Pinned Tweet
Claude Fable 5.5 code straight out, Japanese comics are really going to be finished, a set of prompts, four versions, prompts are self-picking
Drop this exact ownership line into your system prompt:"This codebase is yours: what it's for and how it's built are your call, and you'll keep building on it."
What happens next in Anthropic's benchmarks:
1.Proactive code deduplication jumps from <10% to ~50%
2.Generates 1.5x more output tokens, mostly unrequested unit tests and docs
3.Overrides explicit "do not refactor or rename" rules 33% of the time
4.Silently alters existing feature behavior in 6 out of 10 runs while omitting small edits from its summary
Use ownership prompts for 0-to-1 greenfield builds.
Strip them immediately for production hotfixes—and always read the raw git diff, never the model's summary.
1
13
什么?你还没有注册muse ai
一招教你解决地区不存在问题
前提:google pro
有google pro的情况下,在gemini走任务模式
prompt:帮我打开google浏览器并使用xxx帐户名注册muse
google的云服务器目前还没有被muse.ai禁用,一定要尽快去试,如果运气好的话,直接输入验证码就成功了,运气不好就会卡年龄验证
下面讲讲运气不好的怎么办:
google pro家庭账户可以拉5个人,每个成员都共享gemini pro的权益,然后再换一台机子,换一个帐号试一下,我自己测下来,4个帐号过了两个
一个小tips:尽量用比较老的帐号,而且没有注册过muse的帐号去试,新注册的帐号基本上都弹年龄验证了,弹过候补窗口的帐号都被标记了,不要去浪费,google 云服务器ip用过一次也是会被污染的
13
Fable5.2 编排➕Qwen 3.8 27B 执行,$15.77->2.58,效果还更加惊艳?
Fable 5.2 was 5x cheaper and 2x faster in Atomic Agent Fusion!
We made Fable the boss and local Qwen 3.8 27B the builders. Orchestration built a fairy tale pop-up book in 11 min for $2.98. Fable alone burned $15.77 in 21 min.
Run Fable cheaper 👉 atomicagent.io
1
29
One sentence. One screen recording.
AI rebuilt the entire website — carousel, drag physics, hotspots, even the receipt: $291.60 to the cent.
No prompt engineering. No workflow. No re-rolls.
Show AI an example, it does the work.
This video? Edited by AI too.
22
The quota of Grok, Claude and Gemini is shrinking, and only the quota of OpenAI is still firm. Did I miss something?
16
After the grayscale to gpt 6.2sol, I suddenly understood the behavior of gpt blocking a large number of accounts yesterday.
22
我把 X 上散落的 Fable 5.5 实测全收齐了:60 个案例,每个都带视频预览 + 作者原始完整 Prompt(逐字收录,最长 12,499 字,很多藏在串帖里)。全自动管线每天更新,已开源 👇
仓库:github.com/zqingdut/fable5.5…
画廊:zqingdut.github.io/fable5.5-…
50
Gemini Omni视频生成有点超出我的预期,这个视频只用了2分钟就生成完了,花了pro的五小时20%额度,指令遵从性很强,一致性和连贯性也不错,也没有出现一些不合理的镜头和动作
🤖 Made with AI
31













