@AnthropicAI should add this feature!
Context rot is the silent killer of long AI coding sessions.
The model gets slower, dumber, more forgetful and you lose your whole thread when you finally hit reset.
So I built a “session handoff” into Claude Code. One command writes a clean summary of where I am; one /clear starts a fresh window that auto-loads it. No lost context. No re-explaining.
The best part: I didn’t write a single file myself. I just pasted this prompt into Claude Code and let it build the whole thing 👇
«Set up a "context handoff" workflow in this repo so I can escape context rot
without losing my place. Build these three pieces:
1. A slash command at .claude/commands/handoff.md called /handoff. When I run
it, you write a structured markdown summary of the CURRENT session to
.claude/handoff.md, with these sections: Current task, Key decisions made,
Files touched (with paths), Next steps, Open questions/blockers. Keep it
tight and high-signal — it's a briefing for a fresh session, not a log.
2. A hook script at .claude/hooks/load-handoff.sh that reads .claude/handoff.md
(if it exists) and prints JSON to stdout in this exact shape:
{"hookSpecificOutput":{"hookEventName":"SessionStart","additionalContext":"<file contents here>"}}
If the file doesn't exist, print nothing and exit 0. Make it executable.
3. A SessionStart hook in .claude/settings.json with matcher "clear" that runs
that script, so every time I /clear, the new session auto-loads my handoff.
Merge into existing settings.json — don't clobber other hooks.
Then add a .gitignore entry for .claude/handoff.md (it's transient), show me
the final files, and tell me the exact workflow: /handoff then /clear.»
That’s it. My new loop when a session gets heavy:
/handoff → /clear → fresh context, same thread of work.
One tip it taught me: try /compact first for normal bloat, it summarizes in place. Save the full handoff for true resets where you want an empty window but keep the plot.
Im the same dude! Was testing Sol on my plus plan today. Complete checkup of my codebase, implemented new festures in my app. Used Sol on xhigh and only spent 50% of 5h limit. Ive upgraded to pro 5x to use it as daily driver now. People dont know how to develop with ai, this is the problem i see with everyone complaining about limits!
I also got max 20x with anthropic, but will downgrade as i do t need that high limits when Fable 5 gets disabled again.
@DavidOndrej1 Most people who dont know how to develop with ai experience the same issues. I did a complete checkup on my codebase today, did some fixing and added new features and only spent 50% of the 5h limits. This was done on the Plus subscription!
Access to Fable 5 is extended to July 12th!
Mabe @AnthropicAI is realizing that people will downgrade and run to @OpenAI !
Now just let us use our limits as we want! 50% on Fable 5 for the weekly limit is not enough when the model is this expensive!
@withLinda13 im also missing missing Sol, asked ChatGPT about it and was told that my account might not have gotten the full upgrade yet. i got Plus subscription, but Sol should still be available so just gotta wait
Hi @sama and @OpenAI team,
I've been testing the new GPT-Live feature. While the Norwegian language quality still has room to improve, the core functionality is genuinely impressive — especially the real-time interruption capabilities.
One feature I really miss is the ability to share a live video feed from my camera during a conversation. I can already upload static images, but live video would unlock much more powerful use cases.
For example, when using GPT-Live to train my dog, a live video view would let the model see what I'm doing in real time and interrupt me the moment I make a mistake — exactly like it already does during language practice. This kind of immediate, visual feedback would make a huge difference in hands-on scenarios and lead to much better results.
I think adding live video would be a natural and very valuable extension of the existing interruption features. It would turn GPT-Live into an even more useful real-world coaching tool.
Happy to share more examples or feedback if helpful.
Just tried GPT-Live on my iPhone. The Norwegian voice sounds super broken, like a foreigner desperately trying to learn the language. 😬 Not impressed at all. Hoping it gets better soon.
Gemini nailed natural-sounding voice a long time ago.
GPT-Live is now fully rolled out to all ChatGPT users on Go, Plus, and Pro plans. Free user rollout is in progress.
Update to the latest version of the ChatGPT app on iOS or Android to try it out.
@elonmusk@SpaceXAI Please fix compliance approval for Grok 4.5 in EU and EEA. Being Norwegian i'm feeling left out as i cant use Grok 4.5!
I need my fix xD
@AnthropicAI Where is our limit reset? everyone's pushed hard to spend limits for the OG plan of disabling access of Fable 5 yesterday and now with the new deadline of July 12th we got no more limits!
@SpaceXAI is dropping Grok 4.5 tomorrow
@OpenAI is dropping GPT-5.6 tomorrow
I'd like @AnthropicAI to stay in this AI war, but without limit resets or even extending Fable 5 usage past July 12th most people will spend their money elsewhere, including me.
BREAKING: Elon Musk has just confirmed Grok 4.5 goes public tomorrow.
• Strong positive feedback from beta customers
• Opus-class model
• Faster performance
• More token-efficient
• Lower cost
@bridgemindai@Krazyb44 If grok 4.5 drops today we might see an ai war of releases! @AnthropicAI has no choice they need to reset everyones limits as most people pushed hard to spend every limit for yesterdays disable of Fable!
With GPT-5.6 dropping tomorrow we have some exciting times a head!