The reading notebook
Feeds
Articles we enjoyed, with a few notes on what caught our attention.
Stratégie IA dans le private equity : « On piste les éléphants, pas les petits lapins. »
Our note
The useful distinction here is between isolated AI experiments and changes to an entire operating process. For private equity investors, the question is whether those changes reach the P&L. What interests us is the capability required to get there: understanding the business, working through the data and systems, and carrying the change into day-to-day operations.
Why digital architecture is key to Britain’s energy future
Our note
Having a robust and successful digital architecture means having data sources that speak to each other in a common language.
Proxy failure is an inherent risk in goal-oriented systems
Our note
When a measure becomes a target, it ceases to be a good measure. [...] Similar phenomena have been named and described across a broad range of contexts, such as economics, academia, machine learning, and ecology. Yet it remains unclear whether these phenomena bear only superficial similarities, or if they derive from some fundamental unifying mechanism. Here, we propose such a unifying mechanism, which we label proxy failure.
AI Mania Is Eviscerating Global Decision-Making
Our note
A sharp counterpoint to the enthusiasm around AI. Suresh describes how pressure to demonstrate adoption can displace basic questions about usefulness. Our takeaway: start with the operating problem, listen to the people doing the work, and measure what improves.
Services: The New Software
Our note
Sequoia makes the case for AI businesses that deliver completed work rather than sell another tool. For us, the interesting question is what it takes to own that outcome: fitting into existing operations, handling exceptions, and taking responsibility for the result.
We are all Product Engineers now
Our note
As code becomes cheaper to produce, understanding what to build takes on more weight. What interests us here is the connection between engineering and product judgement: working with the people who know the business, defining the problem, and checking whether the solution actually helps.
Our note
Planning with agents works better when people can explore decisions, test assumptions, and see the consequences. Appleton argues for shared, interactive tools that make complex choices easier to understand. What interests us is how this brings human judgement into the work: using prototypes and visual models to connect a plan with the systems and situations it needs to handle.