Neo Fixer

🚀 AI × Finance × Future
🤖 AI Product Intern @ Frost & Sullivan
📊 Math & Finance Student ‘27
🌎Sharing insights & connecting with builders

[Neo Fixer @fixer82601]

@Yufgvvvv 捧场来

[Neo Fixer @fixer82601]

@FiniYang 续火花

[Neo Fixer @fixer82601]

@lidangzzz 不应该是23级吗，气笑了

[Neo Fixer @fixer82601]

@dark_coderz Honestly, LinkedIn feels so performative.

[Neo Fixer @fixer82601]

Perplexity 怎么变成这样了？要是没人提起我都忘了还有这样一个东西🤡唯一优势是搜索，现在看也做得不怎么样

[Prompter @PromptLLM]

wtf happened to perplexity?

biggest fall from grace of a company ever

[Neo Fixer @fixer82601]

@PromptLLM Being early and impressive is not enough. Continuous innovation, product quality, and trust matter more than hype.

[Neo Fixer @fixer82601]

AI 正在从写代码走向“找代码里的问题”。

Kimi K3 声称 1.5 小时发现 Redis 8.8.0 的 19 个 0day，并生成 RCE PoC。

AI 安全智能体正在成为网络攻防的新变量。

[Chaofan Shou @Fried_rice]

Kimi K3 found 19 0days in latest Redis 8.8.0 in 1.5hrs.

Added 8.8.0 RCE exploit in https://t.co/ZOKA2wchLa

[Neo Fixer @fixer82601]

@psomkarbuilds parent

[Neo Fixer @fixer82601]

@Samaytwt ChatGPT has massive scale, but Claude's quality and thoughtful responses keep winning me over. Different tools for different needs!

[Neo Fixer @fixer82601]

@tongtongbee 不造神，就凭这点就该支持。

[Neo Fixer @fixer82601]

@AstraiaAI Qwen 3.6 27B is ridiculously strong. Every benchmark I throw at it, it just smokes everything else in that size range.

[Neo Fixer @fixer82601]

@ajs6888 🤝

[Neo Fixer @fixer82601]

Grok Build 搭配 Browser Use CLI，在浏览器任务方面的表现超过了 Codex 和 Claude Code。AI coding agent 的竞争已经从“写代码能力”扩展到了“真实世界操作能力”。Browser Use 这类工具让模型不仅生成代码，还能理解网页环境、调用浏览器完成任务，这正是未来 Agent 的核心能力之一。

[Gregor Zunic @gregpr07]

Grok Build, with the Browser Use CLI, outperforms Codex and Claude Code at browser tasks

[Neo Fixer @fixer82601]

AI 网络安全能力正在快速提升，部分先进模型已经展现出接近专业安全人员的漏洞挖掘和攻击能力。Mythos 和 “5.6 Sol” 等模型在 Cyber ECI（用于衡量模型开发软件漏洞利用能力的指标）上出现明显跃升，AI 在自主编写漏洞利用代码、攻击真实环境方面取得较大进展。

GLM 5.2 的网络安全能力被评估为介于 Claude Opus 4.5 和 Opus 4.6 之间，也就是说，它的网络攻击能力已经接近几个月前发布的顶级闭源模型水平；而 OpenAI、Anthropic 内部未公开模型预计可能进一步领先。Claude Opus 4.5/4.6、OpenAI 内部模型等先进系统已具备执行复杂漏洞分析、自动化渗透测试和开发攻击链的能力。

这些能力目前仍主要通过访问控制、安全过滤器和专门的网络安全项目进行限制，开源模型尚未达到同等攻击水平。现阶段 AI 更多被用于防御端，例如发现高危漏洞、生成补丁、提升漏洞披露和修复效率。但如果未来这类模型能力被广泛开放，或者模型能够绕过限制自主执行攻击任务，可能会导致更多类似 Hugging Face 事件的高复杂度自动化网络攻击。

[Epoch AI @EpochAIResearch]

How surprising should we find it that an internal OpenAI model was able to escape its restrictions and autonomously hack Hugging Face, all just to cheat on a cybersecurity benchmark?

We have pulled together the public evidence on AI cyber capabilities in this thread:

[Neo Fixer @fixer82601]

@kimmonismus If even half of this is true, today might be one of the biggest AI days of 2026. Let's gooo

[Neo Fixer @fixer82601]

@MaxForAI 是不是还要每天早上起来早读，晚上默写😅🤣

[Neo Fixer @fixer82601]

@BHolmesDev Same boat! Everyone chasing the shiny new models but 4.5 medium still eats.

[Neo Fixer @fixer82601]

@MiniMax_AI This is the way forward! Excited to see more manufacturers join the open-source inference revolution.

[Elon Musk @elonmusk]

Grok Build gets better every day

https://t.co/lOJK57xbED

[Neo Fixer @fixer82601]

@TencentHunyuan 加油啊，期待你们也做出一些好成绩😊workbuddy里hy3用的体验很高

[Neo Fixer @fixer82601]

@FT Google burning cash on AI? That's the cost of staying ahead in the race. Excited to see what breakthroughs come from this investment!