AI, AI Everywhere

AI is becoming our new working environment!

A few years ago, it felt like you had to make an occasion whenever you opened up a chat for AI. Copy your function into a box, ask a question and paste the response into your editor. Now, AI is integrated into almost every part of your editor, browser, terminal, design tool and inbox. For many of us we can now decide how much we would like AI to help with certain tasks.

The question we now need to ask ourselves is: Are we becoming better builders or are we just using more AI?

1. The AI explosion

Chat GPT led to a quick shift in how we see conversational AI. It was no longer seen as a research tool but instead something that could be used within an hour. Then came copilots to help us with our documents, spreadsheets and code bases. Then agents came along to help break down our goals into smaller steps and then call tools. They will even verify their solution.

New tools have emerged that are AI native. Search, customer service, writing, design and analytics are all being built around the idea that there is an excellent model in the middle.

As a builder this means more tools are available faster than we can form new habits. Tools now can be created in a year that would take a team a quarter to create. However, this can cause pressure to come from expanding. As every product claims to be intelligent and everything is automated, the real challenge comes from deciding what to build.

2. AI became a default

In my usual week I will use AI to summarise a long technical thread in the morning. I use it to create a database schema in the morning and spot an edge case I didn’t consider. I use it to read a stack trace in the afternoon and it tells me 3 potential causes for the error (1 correct). I use it to make the perfect paragraph in the evening.

I use AI to help me code, research, design, write, debug and even make decisions. Some use cases are obvious that they help me, others are habits that have formed due to its availability.

The issue is not using AI everywhere, the problem is using it by default. As it is default we don’t question if it’s the right tool for us at that point in time.

3. The paradox

The more AI you use the faster you should be able to get more done. The reality isn’t quite as simple.

We’ve all had days where we’ve reviewed generated code. The code looks good, the tests pass. However something feels wrong. We spend an hour chasing down assumptions that weren’t written by us and can’t fully explain. We’ve saved time in generating but it’s taken time to verify. In some cases it takes longer. Studies and anecdotes show that in some instances there is genuine gains but in others it’s an illusion. Often we over estimate how much quicker we are.

As everything has an instant answer we lose the friction of having to understand the problem. We can release something without really knowing why. This is great in the short-term but not great in the long-term.

The paradox is we’re not getting more productive, we’re getting more active! It depends on how productive the person controlling the AI is.

4. Skills

Now that generating is easy what becomes valuable? I believe it’s the skills we already knew were valuable but now becoming central.

Problem Solving

AI will solve any problem you tell it with confidence. The builder will need to start questioning whether the problem is real, who it affects and what success looks like.

Judgement

Knowing when something is good enough, slightly wrong or just off the mark. This is a skill that is gained by experience and can’t be delegated.

Verification

Checking your claims, tests and outputs. Just because a model can talk doesn’t mean it’s correct. Great builders have developed muscle memory for testing assumptions, checking the source and running.

Systems Thinking

Most products aren’t individual functions. They consist of data flows, user behaviours, failure points and trade-offs. Whilst AI can generate components quickly someone needs to understand how everything fits together, where it fails and why.

These are not exciting skills but are what differentiate a demo from a product.

5. AI slop

We’ve begun to use the term AI slop to describe the generic articles, polished apps that do nothing, friendly emails that don’t say anything and decisions made based on the first plausible suggestion.

This isn’t always the fault of the tool, it’s often down to no-one pausing to think if it’s good. As it’s quick to generate we tend to generate and release without reading too carefully. This leads to lots of content and products that whilst technically exist don’t have much thought put into them.

As a builder this can be risky as we can create the most volume. We can now create what previously would have taken a small team to create. This can be a blessing if we keep standards high but we’ll just be adding to the noise and losing trust in what we create.

To combat this we should ensure we maintain human standards. We should know when something is bad, take time to improve it and be honest.

The ability to generate more makes the responsibility to curate even greater.

6. What makes a great builder?

What can a great builder do differently in this AI filled world? To put it simply: Learn to control AI.

Using AI means accepting what you’re given. Controlling AI means you have a clear idea of what you want, you’ll give it clear details, you’ll provide it with context and you’ll verify the output. This seems like a small change but makes a huge difference. Instead of asking for a feature you’ll ask for a feature with conditions. You’ll explain why it’s important, what failure points you can foresee and what good looks like.

You’ll also know when not to use AI. Some problems are better solved by reading the documentation, talking to users or spending time on a tricky design problem. Being a great builder means knowing when to go fast and when to take your time.

Having a product mindset will be key. Anyone can build a project. A great builder will ask themselves if it solves a real problem, will it work reliably and will it make sense in a year. Whilst AI makes the building part cheaper to do, it won’t make the thinking any less important. If anything it will make it the main event.

7. Final thoughts

It can be easy to think that AI is taking over but I don’t think this is true for serious builders.

AI isn’t taking over, it’s changing what makes a great builder. We’re no longer trying to create every line and pixel but instead making decisions on what should be created. We’re being precise with our details, verifying it works and taking ownership. These are skills that have always been required from those shipping products but with AI it’s harder to ignore.

I want to ensure I’m creating products, not projects. Using AI where it will make me think better and ignoring it where it will make me think worse. I want to create things that people will miss.

AI is everywhere now and the question is will we still have control over what gets created?