Everyone is going through a bonkers wakeful dream of not reviewing code. Completely trusting agents and their co-reviewers. Their lazy has come out to play and is having a great time.
I have continued to check pull requests, it’s been a little insane because of the amount of code that the AI agents are generating. I keep thinking to myself, why am I bucking the trend here and not trusting the automated reviews? This is just some inner protectionism for my mental health rather than a decision based on experience.
The new software engineer’s reality, being lead-developer to autonomous agents.
Like many, I have been riding the AI rollercoaster. My lows being filled with mourning for the fun of coding. My highs being full of interesting maths, and interesting concepts behind what makes a model. I’ve enjoyed those eureka moments when an agent churns out a huge block of well-written code.
There are some big questions to wrestle with. Questions which will likely need to be addressed by thousands of teams across the world over the next year or two.
I’m a serial project starter, a twisted project starter? … maybe. Ever since I was young I have loved the process of putting my hands to work and creating something. It’s brought me countless hours of joy and satisfaction, but honestly, it has also poured it’s fair share of grief over my head.
What should I do when energy + sweat > output + satisfaction ?
Much of this grief for me originates from reaching a point where I question whether there is any purpose or meaning in the venture I am undertaking. It’s truly very hard to read the signals sometimes.
With the decision to push the boundaries of what was possible with a local first predictive scheduling, I realised that it will be necessary to abstract the various background tasks for keeping the AI Prediction in the Browser using Vector Embeddings up to date.
The first consideration is that these embeddings would need to be ready when the user started to plan their schedule. They don’t take a lot of time to build at the moment but I expect this to slow as the model gets more expansive.