BlazeZhang

BLAZE｜腾讯 AI 产品经理(前)｜AI 初创公司 GTM 负责人
AgentSkillLibrary 作者｜AI 教育与应用实践者｜AI 产品与商业化观察者
从产品开发、市场进入到教育落地，探索 AI 如何进入真实的工作与学习场景。分享 Agent、Skills、AI 原生工作方式与商业化实践

[BlazeZhang @QihangZhan4885]

@RunOnFlux Good point. A 3 a.m. run shouldn’t depend on my laptop being awake. A versioned repo seems like a good starting point, with agents fetching the skills and references they need. How do you handle updates for scheduled agents?

[BlazeZhang @QihangZhan4885]

https://t.co/VdnXYNq9Ai

[BlazeZhang @QihangZhan4885]

@OpenAI 多年以后，OpenAI 面对无数被它取代的人类智性活动，将会回想起它一次性终结四百个四级数学成果的这个遥远下午。

[BlazeZhang @QihangZhan4885]

感觉dot被远远低估了...................................

[BlazeZhang @QihangZhan4885]

https://t.co/YgEk7jO9qs

[BlazeZhang @QihangZhan4885]

@AlanShao111 1580

[BlazeZhang @QihangZhan4885]

@XuchenWAmbn8cu 😆牛逼等我给你点一遍

[BlazeZhang @QihangZhan4885]

https://t.co/jec2widOLB

[BlazeZhang @QihangZhan4885]

Tibo, spare us the bullshit about “models getting more efficient and prices going down, so you actually get more work done.”Every single competitor is doing the exact same thing. Claude improves their models, drops prices, and still keeps the 20x. They don’t turn around and say “hey, because we got better, we’re now cutting your usage in half.” That’s not how this industry works.Efficiency gains and price reductions are the baseline of AI competition. You’re supposed to pass those gains to users as more value, not use them as an excuse to slash the quota and then act like you’re doing us a favor. You’re literally saying: “We made progress, therefore we get to give you less.” That’s completely backwards. If your models are better and cheaper, the normal move is to increase the allowance or keep the old value. Cutting usage because you improved is the logic of a company that wants to extract more margin while pretending it’s “transparent.”Other labs improve and users get more. You improve and users get half. Don’t insult our intelligence by calling that “passing down the improvements.”😅😅😅

[BlazeZhang @QihangZhan4885]

@thsottiaux I want 20x！！！I cannot live without it！plz！

[BlazeZhang @QihangZhan4885]

@thsottiaux I want 20x！！！

[BlazeZhang @QihangZhan4885]

@XuchenWAmbn8cu 丢人的国家！丢人的股市！

[BlazeZhang @QihangZhan4885]

@XuchenWAmbn8cu 好友写的真好😆😆😆😆😆

[BlazeZhang @QihangZhan4885]

GPT-5.6 handles complex projects far better than quick one-off answers. It fixed my PM knowledge base bugs and trimmed a bloated 685-page export into a tidy 178-page handbook with thorough self-audits.#AIProductivity

[BlazeZhang @QihangZhan4885]

@XuchenWAmbn8cu Bravo！Dude, your writing is next level! Way better than most stuff I read online😍

[BlazeZhang @QihangZhan4885]

@thsottiaux Please fix the Codex logging bug soon!
It's writing TRACE logs at ~5MB/s to ~/.codex/logs_2.sqlite, potentially ~640 TB/year and killing SSDs.
Main issue: https://t.co/1D1lcDMeXJ
Thanks!

[BlazeZhang @QihangZhan4885]

@wanderingswann @OpenAI me2😅

[BlazeZhang @QihangZhan4885]

why now codex is so stupid and slow？？

[BlazeZhang @QihangZhan4885]

@thsottiaux Haven't you seen the numerous users on GitHub, X, Reddit, and Rednote complaining about slower speeds and declining model quality? Fix this issue as soon as possible, please!!!🥲🥲🥲🥲🥲