> not claiming that the model wasn't trained on those sessions
The math group inside OpenAI may be training or fine tuning their own models which given some reward functions would definitely bias their usage of the training data towards things that look like math.
You can launder all of it without a human "directly" doing anything.
It's fine. They only need to steal the discoveries long enough to go IPO, then the companies will enshittify and the scientists can go back to doing their original work (which the models can't do anyway).
Just take Baby Rudin and a serious stack of paper with you on these holidays and you should be set.
My two experiences of de-brainrotting:
1. Doing a pilgrimage walk in Japan with a cleared phone home screen. Only the Gaia GPS app on there with the route, and some other local weather and mapping applications. No consumption of anything external during the day. Just steady walking and eating and looking around. It was excellent. You can probably do this anywhere given the right context and commitment.
2. We slept several nights in a hut at 2100m altitude in the Austrian Alps just out of range of cell reception so the icon on my phone was "satellite" all the time. It turns out that without the possibility of communication or fresh input, you can read a bit on your kindle in the evening and then go KO at around 22:00 sleeping like a rock until 06:30. Such an amazing amount of rest and refreshment just from a simple accident of geography.
I have a basic task that I run every day and I use it to eval models and these days the Qwen3.8 model I can run on my laptop is competitive with both Claude and Codex because the models have just been adulterated so far. I have to ask and ask again for it to follow the single skill that describes how to do the task and maybe then will it do it.
For full text search you could also use pg_search (Tantivy) which looks very cool.
But at scale you probably don't want to manage a bunch of mission critical systems that were jacked into your database server. The database is slow? How do we monitor that?
So I would definitely begin like this, but you need to have a plan to break all of these out sooner or later.
We're an AI first company transforming the food supply chain and scaling up rapidly to meet the needs of larger distributors in the industry.
My team is responsible for scalability, reliability and pretty much every other non-product task that comes up day to day. You'll get to work with a top class group of people and lead on making our systems 10x better than they were.
> personalized feedback, quickly identifying where you are stuck and focusing the explanations and exercises on that area,
This has been a huge blocker when I tried to study advanced math myself. Many of the exercise books don't have worked out answers, so often you're either stuck or you have to hunt a variety of sources online for solutions and advice. It kills flow.
Im wary about having my PII hosted on vpses which I suppose makes me a privacy nut. I just host immich on an old laptop and use the VPS to establish a wireguard tunnel.
I think it's ridiculous to expect immich to rearchitect everything in order to make it better able to run on untrusted hardware. It should stick to doing what it is good at.
I've enjoyed the corne I got from typeractive [0]. I am pretty sure I have paired it with my phone for the heck of it but not used it in any meaningful capacity that way. It's running zmk so you should be able to read about others' experiences with it.
The math group inside OpenAI may be training or fine tuning their own models which given some reward functions would definitely bias their usage of the training data towards things that look like math.
You can launder all of it without a human "directly" doing anything.
reply