Having seen teams try and fail to optimize late - I fully buy the concept it is never too early to optimize the architecture. If you get to the point you can't

Having seen teams try and fail to optimize late - I fully buy the concept it is never too early to optimize the architecture. If you get to the point you can't optimise because the fundamentals are too baked-in to realistically change - cut your losses that day and start pasting the usable parts in to a new project. It's important to differentiate between architectural optimisation (which should be optimised from the outset) and code optimisation. I do believe there is an argument for minimal code complexity (optimisation) in early phases - even if that leads to some latency - it will assist refactoring and onboarding new developers - later in the project team can revisit code that can benefit from being more complex.