Follow AI Builders — Not Influencers

BREAKING: Blade confirmed to be behind the recent cyberattacks on french blood banks it turns out vamps are the real AI alignment villain, and wesley snipes is our only hope

不少群友新注册Twitter、Facebook、Github、Vercel、PayPal等平台账号，都遇到过，注册后立马被封号的情况。 这种时候不要急，我们可以从对方平台的角度思考一下，为什么他们的风控策略要做成这样。 目的是什么？ 目的肯定是防止别人用程序批量注册海量账号。 注册好了之后，立马封号，让你做更多步骤的验证。 就可以快速把真实用户和批量脚本注册账号给区分出来。 不要小瞧这个策略，仅仅是多了这一步，就可以吓退很多人。 大多数靠脚本批量注册的人，是不会费时费力一个一个账号去申诉解封的，只有真人真实注册的，才会花这个时间成本去做。 所以他们这一套注册完就封号策略是很有效的，可以挡住绝大多数的程序脚本批量注册的账号。 当然，他们也会根据环境来判断，并不是每一个新注册账号都会被封号，只有疑似可能有问题的，才会封号，让你人工进行解封操作。 这时候，我们要心里有底，不要慌，按照平台给出的认证步骤一步一步往下走就行。 有些要求你手机安装 App 验证，有些要求你写一封信来证明你是真人，反正你就按照对方要求去做就行。 前几天飞嫂注册一个推特账号就是这样，刚注册完就被封号了，飞嫂以为没救了，但是在“见多识广”的哥飞看来，小菜一碟，按照要求，下载好 X App，登录上账号，人脸识别一下，账号就恢复了，App 里可以用，网页里也可以用。

如果你不是深度用户，你是不会发现一些不好用的点的。 微信聊天列表，私聊、群聊、公众号等等所有混在一起。 当很多人给我发了消息，我需要一个一个处理时，就有一个很不好用的交互了。 我只要回复了某一条消息，他就会被自动顶到聊天列表顶部。 假设有 100 条消息等着处理，不管我是从顶往下，还是从下往上处理，都会发现，我好不容易滑动到了某一条还没处理的消息，只要我一回复他，微信就自动把我带到了列表顶部。 我又需要滑动，滑动，滑动，一直滑动，找到一条没处理的消息，回复一下，被带到了列表顶部。 我又需要滑动，滑动，滑动，一直滑动，找到一条没处理的消息，回复一下，被带到了列表顶部。 我又需要滑动，滑动，滑动，一直滑动，找到一条没处理的消息，回复一下，被带到了列表顶部。 我又需要滑动，滑动，滑动，一直滑动，找到一条没处理的消息，回复一下，被带到了列表顶部。 我又需要滑动，滑动，滑动，一直滑动，找到一条没处理的消息，回复一下，被带到了列表顶部。 …… 就这样一步一步，终于把几十上百条消息处理完，这个交互太烦人了。 我没办法去求微信产品经理改交互，于是我就自己尝试用各种方法来解决问题： 1. 先给他在输入框里打字，但不发出去。的确，可以把所有没回复的都打好字，接下来我需要给这几十个一个一个发出去时，一样会被带到聊天列表顶部，方法无效； 2.点开一条未读消息，判断是否需要回复，如果需要，就给他置顶，一样有问题，一置顶就给我带到了置顶的顶部了，方法无效； 3.看到聊天列表里的未读消息，不点开，直接右键给他置顶，这时候由于我还没点开跟他的聊天对话窗口，所以不会被带到顶部，还留在我当前所在的列表位置，可以继续处理下一个。后面就是去置顶那里，处理完一个，就取消一个置顶，方法有效； 4、在聊天列表里，右键，选择独立窗口打开，一个一个待处理的窗口就会堆叠在我视野里，然后一个一个处理，处理完一个，关闭一个窗口，方法有效。 我们自己做的产品，如果不深度使用，也同样无法发现一些不好用的点，而这些点可能就会导致付费用户不再付费了。

`𝚟𝚎𝚛𝚌𝚎𝚕 𝚍𝚘𝚌𝚝𝚘𝚛` - recommend you ship doctor subcommands in your CLIs!

i got a mac mini for everything AI related. 1. codex remote > claude remote. claude through ssh doesn't work when you close laptop. claude remote control is too limited. 2. that said, opus 5.5 is a much more token-efficient daily driver. i've resorted to using it through the claude code cli through ssh/tmux

RT Sam Altman Re @birdabo i want you to have so much ram lets build more

RT Sam Altman People should always be in control of the future. OpenAI is against the notion that it would be acceptable for people to be relegated to a role of doing things that “AI systems think makes sense to reward in humans”. People should not cede control of the future to superintelligence, and that requirement should ultimately constrain AI progress.

i agree that vibe-coded solutions will not replace enterprise-grade saas solutions. however i do think saas margins will be compressed for a variety of reasons: * the value of these tools will be in the data layer, and not in the app/UI layer * it is much easier and faster for startups to launch competing products. coding agents give small companies insane leverage to compete against incumbents, especially as startups are built in an agent-native fashion from the ground up

There are two main types of benchmarks: 1. those that measure frontier intelligence (terminalbench, gdpval, etc.). accuracy matters more than cost. 2. those that measure specialized intelligence. the pareto curve of cost/accuracy more than accuracy alone. both are important, but they are distinct benchmark categories. * as (1) saturates, they become (2). * the largest frontier models at xhigh likely already saturate (2), but they do so at a cost/latency point that's completely price-prohibitive. * every user should define (2) for the tasks they care about the most. we're big fans of both. we have so far dedicated outsized attention to (2) especially related to document tasks. we fundamentally think users should have the most price-efficient tools to understand, extract, and read over their documents. we've built parsebench, extractbench for this purpose. we are starting to invest in (1) as well, because certain document tasks are so challenging that even frontier models struggle (e.g. see extractbench pro). we are continuing our investment in long-horizon workflow research at @llama_index

OpenDocRouter is an awesome playground for trying out different document OCR models: from open-weight to frontier. Check it out! https://www.opendocrouter.ai/