Please SUBSCRIBE to get full access to the newsletter and website

I will post a lot of stuff on covid and some of my experience in government free.

More recondite stuff on the media, Westminster, ‘inside No10’, how did we get Brexit done in 2019, the 2019 election etc will be subscriber only. Subscribers will also get some extra features like community/regular ‘ask me anything’, and a serious dissent section — I’m interested in the best arguments against what I say. Subscribers will find out first about new projects that I make public. Only subscribers can comment.

I will work out the rest as I go along.

If you subscribe, 1) I can get people to help me, 2) you will help me provide free help to those campaigning for an immediate, urgent and open Parliamentary inquiry into what really happened with covid.

Please SHARE

Share Dominic Cummings

Twitter: @Dominic2306

An index of my old blogs here

What is this substack for?

I will explore:

  1. how politics works — campaigns, government, the political media and so on;

  2. how government can be improved;

  3. how do we learn from high performance institutions, given generally very few senior government figures even try to learn and those who do usually fail;

  4. how can we decisively accelerate progress in productivity, science and technology.

In particular I will explore the intersection of:

  1. selection, education & training for high performance;

  2. prediction;

  3. science, technology, and data-intensive science including ‘cognitive technologies’;

  4. communication;

  5. high stakes decision-making in politics/government.

There is valuable low-hanging fruit hiding in plain sight in the intersection of these fields. Exploiting this intersection was crucial to political success in 2016 and 2019.

I wrote about these things on my blog 2013-2019 before going to No10 in July 2019.

For example:

In 2013 I wrote an essay on 1) the growing mismatch between a) the destructive scale and pace of crises such as pandemics/war and b) the abilities and performance of our leaders and institutions — a mismatch Sagan called ‘a combustible mixture of ignorance and power’ — and 2) what could be done to pre-empt disaster including training in how to make decisions under extreme uncertainty.

In 2014 I wrote The Hollow Men about how incentives and culture in Whitehall program poor performance and anti-learning. I explained how both parties and the civil service systematically exclude those with crucial skills from crucial senior jobs and incentivise destructive behaviour. Most of the important destructive dynamics seen in Westminster’s covid response, including the failures of the Cabinet Office shaped by Heywood, were described here.

In 2017 I started a series on expertise and politics. E.g I discussed work by people like Kahneman and Tetlock on what we have learned about expertise in different fields. Some fields like fighting and physics have fast feedback loops for learning. Politics does not and elections do not generate adaptation in the way democratic theory hopes and predicts.

In 2017 I wrote The unrecognised simplicities of effective action #2: ‘Systems engineering’ and ‘systems management’ — ideas from the Apollo programme for a ‘systems politics’. I wrote, ‘Urgently needed projects to lower the probability of catastrophes for humanity will benefit from considering why Mueller’s approach was 1) so successful and 2) so un-influential in politics.’ These ideas, such as ‘concurrency’ (i.e building many vital sub-systems in parallel, rather than in series as governments normally do, to save time and money) had been forgotten by western governments (though are studied intensely in Beijing and Singapore) and were relevant to the success of the Vaccine Taskforce and failure of other projects. I described how the principles behind how Mueller made Apollo successful are like an anti-checklist for Whitehall (cf. from page 26 which is particularly depressing to re-read post-covid).

In 2018 I wrote On the ARPA/PARC ‘Dream Machine’, science funding, high performance, and UK national strategy. This explored how relatively trivial investments can decisively accelerate science and technology if the management follows certain principles. Like Apollo, it is particularly interesting because 1) it is an example of extremely unusual performance, 2) many are interested in the products (internet, computers), 3) but almost nobody is interested in the management principles that explain its phenomenal success. This contrast is fundamental to understanding why learning is so hard for normal political/government institutions: the principles behind effective action and the people able to instantiate them are so hostile to normal bureaucracies, and are so psychologically hard for senior Insiders to cope with, that case studies of high performance are seen as irrelevant or dangerous. I aso described how leaving in place the EU procurement system and EU regulation of science and technology would be dangerous for Britain. Covid exposed both problems. EU procurement and data (GDPR) rules had to be effectively suspended and there were no serious emergency processes to replace them.

In 2019 I wrote Project Maven, procurement, lollapalooza results & nuclear/AGI safety. It explored the crucial question, almost totally ignored by the media even after covid, of how the governmnt can be a good buyer and how governments think about things like existential risk. I pointed out the Bulletin of Atomic Scientists piece on the potential for lab leaks to cause a pandemic: ‘pretty much nobody with real power pays any attention to all this. If those at the apex of power don’t take nuclear safety seriously, why would you think they are on top of anything?… Total failure is totally irrelevant to the senior civil service and is absolutely no reason to change behaviour even if it means thousands of people killed and many billions wasted.’ Westminster did not want to listen — instead they said that Heywood, former Cabinet Secretary, was a genius and we should all be assured when he promised us that Britain was the best prepared country in the world for a pandemic and Whitehall had nothing to learn from the private sector on procurement, digital, data and so on.

In 2019, a few weeks before the Prime Minister asked me to go to No10, I wrote High performance government, ‘cognitive technologies’, Michael Nielsen, Bret Victor, & ‘Seeing Rooms’. This explored why the core institutions of the UK would fail in the next big crisis and what could be done. Having written about the Cabinet room unchanged since summer 1914 and the likely collapse of COBR, I then had to sit through the collapse of COBR in the 1914 Cabinet room, hitting x2 x2 x2 on my iPhone, scribbling numbers on a white board and saying ‘so at this rate the NHS will be broken in X days’.

