SimonZhao

AI algorithm engineer | AI startup founder focused on AI agents.

Personally following the AGI journey.

[SimonZhao @SimonZhaoAI]

@batuhanppn 4 or -5

[SimonZhao @SimonZhaoAI]

@batuhanppn -5 or 6

[SimonZhao @SimonZhaoAI]

@congyuecy_soft 请问用的什么模型？

[SimonZhao @SimonZhaoAI]

@rohanpaul_ai pass@k better than pass@1.

[SimonZhao @SimonZhaoAI]

@CastilloLe58494 you and your mother are so ☺️

[SimonZhao @SimonZhaoAI]

@davi99600 @shachepi @Raz_09_ claude界面不一定用的claude模型

[SimonZhao @SimonZhaoAI]

@colemcfaul @AnthropicAI Once AI becomes an indispensable factor of production for humanity, ideology will indeed constrain its development. This constraint is a limitation of thought, not of cognition. What Harness needs is to serve concrete work, rather than specific vested interests.

[SimonZhao @SimonZhaoAI]

@AnthropicAI Once AI becomes an indispensable factor of production for humanity, ideology will indeed constrain its development. This constraint is a limitation of thought, not of cognition. What Harness needs is to serve concrete work, rather than specific vested interests.

[SimonZhao @SimonZhaoAI]

@lxfater 分数5.5更高点，但是10次成功次数mythos高一点，mythos3/10，5.5 2/10

[SimonZhao @SimonZhaoAI]

@Lonely__MH 这个code areana benchmark主要测评的是前端构建能力，不能覆盖其它代码场景，所以gpt-5.5都进不了前三

[SimonZhao @SimonZhaoAI]

@vikingmute 现阶段复杂问题，复杂项目，codex(gpt-5.4)的稳定性，可靠性，正确率优于cc(opus-4.6)

[SimonZhao @SimonZhaoAI]

Based on recent vibe coding practices, I’ve compiled 5 frequently used skills for your reference. I especially recommend "sdd-plan-maintainer", which I use frequently. GitHub project link: https://t.co/ymsdaa2g2h

[SimonZhao @SimonZhaoAI]

模型能力越强，创造的价值也越大，但伴随的人类社会发展的危险性也越高，两家top AI公司员工在这个时候选择支持“不妥协”的athropic，这对AI的治理，发展，普惠而言，无疑是一个有温度的积极信号，We will not be divided.

[SimonZhao @SimonZhaoAI]

OpenAI和Google的员工联名发起支持anthropic的活动，anthropic在与美国国防部签署合作协议时因坚持两条原则而面临被终止合作，它们坚持的两条原则是:https://t.co/ofJe9ulFFE不能用于大规模国内民众监控；2.禁止��AI用于“完全自主武器”（Fully autonomous weapons）。

[SimonZhao @SimonZhaoAI]

I like this year poetry.

[SimonZhao @SimonZhaoAI]

gemini-3-flash-thinking - the only model so far that can correctly count six fingers.

[SimonZhao @SimonZhaoAI]

@haider1 opus 4.5?

[SimonZhao @SimonZhaoAI]

@frxiaobei 类比自动驾驶L3，人类要随时接管，复杂一点的任务，现在的模型还不支持长时自动化。现在的模型偏向over-confidence(也有对齐尽量解决问题的原因)，无法识别哪些任务自己无法完成，需要人类接管。所以大多数是拆解模块化，文档驱动(SDD)模式开发，并行化在复杂任务上不现实，会让模块边界更不清晰。

[SimonZhao @SimonZhaoAI]

@YonglongT Congrats! I was wondering whether GPT-5.2 has a similar parameter count to GPT-5.1, or if it is larger. If you’re able to share, I’d really appreciate your insights.

[SimonZhao @SimonZhaoAI]

@karpathy I think DeepSeek-OCR mainly verifies the effectiveness of compressing text within images for OCR tasks, but it doesn’t prove the effectiveness of general image-to-text compression. if this approach works on broader tasks like image captioning, can demonstrate truly effective