Arjay McCandless
Region: US
Tuesday 07 July 2026 23:44:16 GMT
114455
5065
41
229
Music
Download
Comments
This Week in Asbury Park :
The tools should handle this automatically. Period.
2026-07-08 17:45:25
252
white_wizard_42🧙🏻♂️🇰🇪🇪🇺 :
they do this intentionally
2026-07-08 16:05:45
78
NetworkPIMP :
Bro... get some shoes.
2026-07-10 02:52:57
2
miltank45 :
let it be dumb. im so sick of this AI BS
2026-07-10 04:20:09
0
futiktok :
Request a handoff file and upload it to the next convo
2026-07-11 08:45:47
3
carlitos guey 🐸 :
this happens to me where if i have a long conversation with lots of data im asking it to recall, it recalls the wrong info a lot. this happen frequently when i use it for holding party hosting plans like shopping lists recipes timeframes etc.
2026-07-10 15:58:48
0
Austin :
or...Just don't use it lol.
2026-07-12 23:00:40
0
Dawid :
How do you know what utilization percentage you’re at?
2026-07-08 06:28:45
7
lokiuruz :
Bad news: it’s not Claude getting dumber …
2026-07-10 18:11:21
5
řichie :
why doesn't the llm summarize itself without having to ask it
2026-07-12 02:30:50
0
Quality_Expert_Writers :
use graphify and thank me later
2026-07-08 08:23:48
7
myk.eth :
I try to keep the context window around 40-50% until I clear
2026-07-08 00:53:38
10
Armand :
This was the case like 1 year ago it’s been fixed for ages
2026-07-10 11:59:36
0
juanitomint :
for long tasks is good to plan first and output to a file then start fresh with task execution you also save tokens if you execute with a cheaper model like haiku
2026-07-08 00:46:27
0
Joshua Koudys :
There's going to be a whole class of product that we'll be building in the coming years that's mostly keeping an LLM focused without the context collapse.
2026-07-08 03:52:41
8
ILikeTuotos :
Socks
2026-07-09 18:48:36
1
Jacques Potts | AI Marketing :
yeppp when you hit a decent % of your context window on to a new chat session
2026-07-08 21:16:25
1
many :
it's a good habit to compact only when in the middle of the work, and in any other case completely clear the context based on the milestones you've achieved. given one needs to implement A, B and C being all separate features loosely coupled or not at all, just completely clear the context
2026-07-09 05:15:04
1
Thomas Tessari :
I’ve got some tasks running miles long and have to constantly remind it, basically like with a person.
2026-07-08 19:26:36
0
user1845891767761 :
After a while, I’ll just have that chat summarize and create a new prompt and then start a new one. Makes the results so much cleaner.
2026-07-24 01:04:49
0
elyonsking :
Chat gpt does this on default as long as you tell it to save what you’ve talked about.
2026-07-21 15:11:28
0
m4xleb 🇨🇦 :
I always try to never go over 40% For a big task, build the plan in phases with one session, have the agent prepare the next session contexte in memory than launch plans in new sessions! That is the best methode I’ve used so far to avoid saturated context and clean tasks. Otherwise, do a task than /clear for the next one.
2026-07-09 23:40:13
0
Blessing.data👩🏽💻 :
I noticed this today
2026-07-09 20:35:28
0
idkman123 :
This is great info. Thanks!
2026-07-08 01:03:43
0
To see more videos from user @arjay_mccandless, please go to the Tikwm
homepage.
