beta_pyxis

curious about all things code

[beta_pyxis @beta_pyxis]

@ziv_ravid This is not LinkedIn. I really don’t get why people do such posts

[beta_pyxis @beta_pyxis]

@whinhosa_sh @shydev69 And fetched all the cans from dustbin to put on the table

[beta_pyxis @beta_pyxis]

@cerebras i thought I was reading 750tps announcement of Sol

[beta_pyxis @beta_pyxis]

@theo @11MMORPG He’s using Gemini to write his tweets that’s how

[beta_pyxis @beta_pyxis]

@zats Obviously I’ll be reading and manually approving every rm rf from now on, thanks to this lol

[beta_pyxis @beta_pyxis]

@thsottiaux

[Glavo @GlavoMeow]

我发现电脑上 .codex 文件夹占了超过 400 G 硬盘，调查后发现我每次发给 Codex 一张图片，它都会用 Base64 编码存到 .codex/sessions 的 jsonl 文件里；每次上下文压缩，它都会在同一个文件内把这些图片全部重新复制一遍；每次开子代理，都会把父线程的日志复制一遍。

我在一个会话中发了 57 张图片，触发了 1052 次上下文压缩，于是 Codex 在一个日志文件里存储了 42895 张图片，其中有一张图片被复制了 1048 次；这个会话触发了 142 个子代理，每次开子代理就把父线程日志复制一遍，所以又把这几万张图片复制了 142 次，吃了我超过 400 GiB 的硬盘，一张图片能重复存储 15 万次😅

[Glavo @GlavoMeow]

我发现电脑上 .codex 文件夹占了超过 400 G 硬盘，调查后发现我每次发给 Codex 一张图片，它都会用 Base64 编码存到 .codex/sessions 的 jsonl 文件里；每次上下文压缩，它都会在同一个文件内把这些图片全部重新复制一遍；每次开子代理，都会把父线程的日志复制一遍。

我在一个会话中发了 57 张图片，触发了 1052 次上下文压缩，于是 Codex 在一个日志文件里存储了 42895 张图片，其中有一张图片被复制了 1048 次；这个会话触发了 142 个子代理，每次开子代理就把父线程日志复制一遍，所以又把这几万张图片复制了 142 次，吃了我超过 400 GiB 的硬盘，一张图片能重复存储 15 万次😅

[beta_pyxis @beta_pyxis]

@thsottiaux @Robertg761_ @thsottiaux

Please take a look

https://t.co/uWzbH6QFbj

[野生の男 @yasei_no_otoko]

I discovered what appears to be a massive bug in Codex.
I’ve been running Goal in a single session for over nine days, and the app’s implementation is progressing almost too smoothly.
However, during that time, I noticed that the storage space on my MacBook’s 1TB SSD was decreasing dramatically. Since I had originally stored over 200GB of Blackmagic RAW data on it, I assumed that Xcode’s cache or Simulator files were taking up the space.
But the cause turned out to be something I never expected.
A large number of JSONL files, each about 1.5 GB, are likely being created in the Codex “sessions” folder with every turn. Since the storage usage is increasing gradually, I suspect that the extremely long thread context is being duplicated in its entirety for every turn. Shouldn’t this be an incremental copy? This is a huge problem in this era of skyrocketing SSD prices.

[beta_pyxis @beta_pyxis]

If anyone from oai is on it then at least acknowledge it??

This is a very very serious issue. You can easily find multiple versions of it, ssd io is not optimised at all, not even for nongoal asks

@thsottiaux @OpenAIDevs

[野生の男 @yasei_no_otoko]

I discovered what appears to be a massive bug in Codex.
I’ve been running Goal in a single session for over nine days, and the app’s implementation is progressing almost too smoothly.
However, during that time, I noticed that the storage space on my MacBook’s 1TB SSD was decreasing dramatically. Since I had originally stored over 200GB of Blackmagic RAW data on it, I assumed that Xcode’s cache or Simulator files were taking up the space.
But the cause turned out to be something I never expected.
A large number of JSONL files, each about 1.5 GB, are likely being created in the Codex “sessions” folder with every turn. Since the storage usage is increasing gradually, I suspect that the extremely long thread context is being duplicated in its entirety for every turn. Shouldn’t this be an incremental copy? This is a huge problem in this era of skyrocketing SSD prices.

[beta_pyxis @beta_pyxis]

@iuditg Which plan are you on? I really hope it’s not 200 one

[beta_pyxis @beta_pyxis]

@nunezvice @OpenAI can we not choose a default model? or voice only works on GPT 5, if yes, is there no option to spin up a reasoning agent with some other model selected?

since we'd not be using that for voice, but plain reasoning, shouldn't it be possible to do that?

[beta_pyxis @beta_pyxis]

@thsottiaux @theo Please port it into codex

[Kimi.ai @Kimi_Moonshot]

Kimi K3 has received far more love than we expected, and our GPUs are feeling it.

Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing compute for current members. Existing subscribed users are not affected.

We're adding capacity as fast as we can and will reopen new subscription spots in batches.

Going forward, we'll also split membership into two more focused plans: Kimi Membership for Kimi Web, App, and Work; and Kimi Code Membership for coding workflows. This will help us match compute more precisely and keep the experience stable.

Thank you for your patience and understanding!

[Haider. @haider1]

PREDICTION

after kimi k3 successful launch and mogs opus 4.8, Dario Amodei is gonna write an essay explaining how open-weight Chinese AI threatens democracy, national security, and the future of humanity

[beta_pyxis @beta_pyxis]

@thsottiaux No resets for business plan people? Not even a single one??

[beta_pyxis @beta_pyxis]

@BraedendotTECH Omg I’ve having the same exp

I’ve added lines in all my agentmd that’s not to reply with more than 15-20 lines unless i explicitly ask for it

I have dedicated review agents so i don’t have to read anything.

I rather come back and try to fix a bug than review it firsthand

[beta_pyxis @beta_pyxis]

@maria_rcks Exactly.
I’d die on this hill that I’m making this ‘tweet’ on ‘Twitter’ and my ‘codex’ and chatgpt are not the same

[beta_pyxis @beta_pyxis]

@noahzender @tibor_tee Why not do a collection of such prompts a standalone feature?

For starters, keep this prompt marketplace seeded with prompts that cursor devs feel are really good, can accept user feedback and add ranking on it later on as well

[beta_pyxis @beta_pyxis]

@skeptrune Biggest issue with ai assisted review is, intent is missing. It can check the syntax validity without evaluating if said change is even needed.

[beta_pyxis @beta_pyxis]

@steipete What’s your review of sol 5.6 frontend capabilities? Assuming you used that for the facelift?