Learning to Code before & during the AI Revolution - Part 4 - You are Superhuman

AI has turned developers into something close to superhuman. One person can now build systems that once required teams, time, and deep specialization. This part explores why that feels exhilarating and why it should also provoke caution. The concern is not that the bar has been lowered, but that effort is increasingly detached from comprehension. Whilst this moment is most definitely extraordinary, it will not last and so what should we be doing to embrace it and make it work for us?

We are currently living in the most extraordinary period in the history of software development. Developers today are capable of producing work that would have seemed implausible even 3 years ago.

With AI assistance, one person can design, build, test, and deploy systems that previously required entire teams. The productivity gains by my reckoning are about 1000%. It is not just little old me that says this either. The OpenAI team are using a system coded and tested 100% by AI with their team. They have been using it for months and they cite the same 10x productivity gain.

In practical terms, developers have become something close to the “Superman or Supergirl” of this time.

Amplification, not magic

AI does not understand problems in the way humans do. It is a kind of massive prediction engine using lots of math (mind you, we might be the same!!) and fast and massive computational power that is getting bigger and fast all the time…phew!

This begs the question - when will Terminator 2 happen? That’s for another day though.

For experienced developers, this (prediction engine) can be transformative. You can move faster, explore more options, and offload large amounts of cognitive overhead. In other words, you can start to use AI like an assistant rather than a Prophecy Engine or an Oracle like in “The Matrix”. Actually, I use my AI - whether it is ChatGPT or Alfred (my Clawdbot) - like a partner albeit a junior partner in seniority but senior in broad based knowledge.

This leads to better systems built faster. The ‘better’ in many ways will come from you because whilst your AI will build you an app or system - it could easily end up as a ‘house of cards’ unless you provide a lot of intervention, refactoring and questioning of what is going on there.

Remember the outsourcing to India in the 1990s - what a pile of poo right? AI was ‘Actually Indian’* in those days and we got systems that were unusable because we didn’t know how to manage the process - the fault was ours. I worked for companies doing this at that time and the whole thing was a nightmare for those of us at the coalface. (Full disclosure - I am Indian born in the UK - so don’t go thinking silly things like I’m getting at Indians).

Effort and Understanding drift apart

Now we are all experts right - the field has been levelled - yaaay!!!

Er…no. This is not good. My view of learning which isn’t at all outlandish is that learning comes with struggle, mistakes and pain - you don’t just ‘get to be a butterfly’ - you need to strain (unpleasant thought but…).

You can generate output without struggle. You can fix bugs you do not fully understand. You can ship features without being able to explain the system beneath them.

And there’s the problem.

Look - if you want to build these things for yourself or your internal team, they might work for you and you might learn something. But if you want to build them for the World - think hard about what you need to cover because the list is extensive and AI can’t do it for you.

In earlier eras, effort and understanding were tightly coupled. You could not make progress without learning something along the way. That friction was not always pleasant, but it forced alignment.

Effort and Understanding used to be the journey. Things have changed and we just need to know that.

The economics begin to shift

This separation has consequences beyond individual learning. It reshapes the economics of the industry.

Senior developers can now amplify themselves dramatically. A small number of highly capable people, assisted by AI, can produce what once required larger teams. As a result, junior hiring is already slowing. Is this a good thing. To me it feels like “no it isn’t”. It is genuinely worrisome.

This does not mean juniors will disappear entirely. It does mean the traditional apprenticeship pipeline is under strain.

If seniors are more productive and juniors are less necessary in the short term, the question becomes unavoidable: how do new people gain experience in a functioning team? And what happens when the current seniors eventually step aside? Well - maybe that last question is moot anyway because of the pace of change.

Nobody really knows the answers - what I see right now is an insane and brainless race to win by some exceedingly rich and powerful men who do not have the World’s interests at heart - merely their own.

This moment will not last forever

So where does that leave us. Things are great and not great at the same time!

Right now, it is an extraordinary time to be a developer. The tools are astonishing. The leverage is real. But it won’t last - yet who knows what is the next phase.

One thing is certain right now! You are superhuman.