DCF landed at 719p.
→ Banks$21.2bn net income, revenue up 27%, equities trading up 86%.
→ MobilityThree parties need the same pothole. We scrapped our own idea to get under £1 a month.
→ Agentic AIBuilt and pitched in a day. Retrieval-grounded, no investment advice.
→ FX and savingsA four-currency allocation engine. The diversification floor was the constraint that mattered.
→Last card is a short argument for why AI does not replace human intelligence. It sets up the experiments below: the question is not whether a model can produce an answer, it is which parts of the work still need a person.