Chase AI
Region: US
Wednesday 14 January 2026 20:43:17 GMT
22780
733
19
89
Music
Download
Comments
DuaneSr :
Ai needs to figure out a way to fix this
2026-01-15 04:10:12
10
J :
Isn’t this what the Ralph loop fixes?
2026-02-24 23:34:34
0
ISTANI FITNESS :
What about local llm which does not log memory?
2026-02-18 09:35:08
0
Aquarius :
I wish I saw this after asking help from 3 LLMs each for 20 hours
2026-02-20 00:02:53
0
Kim Diesel :
I just open a new chat and tell it to reference our previous chat about (topic). Does the summary generation add any benefit if Claude can just dig into its chat history for full context?
2026-03-05 10:37:30
0
Will :
/memory
2026-03-07 23:03:06
0
ESTERLUST | ELUST :
As a context window fills... attention spreads thinner, local coherence beats global coherence, & constraint interference increases. Crowding, not decay.
2026-02-24 23:04:49
0
GhostPrimerib :
claude.md files ftw
2026-02-25 16:47:22
2
Coconut_Tree_Mafia :
😳😳😳
2026-03-09 01:46:21
0
john_stone7 :
Vitadigital.ai has it figured out. They give you a memory layer so then AI can pull from what you want. No context rot!!
2026-01-15 06:03:40
3
Capitán Obvious :
for larger prompts you got about three shots to get it right before degradation kicks in.
2026-01-15 03:26:33
3
capet :
This is true, and then sometimes not true because sometimes and the weirdest ways I could be in a long context because say the plan we’re in required such a large context window to just arrive at the frame of being able to have an educated decision tree, and as you are correct, sometimes I find that if I switch into another contacts fresh, even if the plan is fully flushed out sometimes I could go back to the old context window, where we left off and it’ll know exactly what to do. Where is the fresh context window even with all that previous testing paper of knowing what to do it’s still won’t always get it right so I don’t know I agree with you and then sometimes AI throws me on a loop cause I’m like dang man sometimes I don’t wanna change out of the context window because it’s just hitting 100 out of 100 and I don’t wanna leave because I’m like damn this things like just keeps going always look at it like this whenever the AI stops batting 1000 is when I’ll switch out and then sometimes I look up to my surprise dude I may be deep in the context window like at the tail end and it’s just hitting and I’m like dang if I went to a fresh con window, it would not hit the way that other one does so I don’t know sometimes there’s like a miscellaneous miracle that takes place
2026-01-15 08:21:40
2
lillyflowers :
does each new chat start with full tokens
2026-01-15 12:48:52
0
pancity :
Having a strategic plan for this is so critical, and I appreciate your attention to this as I understand it after having my chats basically deteriorate. Now I will go back in and structure new context windows, and build out the mechanism by which I can rebuild… super interesting and super important. thank you
2026-01-15 10:08:52
0
Erv Lenzy | ErvTV :
If I know, I'm doing a long project with ChatGPT, I will purposely set up all the information I will need for the various phases of the project. So we'll discuss the stages, what steps defined the steps to find the stages, and discuss all the details needed. That usually doesn't take a whole lot of context window. And then I will branch a chat off of that to start each new part or stage. If I come up with something new that we didn't talk about, I can just go back to my starting chat and talk somewhere, add some more context, discuss some new bit of information. And then I just branch off the bottom of that chat continuously as I continue to add new context. This is a really great way to keep things fresh.
2026-01-20 07:36:15
1
To see more videos from user @chase_ai_, please go to the Tikwm
homepage.
