Not a fan of their context compaction and I feel like 1M-token context should minimum today. Each day I see how GPT 5.5 and 5.6 struggle a bit after each compaction before they get to the full speed, sometimes focusing too much on some older steering message that made it into the compacted context.

▲

rbranson 20 hours ago | parent | next [-]

Note that GPT-5.6-Sol is ~2x as token efficient as Opus/Fable, so a 258K max is equivalent to ~516K on Claude models. You also may just be fooling yourself, as context rot is still a problem[1][2] and there's evidence compaction can be equivalent or better than longer context windows for agentic tasks[3]. Obviously it would be best if the models could reason over 1M context like they do over 256K context, but they can't yet.

[1] https://arxiv.org/abs/2605.12366 [2] see F1 of GraphWalks 256K vs 1M in Opus 4.8 System Card: https://www-cdn.anthropic.com/0b4915911bb0d19eca5b5ee635c80f... [3] https://context-folding.github.io/

▲

bigmadshoe 19 hours ago | parent [-]

The improved token efficiency would only account for the output and reasoning tokens occupying context, right? What about the user messages and other context like file contents? I’m not sure how these measure up in a long conversation

	▲	rbranson 15 hours ago \| parent [-]
		For agentic use it impacts input tokens because the output is fed back into the model as input on subsequent turns.

▲

jubilanti a day ago | parent | prev | next [-]

The fact there is no way to disable auto-compaction like every other coding harness makes me scream. It fires randomly at 10-20%, so really you now only have guaranteed context of 272k * 80%. So many times I've had it work on a large codebase, it is almost done with a task, it needs maybe a 2000 token response, but it drops below 20%, it churns and churns and churns and then I see the dreaded ``Context compacted`` and scream. You can't go back in the history to the state of the conversation before the compaction. So then it has to investigate the codebase again gets to 20% and then auto-compacts. So it then has to read more, and then auto-compacts again, and then I'm out of tokens.

▲

n4te a day ago | parent | next [-]

Use pi. Can't trust OAI or Anthropic to make good tools.

▲

solenoid0937 a day ago | parent [-]

Claude Code is much nicer than Pi, especially for long running autonomous work. Pi might be OSS but it's just not nearly as featureful or agentic as Claude Code.

▲

EsotericSoft 21 hours ago | parent | next [-]

Pi can do all the same that CC can. OSS is nice but the reason to use it is to control what is happening between me and the LLM. See: OP and many others for many reasons, bloated system prompts, the harness changing day to day, Anthropic shipping terrible and untested [features](https://news.ycombinator.com/item?id=48947776) that don't appear in the changelog, Codex ultra subagents nuking usage, etc. You can tell which harnesses are vibe coded.

▲

JLO64 20 hours ago | parent | prev [-]

I agree, but when it comes to the Pi defaults. I’ve spent a good deal of time (admittedly via vibe coding) writing extensions that completely alter the look/feel as well as functionality of Pi.

I view Pi as similar to NeoVim, both programs have kinda crappy defaults that nobody should really use but are incredibly extensible and customizable for power users willing to invest their time into them.

	▲	tlamponi 20 hours ago \| parent [-]
		Interesting comparision with NeoVim, given that it was basically created to provide a better default (config) experience than classic vim. We might be close to PostNeoVim (or is helix [0] already that?). [0]: helix: A post-modern text editor. https://helix-editor.com/

▲

bredren a day ago | parent | prev [-]

Pardon the plug, but I have built a tool that ingests the state of conversations to a local DB in realtime. It has both macOS and Linux clients.

When context compaction introduces a gap, I use the /total-recall skill to pull prior turns back into context and off it goes.

The tool is free for personal use and has a source available local cloud option to sync convo histories across multiple machines.

http://contextify.sh/docs

▲

iugtmkbdfil834 a day ago | parent | prev | next [-]

Full agreement. I sincerely hope that reducing token is mostly a cost saving measure and not some roundabout way to get more token use. It still does not help. In my corp, bean counters rendered initially useful internal llm largely pointless as they severely limited context.

I swear there is a executive convention out there sharing worst practices.

▲

Sammi a day ago | parent | prev [-]

Just use md files for working memory. No need for large context. LLMs get dumber as you put more in context, as it stretches their attention. Keeping context small is better for quality.