DZ

Creative-Driven AI-Assisted Programming Practitioner Focus on mobile internet technology iOS tech manager Mobile game developer

[DZ @prodbitz]

这几个月高密度的连续使用了 ClaudeCode, Cursor, AugmentCode, Cursor, Windsurf, Trae, Codex 之后。我得出 Coding Agent 的能力计算公式 = Agent x LLM

总的来说 ClaudeCode 比 Codex 强太多了，虽然最近很多人狂吹 因为Codex 和 ClaudeCode say good bye. 但是....

[DZ @prodbitz]

@OsaurusAI 大家对gemma4 的要求太低了，能写文件只不过是蹒跚走路而已。你们是在嘲讽Google吗

[DZ @prodbitz]

@md_kasif_uddin macOS for creation，Linux for running stablely，Windows for shitting

[DZ @prodbitz]

@shzw47yzs6 @justbyte_ Exactly

[DZ @prodbitz]

@barckcode 本质上只是个 客户端而已，ChatBox + Tailscale 也可以做大

[DZ @prodbitz]

@twtayaan 癌症变异而已，敬而远之最好

[DZ @prodbitz]

@kis 这不值得惊讶，MoE 架构都可以使用全尺寸的 50% 内存，高速运行。

[DZ @prodbitz]

@bnjmn_marie @WaleedAhmad1a10 gemma4-12b reach 49tps on my 4070ti, waiting your MoQ to increase 30 percent

[DZ @prodbitz]

对设备应该有要求吧

[LM Studio @lmstudio]

Meet LM Studio's mobile app.

Your local models, now in your pocket.

[DZ @prodbitz]

@huacnlee 没事不用windiws

[DZ @prodbitz]

@Its_Nova1012 Prefer to design A

[DZ @prodbitz]

@TheGeorgePu Cannot compare each other, OpenAI didn't develop 120b further more.

[DZ @prodbitz]

@valigo 他们就是想怎么恶心怎么来

[DZ @prodbitz]

@bnjmn_marie mtp 对于资源不富裕的系统来说是个减分项

[DZ @prodbitz]

@atomic_chat_hq @atomicbot_ai Looks like 26b-a4b is more better cause MoE

[DZ @prodbitz]

@HowToAI_ 你们忘了cuda 是nvidia 创造的吗？ai 可不会创造没见过的东西

[DZ @prodbitz]

@shub0414 微软要开始封锁他的 github 账号了，你也得小心，因为你转发了

[DZ @prodbitz]

@barackomaba I use mulder‘s uncensored apex balanced model,fast and high quality

[DZ @prodbitz]

@mark_k too late，you lost us，we dont need windows not only wsl

[DZ @prodbitz]

任何庞然大物都是由这种有机机器人的移动带来行动力，我们何其幸运

[cumi @sambellcumi]

Whenever you feel sad, remember that there are tiny kinesin proteins trying their best to make you happy 🤓

[DZ @prodbitz]

@JusBili @tunahorse21 你在用Windows 羞辱codex，它不快乐，so …