Escaping the flatland

I don’t consider myself an avid reader. But I read more than I watch, all kinds of watching included, which still puts me in a weird very-2026 minority. Intersect that with my screen time, which is probably close to someone who consumes content on Netflix, YouTube, etc… it’s not surprising that various Substacks are some of my frequently visited pages.

One of the Substacks I really like is titled “Escaping the Flatland”, by Henrik Karlsson. Some of his writing really resonates with me, but the title resonates with me the strongest.

(Okay, back to AI. Thanks for your patience!)

I was off work half of 2025. I read books, tried to find a good problem to start my own company, and enjoyed a long awaited pause. All the while, I couldn’t believe that OpenAI would let me use ChatGPT so much for only $20.

The speed of development of LLM-based general purpose tools was is dazzling to me. I learned a lot. Have forgotten a lot. I’ve first-hand witnessed the steps taken towards richer experiences. Of the things I’ve learned, I also digested (and thanks to it) retained a lot. I started feeling like the 20-year old me, back in the first two years of my bachelor’s years, spending hours at the library of my university with my laptop in front of me, following my curiousity shuffling through shelves filled with some of the best texts the humanity has produced.

While speed was dazzling, a gross majority of my conversations felt “flat” to me. Yes, LLMs trickled my curiousities in unexpected ways. But besides the times I had focus, motivation and energy, I couldn’t shake the feeling that I was reading texts that didn’t resonate with me.

Dazzled by the speed, I waited with excitement that someone will build something… They’ll then flip a switch, and voila, we’ll escape the flatland. Information will feel rich, alive, relevant.

This didn’t happen. The progress has been noticeable, but we didn’t hear the switch clicking.

Six months ago, LLM outputs were more spiky than they are today: sometimes they’d be precisely what you want; sometimes you couldn’t get it to inch towards what you want. This still happens, but at an impressively lower rate.

That progress made it easier for me to get an LLM to produce texts that felt richer, with less effort. But in the meanwhile, I still didn’t hear the switch flipping. Perhaps “Finding the Switch, or Building It, and Flipping It” wasn’t a general problem? I don’t know. It’s a difficult question.

My first-hand experience, and the shared aspects of my experience with intellects that I respect, makes me believe that it’s not a general problem, but a difficult one. Regardless, it’s worth making an attempt, since the promise is big.

Like the 20-year old me whose ambitions got fueled by the text that surrounds him, 33 year old me is very much motivated. And yes, chances are that the 46 year old me will look back and find it naive… but those will be 13 years in the right direction.