neteroster's recent timeline updates neteroster's recent timeline updates |
neterosterV2EX member #191331, joined on 2016-09-11 21:01:55 +08:00Today's activity rank 890 |
 |
Per neteroster's settings, the topics list is hidden |
Deals info, including closed deals, is not hidden
neteroster's recent replies
9h 21m ago Replied to a topic by 380cc › 问与答 › 求助 Gemini api 对 srt 翻译后造成时间轴错乱的解决方法 |
这些都有成熟工具了，LinguaGacha AiNiee 之类的可以看看
11h 40m ago Replied to a topic by andforce › OpenAI › 推理强度：最高，在 Mac 上是默认关闭的不可选 |
这个最高（实际上叫做 max ）比较容易过度思考，慎用，oai 隐藏不是没有道理。推荐日用 xhigh ，spec 实现 xhigh+goal
12h 7m ago Replied to a topic by burnsby › 程序员 › grok-4.5 真的比 gpt-5.6 sol 好用吧 |
grok4.5 特化有点太严重了，写的是常规、朴素逻辑还好，到稍微专一点的领域涉及一些决策就不行了（另一方面就是通用逻辑和推理能力甚至不如 glm ）。当个执行模型还算不错
Jun 18 Replied to a topic by bihui › 程序员 › 你们用啥 ai 编程工具可以当生产力? |
建议先了解 AI 的能力边界和常用 workflow 构成
Jun 12 Replied to a topic by haoofun › 程序员 › 订阅制还能持续多久呢？ oai 和 anthropic 的订阅制都是赔本赚吆喝吧？ |
虽然所有人都打满可能会赔钱，但拿 API 当成本显然不对，而且订阅制的假设就在于不是所有人都会打满
Jun 10 Replied to a topic by Jensond › macOS › macOS 27 已知问题汇总 |
@SuzhaharCan 不是全屏闪烁，就只是窗口边缘闪几秒，而且是从最小化情况拉出来才会这样，首次启动软件打开的窗口不会，就很神秘
Jun 10 Replied to a topic by Jensond › macOS › macOS 27 已知问题汇总 |
> 新的 Liquid Glass 外接显示器，外接 4K 显示器 UI 上有时候会出现窗口边缘毛边闪烁
我的 26 内屏一直有这个问题...
我的 26 内屏一直有这个问题...
Jun 9 Replied to a topic by hansonl › 程序员 › 1 个 gpt pro 账号有多少并发 rpm |
如果你说的是 codex ，很高，前段时间有人测过 pro 20x ，能打 50 并发以上，rpm 不太确定
Jun 6 Replied to a topic by xyz8899 › OpenAI › 总结一下这次 OpenAI 封号的问题，同时谈谈大家忽略的一个问题（可能这才是封号的元凶） |
有点幽默了，这次是纯误封，可以看 status.openai.com 和相关负责人 X 上的发言
Jun 5 Replied to a topic by w568w › OpenAI › 我的 GPT 5.5 怎么和你们的不一样？ |
@w568w 那我很怀疑是你的 harness 问题了，GPT 在各种 bench 和体验反馈都是执行强，deepswe 有案例分析，原文
```
GPT implements exactly what's asked
On DeepSWE, GPT-5.5 has the lowest rate of missing stated behaviors of any configuration in the chart; GPT-5.4 sits just behind it.
GPT reads the prompt and the visible repository contract literally, and produces a patch that honors both. The behavior is consistent across runs: when several GPT trials attempt the same task, they tend to converge on the same interpretation of the prompt, suggesting this precision is a stable trait rather than per-run luck.
A natural follow-up would be to examine whether this precision comes paired with related stylistic traits, like overly defensive code, surplus error handling, or other markers of a tightly instruction-anchored coding style.
```
```
GPT implements exactly what's asked
On DeepSWE, GPT-5.5 has the lowest rate of missing stated behaviors of any configuration in the chart; GPT-5.4 sits just behind it.
GPT reads the prompt and the visible repository contract literally, and produces a patch that honors both. The behavior is consistent across runs: when several GPT trials attempt the same task, they tend to converge on the same interpretation of the prompt, suggesting this precision is a stable trait rather than per-run luck.
A natural follow-up would be to examine whether this precision comes paired with related stylistic traits, like overly defensive code, surplus error handling, or other markers of a tightly instruction-anchored coding style.
```
Select Language