dug 
This is good.
Especially like the point about state capacity being something people feel – in whether a service works, whether something gets built, whether government appears to be in control of its own machinery.
Sort of reminded me of two books I failed to finish last year. Sam Freedman’s “Failed State” and Polly Toynbee’s “Dismembered”. Freedman shows how failure is baked into the operating model – fragmented accountability, outsourcing, and the slow loss of technical and operational sovereignty. Toynbee shows how repeated reorganisations and erosion of capability, particularly outside Whitehall, have gradually thinned the state’s ability to act coherently at all.
Both are saying that the problem is about the systems individual civil servants inhabit (and less about the calibre or intent of the individual).
Where I’d push the argument slightly further is on where capacity actually lives. In practice, it sits less in institutions themselves and more in the operational substrate – the services, workflows, data models, and feedback loops that translate policy into reality. That’s why small taskforces often outperform entire departments: they rebuild the machinery locally, rather than trying to steer it from above.
New institutions and a Northcote–Trevelyan moment may well be necessary. But unless they deliberately replace underlying systems – and create real pathways for legacy structures to be retired – they risk becoming another layer resting on the same foundations.
Ultimately, I don’t think state capacity is an abstract property of “the civil service”. It’s a property of whether the state’s operational systems can perceive, decide, and act coherently in the real world.
That’s the level at which renewal has to occur?
In any case Andrew’s piece is well worth the time to read it:-)