There is an interesting discussion to be had about contact tracing apps and choice of model, most of the internet and all of the press will not be having it. (

There is an interesting discussion to be had about contact tracing apps and choice of model, most of the internet and all of the press will not be having it. (Ill informed and barely remembered fragments that somehow Palantir / Cummings / Satan's niece was involved in this app are nonsense. Palantir is doing some internal data vis work and both apps are being built by the same team anyway) The first thing to consider is whether this has led to a real delay in roll-out. We don't know that so far, but we do know that they actually started work on an app based on the Google-Apple API more than a month ago and very few countries have done a full roll-out of an app based on that API yet (and those quite recently). We also know that these approaches work better after "full lockdown" because during the lockdown contacts are already massively suppressed. If it hasn't led to a delay, then the inevitable wailing and gnashing of teeth will be completely irrelevant. Optionality is good, the more things tried the better. The second is to acknowledge that one of the issues with the (very good idea) of partially or wholly anonymous contact tracing is that apparently people don't like receiving this news from an app interface. That makes an app which interfaces which the rest of the contact tracing system like the one that NHSX originally built a better bet. The third is to realise that there is as yet no evidence that these solutions work at all. We know that fully centralised contact tracing (which includes a geolocation tracking app) works in Asia. Nobody knows if either the fully decentralised DP-3T system developed by Apple/Google or the partially decentralised PEPP-PT/Robert system developed by NHSX will works. The fourth is that if you can make it work around phone OS restrictions, there are very real advantages to the NHSX original design which simply cannot be replicated with a fully decentralised system. There are two which are particularly noteworthy. First, you can extract aggregated local infection data from the NHSX system which can drive granular local lockdowns. Second, rather more significantly, because of the central control element, you can revoke notifications. That's important because revokeable notifications allows you to take the risk of allowing some notifications based on symptom onset before a test has been conducted. Since the peak transmission window is around symptom onset and the incubation period is short, notifying early has very significant advantages. People will be irritated if they get a notification which is revoked 24 hours later and you do have to consider the behavioural elements but it is a useful capability to have. You can also do exposure cohorting. If person A had 10 contacts in a three hour period and on day 8 after exposure none have developed symptoms, you may consider releasing them all from quarantine early. This requires developing quite a sophisticated risk model and carefully weighing up the tradeoffs - you may decide not to use it - but it is an extremely powerful thing to be able to do. It has never been the case that the Google/Apple solution is "obviously better". Rather there are complex tradeoffs to be considered between different solutions. It would be one thing if we had data to show that the Google/Apple supported solution worked brilliantly. That would mean that the additional features the NHSX design has are not required and therefore the privacy tradeoff is not worth it. We don't though, we know that the a fully centralised panopticon nightmare system works in South Korea and China and we know that we don't want that. (edit: obviously if they can't make their own app because of Bluetooth issues then the whole game ends and they have to switch) (edit 2: it was apparently due to technical issues that they were not able to resolve. In denser areas the keepalive mechanism they built works well and keeps most iphones and androids visible for long periods of time but if you go through a period of not being around other phones, too many will drop out. Note that Bluetooth is an inherently challenging way of mapping distance and that this applies to any app which uses it)