As I continue to write more software, it seems like software engineering is less and less of a solved problem. At the same time, I am not writing code by hand anymore. It is certainly a strange position to be in.
When I was a freshman in college, AI was already here but it was frowned upon to use it. The reasons given were that they would impede learning. Now, just three years later, it seems the use of AI is ubiquitous.
It seems that AI has completely changed learning. I have been trying to build more and it has allowed me to tackle way beyond what I know. The only way I have been growing and learning is continuously tackling new problems and going deep in it. It has to be very intentional and the desire to one shot and get some semi-working shitty prototype is more than ever. I have had to intentionally increase friction to allow myself the space to think as constant steering is still required for the agent and for that end, I need to know enough.
It seems it is easy to get to the 80% of the solution but further intentional refinement is what differentiates and scales. If I do not know a topic, I just continue to talk to the LLMs till I know how to. The way I keep the models in check is by constraining the models to specific requirements and using strict types. This is something I learnt from old man
@badlogicgames
Here’s an aggregate of tips to maximize learning -