Arise: Lessons from Solo Levelling for Software Engineers
In the anime Solo Levelling, Sung Jin-woo starts out weaker than the hunters around him. A rare opportunity lets him grow stronger, and eventually he can raise defeated enemies as Shadows that fight on their own. He can step back. Does his strength stagnate? No, it accelerates.
Software engineers can do something similar with LLM-based tools. Many of us tried them early, found them more hindrance than help and moved on. I did too. Tools like Cursor and newer models have come a long way since. Is it still right to dismiss them without looking again?
As the dungeons get harder, Sung's Shadow army becomes more valuable, and leading it teaches him new skills, such as instructing and strategising. An engineer who hands the mundane work to generative AI gains time in the same way. Relying on the Shadows doesn't stop Sung growing. It's what lets him grow.
The amount of information in front of us keeps growing, and the time we spend reading and digesting it keeps falling. There's a limit to how fast we can take information in, and we passed it some time ago. Skimming has become the default, and nuance gets missed. We're drinking from a firehose every day, and the burnout comes from overstimulation, not overwork.
The next layer of abstraction
Programming has changed beyond recognition in 70 years, from punch cards to assembly to object-oriented languages. Most steps were an abstraction over the one before, and nobody misses waiting hours to rerun a stack of punch cards. Every step also had people who resisted it.
When something seems to threaten the value we add, it's tempting to ignore it. We've been conditioned to measure productivity in keystrokes, when the real output is the quality of our ideas and the creativity behind our solutions.
LLMs are the next layer of abstraction, like a better paintbrush for a painter. It's still on us to apply our expertise, use the tools selectively and understand the complexity and unknowns in what they produce.
Hard, unclear problems are what sharpen the tool that matters most, the mind. Better tooling frees up the bandwidth to spend on them, as the Shadows did for Sung. Do you need to type every keystroke? A compiler already optimises the code you write.
Why not just slow down? Because the pace isn't ours to set. Used well, though, these tools give us room to slow our thinking down and spend it on deeper insight while output holds.
Ship faster and you learn faster. Done well, competitors stay a step behind and can't catch up.
A history lesson
What we can control is fitting the world into our cognitive bandwidth, and LLMs help with that. Socrates worried that writing would ruin people's memory. Instead it gave humanity a collective one. Used thoughtfully, LLMs won't weaken us as engineers either. They'll raise how much complexity we can manage and what we can build. Wisdom isn't fearing what change might take from us. It's seeing what it might unlock.