@bigpictureclub: In a new chat, type: Create a skill file called "handoff" with this exact trigger: any time I type the word "handoff" or ask you to summarise the chat for a new session, run this skill automatically. When triggered, go through the full conversation and produce a plain text block I can copy and paste into a new chat. Include these five sections: Who this is for: one or two sentences on who I am and what I am working on. What we covered: a paragraph summarising the main topics and decisions. What was confirmed: key facts, conclusions, and anything we agreed on. Still in progress: anything unfinished, flagged, or left open. Next steps: what to pick up in the next session. Rules: no markdown, no bullet points, write in plain sentences. Be specific. Do not pad. Do not invent anything that was not discussed. Write it as if briefing a new Claude who has never seen this conversation. After producing the file, tell me I can edit any section before pasting it into a new chat. #ai #aiforbeginners #claude #anthropic #chatgpt @Claude
Samia
Region: GB
Friday 03 April 2026 16:33:47 GMT
Music
Download
Comments
eian801 :
I just use obsidian as a memory file and works amazing
2026-04-04 06:24:25
139
user43017974403827 :
Claude's biggest problem isn't memory, it's message limits
2026-04-03 21:30:46
868
Anton O :
I use Notion as my memory file and also have it connected to my drive for back up efficiency. If I change my Masters via Claude the memory and back up drive gets updated, if I change the memory Claude and drive updates and vice versa
2026-06-09 05:22:27
0
Camby :
No no. You need to use both GPT & Claude. Even the models will tell you. GPT is best for ideation and strategy. Claude is best for technical work. Sounds very ‘22, but I actually do have a master prompt to generate prompts which is a god send.
2026-04-05 08:33:32
99
Jake | AI :
You can ask Claude to go back and read the chat as well. I discovered this when I was so frustrated and told it to figure it out and it did 😅
2026-04-03 22:02:35
151
ronaldusmaximus :
I got the weekly message limit. Claude is good, but its limits are fracking annoying and a pure money grab. I beyond hate that they do this.
2026-04-04 17:03:54
14
Mountain man :
Just connect your Gmail to it, and then it can read and write its own memory through Gmail. 155 kB per draft. And Gmail doesn’t have a limit on how many draft you can make. Your clot assistant can read and write its own memories with almost unlimited space with one word/phrase activation.
2026-04-04 13:42:23
16
PNW4Life :
10-20 messages per chat? Whaaaa? I’m doing EXTENSIVE chatting building a site. I use one chat per day and have to write a session summary .md file that I swap out in the project docs eod. The next day new chat, it ingests the file and all background is there and my context window is much smaller. And for Coworker and Code I simply let it read that same summary.md file and all three tools stay synced to what’s latest for that project. Better yet I use Obsidian for the .md editor so if I want to edit/augment the file I do so. AND make use of the Project Instructions as well as Skills to have even more robust usage. On Pro each chat session is about 200k tokens so yeah it can fill fast but if you setup for efficiency you can work around it. Easiest hack; watch your session Usage windows and always start working at the middle of the 5hr window *NOT* the beginning. You’ll most likely float between the windows without ever hitting a limit. ✌️
2026-04-05 21:12:53
5
Cain Marko :
I’ve actually solved this in a completely different way. If I know I’m gonna use Claude for a big project or if I know I’m gonna have a long conversation with it, but I do as I plan out my conversation and I create a project as the project Rose, I haven’t create a file that I then add to the project file and then I continuously use that project and as the project files grow and the project grows, the conversation gets better and better
2026-04-05 13:31:00
6
Thewokeslowpoke :
Um Claude has memory, just start a new chat just tell it to refer to the previous chat
2026-04-06 01:28:13
6
ChiknPox :
His biggest problem is that he never knows the date and time.
2026-04-10 12:11:26
13
voiceofthechild :
Please drop the instructions for creating the skill
2026-05-05 06:38:35
1
Kelly Wotherspoon-Rawle :
The context window is now huge (a million tokens), I haven’t had this issue since the last update.
2026-04-04 03:04:40
5
japanexposed :
This is why ChatGPT is still better than
2026-04-04 01:55:04
7
suMoKo.xyz 🖤🤍❤️ :
Claude desktop - Obsidian using metadata for wiki. All projects under one dir. connect all 3 desktop code/chat/cowork to dir and Obsidian Vault. Use rules in top dir to define this setup. Skill files set up per project. Some skills global but only what you need. This gives you context across everything.
2026-04-04 05:56:08
6
DaScorpio :
handoff
2026-04-04 09:32:37
2
Moorepc :
Nope not memory, it’s the limit. If you came from gpt, it will be painful.
2026-04-04 00:14:24
12
siya_mats :
Create instruction’s in Claude Projects where you tell it to log and read your sessions in Notion as it acts as your agents brain (memory) without losing context in new chat windows.
2026-04-05 21:59:19
5
Chelley 🇪🇺🇧🇧 :
Oh the same with Gemini. It was stucked in a horrible loop and I started cursing it. Did it right and next question qnet straight back to something prior.
2026-04-03 16:37:17
5
N0_Vaseline :
Just connect to Obsidian.
2026-04-04 21:03:04
7
Promd879 :
just download desktop app, mcp to your c drive and save chats in notepad. That's what I do, works like a charm
2026-04-03 23:23:51
4
Adam Next AI :
I’m trying to automate my entire life with AI 🤖 on day #3
2026-04-20 05:49:38
1
rentflowsolutions :
10-20? I'm overworking that poor guy
2026-04-17 04:04:00
1
Original Bible Foundation :
here's another solution: use GPT or Gemini
2026-04-06 03:13:46
2
Ekwy Chukwuji | AI & Business :
Also add ‘pickup’ command for the new session to pick up from the last handoff. And you should be good 👍🏽
2026-04-05 04:59:28
2
To see more videos from user @bigpictureclub, please go to the Tikwm
homepage.