ai.field.notes
Region: US
Thursday 16 April 2026 13:26:22 GMT
6829
175
45
15
Music
Download
Comments
Santa's Apprentice :
and tell Claude not to narrate as it works.
2026-04-27 18:40:09
1
CasaTechMedia :
I use codex to plan, Claude to review the plan and fill in the gaps and redo some areas of my plan, then have codex implement the final revised plan it's allowed me to use my Claude in frugal way
2026-04-16 16:51:25
2
jplpagan :
it's wild to me that we're in an era where the (seemingly) most advanced technology keeps shipping without a user manual, and not only that, keeps changing how it works, usually without explanation. instead we're all chatting to each other about what we have discovered or think we have discovered, which is also constantly evolving.
2026-05-04 13:51:43
1
Twister Tipster :
I hit my limit in Claude instantly and on the paid plan - frustrating
2026-04-17 07:19:18
3
Doug :
This is very helpful. How do you avoid Claude forgetting everything? I have resorted to this prompt. Generate a full session handoff document I can paste into a new Claude chat to resume exactly where we left off. Structure it as follows: System State — current values of every tracked signal (scores, flags, blockers, staleness). Be specific with numbers and timestamps, not summaries. What Was Built — completed items only, each with a one-line proof of completion (log line, exit code, confirmed output, etc.). No items without proof. What Was Discussed But Not Built — unfinished items with the exact reason each was deferred (missing precondition, time, risk, etc.). Direct Plan of Action — ordered steps to resolve every open item, taking the most direct path. Each step must include: the exact commands to run, the expected output that confirms success, and what it unblocks. No placeholders — expand all paths and variables. Permanent Rules — rules that must be enforced every session without exception. Write this for someone who has no memory of the session. Every claim must be grounded in actual observed state, not inferred or assumed. If something is unknown or unverified, say so explicitly.
2026-04-17 05:21:10
1
cory gibson :
ChatGPT has the same context rot. At first I disliked the whole token usage model when I started using Claude. Now I feel it has conditioned me to use all LLM more purposefully. Fitting my interactions closer to the architecture of the technology. I have a Claude Max and I have noticed if I use /handoff after about 5 prompts, copy that into a new chat I use far less tokens and get really good results. I’m constantly hitting my session limit about 15 minutes before it resets now which is isn’t bad. I just take a 15 minute break, come back to a fresh session and repeat.
2026-04-17 05:36:40
0
Broward Bark :
I’m using the $100 plan for coding and it is working great! I’ve done some pretty impressing plugins in size.
2026-04-16 20:36:24
0
Kenny R :
I love Claude, but I’ve never hit my limits with chat gpt
2026-04-16 17:21:24
2
gutterwalker :
you need to refine your workflow talk to chat and establish what you want to say have chat make it more concise and you need to start using pictures chat can make a diagram for your builds that's very simple looking to you but says thousands of things to another AI be specific with your language and you won't have these hurdles start using pictures and you will totally revolutionize not only the way you use AI but the way you fill up the context
2026-04-30 13:55:12
0
KVruno :
I’ve never hit my limit on chatgpt
2026-04-16 21:59:30
0
Fed :
I’m more confused now. I thought I understood it. What changed in the last month?
2026-04-17 02:20:39
0
wheres.jared :
It makes me not want to use it. I know Claude is better for coding but I can barely get a plan made before running out and I have pro. Codex you can actually get things done.
2026-04-16 14:13:41
1
This MF'er :
After listening to all of this, I understand why you hit your limit.
2026-04-17 02:45:07
0
Carmen :
Thank you for explaining this! I couldn’t understand how I only get one interaction every 5 hours with Claude but I can go back and forth with chat for hours before I hit
2026-04-16 23:50:25
0
mojocat01 :
Super interesting and helpful actually. I never knew this.
2026-04-16 19:16:49
0
Colton_Wade 🇺🇸 :
The practical context window limit is smaller than the advertised limit because some of the window is already spent before your own prompt fully arrives. The cleaner your context is, the more of that budget stays available for the work that matters. That is why context hygiene helps: keep only the load-bearing facts, avoid dragging old chatter forward, and use authoritative handoff file(s) so each new session loads a compact, intentional state instead of a messy conversation history. The same applies to tools: use the fewest MCPs possible, because every attached tool can add instructions and overhead; when a CLI equivalent exists, it is often leaner and easier to control, so it preserves more usable context for actual reasoning.
2026-04-18 11:21:23
0
London Edwards :
Thank you for this information!
2026-04-18 05:22:39
1
🌲Bryan in the Northwoods🌲 :
thanks!
2026-04-17 00:23:58
1
Wolfpack :
Thank You!
2026-04-17 08:35:56
0
Andy Magnus :
Do your huge verbal thinking in a planning session, then ask it for a handoff to a cowork (or code) session. Huge unlock in the Claude system for what you just described as your workflow.
2026-04-27 12:11:56
0
cory gibson :
Man go down the rabbit hole of performance per percentage of context window at your own peril.
2026-04-17 05:28:38
0
paleoheritage :
Great TikTok! Expected you to have 10k followers! Keep it up, please!
2026-04-17 01:07:47
1
Andrew Begnoche :
😂😂😂
2026-04-17 05:04:20
0
To see more videos from user @ai.field.notes, please go to the Tikwm
homepage.
