DataLearner

Codex usage on Plus: • GPT-6 Astra: 20.9% of the 5-hour quota per task • GPT-6.1 Sol: 6.1% per task That's just ~5 Astra or ~16 Sol tasks to exhaust the entire 5-hour allowance, ignoring resets. Unless the shared quota increases, Plus could become very limiting.

gemini-4-argon 我认为在文本咨询领域， 已经可以封神了， 去年十月 claude-4.5 ， 我是上架第二天开始喊这个 sonnet-4.5 是文本的新王吧。 今年这一次是 gemini ， 带给人前所未有的强大辅助思考能力， 一款类似于高达的认知外骨骼， agent和chatbot的近乎完美平衡融合model

Day 25/ We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today. We heard your kind feedback and won’t be reducing usage limits for Pro 200, so your Work and Codex usage will stay the same. We’re also going to continue offering extended limits for dots for now, and usage of your dot will still continue to be separate from your default usage. And we'll of course also keep shipping better models and other efficiency improvements that increase your effective usage over time.

升级 #IOS 27.2 + 国行账号，限制百度网盘运动 #传感器 权限后，百度网盘还是能跳转摇一摇广告？网友实测禁用传感器权限也不能阻止摇一摇广告继续跳转，其他网友猜测百度网盘可能设置了随机延迟时间，到时间后无论是否摇晃手机都会触发跳转。 最后网友还是在百度网盘、我的、设置、隐私与个性化设置、个性化管理、广告兴趣管理里，手动关闭了摇一摇广告展开权限。

Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO. Read more: anthropic.com/news/s1-sec Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.

Built it in ~2 days. Started with Claude Code + Fable 5.1 and somehow burned ~$80, then moved most of the work to Codex + GPT-6 Astra Low (~10M input tokens). ChatGPT did the artwork, Gemini did the music. visit: llmlearner.com/graveyard

Been using Codex heavily for 11 days, and GPT-6 Astra has been better value than I expected. I mostly run Astra on Low/Medium, while Sol is mostly High. Astra: ~$0.17 per request / Sol: ~$0.155 Astra Low is fast, good enough for a lot of coding work, and the cost isn’t crazy.

Demand for Astra is really unprecedented. We're pulling all the levers possible to sustain the demand, but I've not seen anything like it until now and we went through very steep growth before. Priority will always be to keep excellent service for existing users, but we might have to pause new Pro subscriptions for a bit if this continues.

It seems that Gemini 4.0 is coming soon, considering gemini pro has not updated for a while, so it will be gemini 4.0 pro?

GPT-6 Astra burned 90% of my Codex 5-hour limit in ~11 minutes. The weird part: 97.8% of the input was cached, and it generated only 11.3K output tokens. So I dug into my local Codex history to see what actually happened. Two runs stood out.

So does Tibo hint that the Codex will have a new reset in a while?

Another implication: frontier models can be useful internally well before launch. GPT-6 Astra-class RL was active by July 20; GPT-6 Astra launched Sep 3. That’s ~6 weeks from documented internal RL/safety work to rollout—not “training finished,” but a short deployment window.

目前为止我还没觉得 GPT 6 Astra 在开发方面比 Fable 5.1 更强，差不多是 Fable 5 的水平，但是速度很快，比 Fable 耐用。 这其实带来的优势很明显，首先绝大部分场景不需要 Fable 那么强的智能，但是 Opus 5 又不聪明又慢。而现在 Astra 就是绝大部分任务表现的足够聪明胜任，速度又快，而且 Token 消耗没有 Fable 那么夸张，加上 Tibo 时不时送重置卡，这套组合下来优势会相当明显。 另外还有 UI 方面 Astra 有明显提升，但是目前我测试下来还是不如 Fable，细节上很多问题，但是你纠正一下它还是能改好。 写作方面还没有深入体验，比 GPT 5.6 强。 综合来说我接下来应该会更多的使用 Astra 在大部分任务上，但是更复杂的任务或者是 UI 设计产品设计的任务，还是会优先考虑 Fable 5.1（如果还有额度的话，太不耐用）。 当然这只是我目前为止的结论，我也没去测试做游戏和3D，那些不是我目前关心的，估计后面还会有些修正。

Benefits All