In January 2020, after the election, I wrote a blog, Two hands are a lot, to start hiring different sorts of people with different skills — including data and project management. In it I wrote, ‘There is a huge amount of low hanging fruit — trillion dollar bills lying on the street’ if we could improve broken political institutions. Westminster howled with laughter at the idea that data scientists and great project managers combined with changing Whitehall ‘HR’ policies could exploit trillion dollar bills hiding in plain sight. Tragically the trillion dollar bill was already lying there on the pavement, playing on the news, and a few weeks later tens of thousands died for lack of data, the skills to interpret it, and great procurement and project management skills. Though my hiring/reform efforts were too late to avert disaster, some of those I started hiring in January played vital roles in the crisis and helped save lives. We created the Analytical Private Office in No10 to embed crucial, but previously unavailable, skills in the PM’s private office. Just as Red Teams have often been ignored, having the right skills/advice cannot guarantee decision-makers will act sensibly (cf. 22/9/2020) but at least it gives us a chance.

The media portrayed my push for reform as ‘Cummings at war with the system’. In fact, many officials supported the push for serious change. Indeed, many more officials than MPs support serious reform because serious officials are much closer to the reality of dysfunction than the MPs, few of whom understand how power really works in Whitehall — just look at how hapless most ministers are about basics like how they are manipulated via agendas, pre-meetings, action points, ‘legal advice’ etc. Further many officials saw the system implode over covid and below the top level, people who don’t have to feel responsible for the implosion, the need for change is obvious. For example, many officials supported the push to make all civil service jobs ‘open by default’ — the Cabinet Secretary supported it too, though he also said, obviously, that there’d be fierce resistance and it would need the PM’s explicit support. One of the big problems in building a coalition inside the system for change is that officials understandably do not have confidence that elected politicians really care about ‘the wiring’ and will make sacrifices to improve it — and while officials can easily block progress they can only make progress if the PM actively helps.

Some core lessons:

  1. The fundamental problem is not one of knowledge. We can study how to create extremely high performance institutions in case studies. Crucial high status people with power don’t bother reading them.

  2. The fundamental problem is not one of resources. Crucial improvements are near-zero cost and don’t need many people. Crucial high status people with power don’t bother looking for the resources.

  3. The fundamental problem is not solved by the apparent incentive ‘winning elections/grabbing power’. Crucial high status people with power are provably irrational about basic issues of strategy and communication even when strongly personally incentivised on a short time scale - e.g you win/lose an election in X weeks.

  4. The fundamental problem is that dysfunctional systems, like Westminster/Whitehall, fight to ignore crucial system dynamics and to stay dysfunctional. Improvement means no/less power for powerful people/institutions. If they can keep power by keeping the ‘unrecognised simplicities’ of effective action unrecognised, they will — and usually do. Key players in the system know that it is career death to threaten the system. Only very odd people will move before there is an obvious critical mass. You will see practically zero media analysis of the fundamental reasons for the covid failure — everything is surface phenomena, conventional tales of ‘incompetent X replaced by Y’, ‘power struggle over Z’ — but never an analysis of why incompetent X was put in that job then replaced with another incompetent Y, or why the ‘power struggle’ is programmed by the players’ incentives. This is why you see recurrent patterns of failure in history — things like Whitehall getting deterrence wrong in 1870, 1914, and 1939, or No10 now lying about covid to avoid facing reality, as facing reality would make the case for regime change obvious. This is why people like General Groves (Manhattan Project), George Mueller (Apollo), Robert Taylor (PARC) tend to be pushed out of normal bureaucracies and after they go performance reverts to the mean. The assumption that the epic failures of western governments over covid ‘must’ mean serious changes to the wiring of key institutions is false. Often political institutions keep failing until they are destroyed even/particularly when failures are huge and obvious because the closed system resists any group changing course. This is why improving politics is a systems problem requiring a systems solution.

    As the complexity of civilisation, the destructive scale of technology and crises, and the speed of crises all increase, there is no greater multi-trillion dollar bill lying on the street than this problem: how to get the most able people — who don’t provide 50% more value than average leaders but many orders of magnitude — into the highest leverage positions with the right incentives despite the tendency of existing political bureaucracies to entropy and anti-learning? We have discovered no reliable method to solve this problem and even after covid you will see close to zero discussion of it in the media or the discussion of political elites. (Other elites do discuss it but privately for obvious reasons.) Without progress on this problem humans are playing round after round of Russian roulette. We got very lucky a few times with nuclear weapons. After covid it is easier for many to imagine how our system is vulnerable to a 1914-type collapse. This does not mean we will solve the collective action problem.

I will also explore some issues around science, technology and the startup ecosystem. This absorbed something like x20-x50 more of my time in 2020 than politics and ‘communication’.

I will also explore international politics. Connected to the problem of coping with crises better is changing the psychology around international competition. If we continue with things like ‘launch on warning’ nuclear protocols, our chances are low of making some sort of transition to a state in which we are not regularly threatened with 1914-style system collapse. If we could develop international projects that strengthen the idea of ‘humans against the universe’, instead of ‘humans against humans’, then we will be much safer. And if we avoid collapse then science, technology and markets will keep generating greater prosperity and knowledge.

The intelligentsia that dominates political and cultural institutions know little about these processes and how our institutions really work and are, in David Deutsch’s phrase, like people in a submarine opening the hatches…

Please subscribe!