affenlehrer
@affenlehrer@feddit.org
- Comment on How to I prove to someone that the U.S. moon landing wasn't staged? 2 days ago:
The flat earth people often say it’s a hologram
- Comment on 2026 is the year of the horse 4 days ago:
Our beloved HOTY
- Comment on The Hole 5 days ago:
Dangerous
- Comment on ard 5 days ago:
Doesn’t work in German (also a Germanic language like English):
- Mustqerd: Senf / Mosterd
- Bastard: Bastard
- Wizard: Zauberer
- Coward: Feigling
- Drunkard: Säufer / Trunkenbold
- Comment on don't 👏 claim 👏 to 👏 be 👏 a 👏 fan 👏 of 👏 tag 👏 team 👏 wrestling 👏 if 👏 you 👏 don't 👏 know 👏 who 👏 this 👏 is 5 days ago:
Ah, the Duras sisters…
- Comment on ELI5. Limit of current gen AI/LLMs 6 days ago:
I have mixed feelings about it. I wouldn’t give code a full production application but I think it’s sometimes helpful if the LLM is able to generate a prototype or scaffold to get a head start. Removes some of the friction of starting a project.
The fully vibe coded stuff I’ve seen so far were usually unmaintainable dumpster fires.
- Comment on ELI5. Limit of current gen AI/LLMs 6 days ago:
The LLMs will just predict probabilities for the single next token based on all previous tokens in the context window (it’s own and the ones entered by the user, system prompt or tool calls). The inference engine / runtime decides which token will be selected, usually one with high probably but that’s configurable.
The LLM can also generate (predict) special tokens like “end of imaginary dialogue” to end it’s turn (the runtime will give the user a chance to reply) or to call tools (the runtime will call the tool and add the result to the context window).
Some of the tool calls are e.g. web searches etc. and the search results will be added to the context window. The LLM can decide to do more calls for further research, save data in “memory” that can be accessed by later “sessions”.
Models tend to get larger context windows with every update (right now it’s usually between 250K - 2M tokens but the model performance usually gets worse with more filled context windows (needle in a hay stack).
To keep the window small agentic tools often “compact” the context window by summarizing it and then starting a new session with the compacted context.
Sometimes a task is split into multiple sessions (agents) that each have their own context window. E.g. one extra session for a long context subtask like analysis of a long document with a specific task and the result is then sent to an orchestrator agent in charge of the big picture.
- Comment on Current events dictate that I post this. 1 week ago:
Aggressive bomb definitely bad, defensive bomb probably impossible, preemptive bomb maybe ok
- Comment on It's literally science 1 week ago:
I popped my two discs around L5. So far I’m trying to avoid surgery, pain is mostly under control at the moment but I’ve had extreme ups and downs for the last few months so…
- Comment on i love positivity 1 week ago:
Good job!
- Comment on The bad behavior of the richest: what I learned from wealth managers 1 week ago:
One of the worst things about this is in my opinion that they call their boredom activities “work” and claim to work 16 hours each day while calling their employees lazy for not working round the clock and on weekends.
- Comment on Would you reboot the router for a Scooby Snack? 2 weeks ago:
True Norwegian Black Raggae?
- Comment on Would you reboot the router for a Scooby Snack? 2 weeks ago:
I’m sorry
- Comment on Would you reboot the router for a Scooby Snack? 2 weeks ago:
I can’t really tell why but he also looks extremely Scandinavian to me
- Comment on Phrog 2 weeks ago:
Impressive. Now let’s see Paul Allen’s drawing
- Comment on Find what you like and stick with it 2 weeks ago:
Butterbrot geht immer
- Comment on What a courageous man 2 weeks ago:
Do you by chance know the backstory of the Mark Ruffalo blowing thing? I started watching the channel a bit late and it was already established at that point.
- Comment on im not telling you woof 2 weeks ago:
If there’s on thing I learned from watching a lot of true crime, it’s this: ALWAYS ASK FOR A LAWYER
No chit chat before, first thing you do is ask for a lawyer and then keep your mouth closed.
- Comment on my bf says sleeping with your balls out is the male equivalent 2 weeks ago:
I don’t know what that means. My balls are always out of my body. I prefer underpants with a little tighter fit since I’ve had experience with this: en.wikipedia.org/wiki/Testicular_torsion
- Comment on What a courageous man 2 weeks ago:
Ordinary Sausage
- Comment on Baby pelicans 👎 3 weeks ago:
- Comment on The simulation is coming apart 3 weeks ago:
The first one looks a lot like Crusader - No Regret
- Comment on So Deep 3 weeks ago:
Indeed, and I thought the deep fried hand was the worst idea
- Comment on So Deep 3 weeks ago:
The YouTuber “Ordinary Sausage” managed quite a lot of spectacular deep frying failures. It’s often bad technique but still…
- Comment on They do what? 3 weeks ago:
What kind of assistance?
- Comment on Before you get vaccinated, consider this 4 weeks ago:
Everyone reading this right now lives. So this right here is the solution to immortality!
- Comment on YouTube sponsorship starter pack 4 weeks ago:
EG1 is missing
- Comment on purely transendantal 4 weeks ago:
Thank you. In the meantime I’ve read his statements regarding the affair. I liked his talks and approach regarding AI. Interacting with a sociopath like Epstein got him burned now but I hope he recovers and learns something.
- Comment on purely transendantal 4 weeks ago:
Never expected to find him in the Epstein files
- Comment on Bears or no bears? 5 weeks ago:
We could accidentally release some polar bears in Antarctica, guess they could survive there.
