You absolutely should not stay where you are. You are also extremely unlikely to be able to make any positive changes there, the company product and needs “are what they are.”

My parents had a laser printer for their business. I realized that it had a very high DPI and also very little ink bleed. I started printing whole books I downloaded at the smallest font size that I could managed to still read, just a few point. I removed line breaks and printed out whole books on a page or two. I found it incredible how much tiny text I could fit on a page.

In class I would read with a little half folded sheet of paper hidden in a notebook. Sci-fi, Russian lit, biographies, classics. I was never caught, but it’s bizarre to think back that I was reading Crime and Punishment while the rest of the class was learning fake American history propaganda.

Honestly, I think the very concept is broken from the start. Neurotypical people do not need to use online dating or hookup apps. In fact, it robs much of the fun out of the experience. Meeting people in person, seeing how they interact with you, with friends, with others, how well they pick up on signals and how they make you feel and letting that pheromonal magic work is how our species has been doing these things from the start.

The normies should polish back up their social graces and head back to the bar and leave the online dating for the folks for whom it really is the best option. If we rob the apps of their mass market appeal, the profit model can collapse and online dating can go back to the niche it belongs in.

Another variation of this seems to be the “thought loop” that agents such as Devin and AutoGPT use.

For example, if a major event or discovery happened, the model would know about it once a critical mass of news stories and discussions had been generated online. You’d probably be looking at a few days before the content accumulated to the point where it would affect the model weights, which encode all human knowledge ever digitized, so some single news article in a training set of trillions of tokens is not enough.

If you want a long term memory of user interactions, long context and RAG seems to do the job nicely as a single fact can be pulled out of a context length of millions of tokens.

Basically pot makes you less self-aware.

It would also be reasonable to rename an aircraft class if it was significantly different from other aircraft. For example, an electric, vertical takeoff fixed wing aircraft that doesn’t require a pilot could be reasonably different from airplanes and helicopters to warrant a new name.