Introducing Grok Bot, now in early beta. Bots are AI teammates that do real work for you. They sign in to your tools, use them just like you do, and come back with finished work.

A Note to Our Users Manus will soon resume operating as an independent company. As part of this transition and to comply with regulatory requirements in specific jurisdictions, some users will need to back-up their data before 7:59 a.m. on August 23, 2026 (SGT) and then begin restoration starting from 8:00 a.m on August 25, 2026 (SGT) to ensure normal usage. More information can be found in the blog post and we’ve provided detailed guides, including to understand if you’re affected, in our Help Center For affected users, these are the key details: We’ll notify you by email and in the Manus app. If you signed up with an Apple ID or Facebook account, please check your in-app notifications. Back up your data from now until August 23 (SGT) before 7:59 a.m. You can back up more than once, you can create new data after backing up, just run it again. Starting from 8:00 a.m. on August 25 (SGT), the restoration portal will open. Restore your data and pick up right where you left off. Affected users will not be charged during the transition, and we will be providing welcome back bonuses. If your account is not affected, you don’t need to do anything - keep using Manus as usual. You’ll get an in-app notification confirming this. For those affected, we know this is disruptive and we’re sorry for the inconvenience. Supporting you through it is our top priority. In addition to the full guides in the Help Center, our customer service team is ready to help 24/7 with any questions. As we look ahead, we couldn’t be more excited by the future. We’re preparing a series of new features that will push the boundaries of what’s possible for general AI agents once again. Thank you for being with us as we go through this process together. The best is still ahead.

用pi统计了各主流 harness 的缓存命中水平，这里是基于我的实测数据分析的，在pi当中甚至没有使用skill 来分析。开始的识别有一点点错误，一开始的统计见图二。排在最后的是Claude Code、Kimi Code 。 pi、Command Code、ZCode，分别占据前三。接下来的DimCode、Grok Build 则是总量最大的里面，缓存命中率最高的。可以对比我之前的分析，看一下。pi、Command Code 都相当优秀 这个缓存命中水平不出我的意料。DimCode、Grok Build也各有特色。

梁圣的含金量还在上升！！！ 桀骜不驯的Claude也开始降价了！！！ 比智谱GLM 5.2都便宜了！！！ 跑 agent 的、用 Claude Code 的、token 按百万烧的， 这回真的要省钱了！ 原计划 9 月 1 号涨价 50%， 今天Anthropic说：不涨了，永久 $2/$10， 对 API 开发者和 Claude Code 重度用户， 这是今天最实在的新闻。 Claude Sonnet 5，能力接近 Opus 4.8 的那个中间档模型，优惠价直接变永久价。 刚用黎曼假设证明了能力天花板，然后就把主力模型的价格焊死了，一边证明我能做多难的事，一边告诉你用我做多便宜。

Pi ships with 4 built-in tools: - read - bash - edit - write These tools are mostly all you need for an effective coding agent. Models know how to use bash and can do more than you think with the basics. Here's the code to add these primitives to your app using Pi’s SDK.

现在给国产大模型厂商就三条路，第一条：这时候就别藏着掖着了，直接把牙膏挤爆，压箱底的大模型赶紧端上来拼命。第二条：和 DeepSeek / OpenAI 打价格战，往死里打，各种渠道免费送。第三条：死

the average OpenCode Go user spent $1.14 per day on deepseek flash v4 this past week the dual DGX setup people are running to do the same costs $10,000 it takes 24 years to break even at 10x the usage it takes 2.4 years