currently doing my Masters in AI & Robotics I like to build & design stuff. https://teapawt.net/t.co/RdcsU08idh
Joined July 2025
- Tweets299
- Following91
- Followers216
- Likes144
Genuinely, where is the line between AI slop and non-slop? I’ve been thinking about this a lot recently, and I’m not sure I have a clear answer yet.
My best guess is that the difference comes down to whether someone actually cares about what they’re making. I’ve seen people build things with AI that made me go, “Holy shit, I didn’t even know that was possible.”
But at the same time, I feel like the better the models get, the less care you actually have to put into something to get results that look good. So where does that leave us?
Also, back in the pre-LLM days, one of my biggest differentiators was that I could write code, which most people couldn’t. Today, LLMs write better code than I did at the peak of my unassisted coding ability, which was a bitter pill to swallow initially.
Today, no matter what I do, I have the feeling that my work isn’t my own anymore, and that everyone could do the same (which might be true or untrue idk)
Might sound obvious to some, but I wasn't fully aware of how much every piece of modern media is designed to take your energy and attention, until I started my 2nd postgrad in parallel and got busier. It’s actually crazy and instagram is definitely the worst of them. Problem is I can’t really uninstall it because some people actually use it as a messenger for some reason
If @bryan_johnson saw my sleep schedule, his biological age would jump forward by 2 years.
Seriously though, for some reason I haven't been able to fall asleep before 4–6 AM for the past few days. It's really annoying.
“Nobody takes any risks and tries stuff out. “ Very true sadly
Germans are bad at software because of the class system that you find across Germany.
“Important projects“ are allocated to established companies that “know what they’re doing” hence all work goes to SAP and done.
Nobody takes any risks and tries stuff out. Except when it comes from the US and becomes established market standard.
Even the VCs only invest into to series a companies that already went through hell and are making substantial revenue. Hence German startups can’t grow, while in the US, kids can raise a million on an idea and burn it trying to do some stuff Germans dream of due to lack of funds.
I think as all the major AI labs get closer together in terms of capabilities, having a good user experience and awesome products built on top of their models (by the labs themselves) will become increasingly important. Anthropic’s product design has never really clicked for me for some reason, which is why I’m still using the CLI. The Codex client is way better imo.
pretrained attention projections can function as a classification interface without adding a separate head or generating tokens, that’s so cool
arxiv.org/pdf/2610.22276