## Notes from 31 December 2025 [[2025-12-30|← Previous note]] ┃ [[2026-01-01|Next note →]] Read "[How to Train a System: Methods, Routines and Institutions to Build National Strategic Capacity](https://www.bsg.ox.ac.uk/sites/default/files/2025-12/Heywood%20Fellowship%20-%20How%20to%20Train%20a%20System.pdf)", a report from the [[Heywood Foundation|Heywood Fellowship]] at the [[Blavatnik School of Government]]. The fellowship is named after Jeremy Heywood (Cabinet Secretary 2012–18) and gives senior UK civil servants space to explore public service issues. The 2024-25 fellow is [Lucy Smith](https://www.linkedin.com/in/lucy-smith-b21bb7367/). ## The report The report makes the case for building system-wide capability across UK government to think, plan, and act strategically. The practical recommendations focus on learning methods: regular simulations and case studies to rehearse strategic pivots, foresight and backcasting exercises, real-time challenge techniques (red-teaming, pre-mortems), and rapid post-implementation reviews to learn from failure faster. These are sensible recommendations. The examples are good: Microsoft's AI red teams with deliberately diverse composition, the rapid US Senate hearings after Hurricane Katrina, Singapore's Centre for Strategic Futures embedding foresight officers across ministries. The report then argues these capabilities should be housed in a "National School of Government" with responsibility to coordinate system learning, curate simulations and cases, develop red team capacity, and deploy expertise for rapid reviews. ## The problem The case for why this should be a "school" is never made. The report identifies real problems: fragmented learning, siloed institutions, weak institutional memory, slow response to failure. But then it jumps to "therefore, school" without showing the work. The functions they describe (simulation design, foresight coordination, red team deployment, rapid review capacity) don't obviously require a school. They require coordination, resources, and mandate. Those can live in different organizational forms. What the report doesn't explain: - Why a "school" rather than strengthening existing Cabinet Office functions - What authority or resources it would have that current structures lack - Whether this (the "school" they propose) is about provision (delivering training), certification (validating others), or standard-setting (defining what good looks like) - What problem the institutional form of "school" solves that a directorate, a unit or a quango like [[Nesta]] wouldn't - Who would staff it, fund it, give it priority in ministerial attention They cite the French [[Institut National du Service Public]] (INSP) as a model, but the box about INSP is almost entirely descriptive. It replaced ENA because ENA was seen as elitist and disconnected. That's context, not an argument for why the UK needs a school. ## Why do I hate schools of government? The "school" form carries assumptions: centralized curriculum, learning separated from work, credentialing logic, classroom-based delivery. It's a 20th century organizational answer to a 21st century problem. If the goal is cross-government capability building, there are more modern alternatives. A platform model where government curates and certifies providers rather than delivering training itself. An embedded model where capability builders sit inside delivery teams, like [[UK Government Digital Service (GDS)]] did for digital. Networks and communities of practice for peer-to-peer learning. Internal consultancy teams that deploy to problems and build capability by working alongside. Rotation and secondment programs where learning happens through movement between contexts. The report doesn't engage with any of these alternatives. It doesn't ask: what organizational form best fits the problem we're solving? It assumes the answer is a school because some other countries have schools. This is the kind of recommendation that sounds substantive but isn't operational. "The UK needs a National School of Government" is the equivalent of "the government should have better policies." Fine, but: what specifically needs to happen? Who does what? What changes? The report's "next steps" section is three bullet points, one of which is literally "trial a simulation exercise." The diagnostic sections are worth reading. The institutional prescription feels like it was decided in advance.