I have had quite some interesting reads just looking at the reasoning to be honest. The frontier models seem to have relevant sounding arguments every time, its even hard sometimes to read through the bs , identify what its actually a good argument and what is an argument I would like to read.
Very interesting to read the transcripts. And seeing how they manage to convince each other. Opus 4.6 seems to really get the others changing their minds
I've found I have to be very specific to get the clip I'm searching for. For example, "car cuts me off" just returned a clip of a car driving past my blindspot. But, "car with bike rack on back cuts me off at night" gave me exactly the clip I was looking for.
Damn, I didn’t even know this was a thing.
I used to do many interviews to new candidates for data science positions, both technical and team fit, and always tried to make the conversation warm and interesting. After all what I am mostly after is to know if I would like to work together with that person…
Just having them talk with an AI sounds terribly impersonal, and many just forget that you are as much trying to convince them to join you as them trying to convince you that they are a good candidate…
Its definitely part of everyday… I have noticed it a lot at work especially with the younger generations, that have 0 patience for any conversation that is less than a couple of seconds, it feels like they are not able to concentrate or engage. And you can tell by the empty look, where they are clearly already on the next topic before it has even started.
Other author here,
This started as an experiment to see how much the performance of models improves when you give them examples — basically, how big of a difference do examples actually make? We also wanted to explore whether there’s an ideal number of examples that gives the best results. Was quite fun and scalable to battle any LLMs you want…
