herrkaefer

Introducing: Video Use. Edit videos with Claude Code. 🫡 I got tired of paying for video editors, so I made a Claude Code skill that does it for me. > Talk to camera, get final.mp4 > Auto cuts fillers, color grades, adds subtitles > Adds Manim and Remotion animations > Self evals the render before you see it 100% open source, 100% free.

Shocking result on my pelican benchmark this morning, I got a better pelican from a 21GB local Qwen3.6-35B-A3B running on my laptop than I did from the new Opus 4.7! Qwen on the left, Opus on the right

Codex for (almost) everything. It can now use apps on your Mac, connect to more of your tools, create images, learn from previous actions, remember how you like to work, and take on ongoing and repeatable tasks.

Vibe planning / architecture design is one of the few joys left for programmers today, because it still depends on your domain knowledge and experience. Cherish it. User your brain to hold onto that thread. Don’t let AI run wild on its own. Output comes second. Waiting & coding: a minus. That’s why I prefer models that don’t overthink but give fast feedback. Faster pace, even if slightly off, helps you remain actively engaged in the developing loop. That’s why codex 5.4 high feels just right.

My current take: vibe coding still has limits. You need a human in the loop to keep complexity under control and to handle the details efficiently. But vibe planning is real productivity. Every developer has limits, and can't be an expert at everything. Traditionally, we tend to frame problems and look for solutions within the space we’re already familiar with. But AI has seen far more patterns and possibilities, and can explore a much larger solution space. That’s why it often comes back with better or unexpected options. Our job is to be good at asking it for directions and suggestions — and (for now) keep the final decision in our hands. Also we have to learn quickly during the planning stage, with the help of AI of course.

Programmers worried about AI agents should take a page from OnlyFans. OnlyFans was revolutionary. It made creation trivial and removed gatekeepers from idea to income. OnlyFans has been the Claude Code for the people who vibe create their way into the adult industry. So the great flood of supply came, ie those who had never been in the industry suddenly found themselves in it. What happened to those who were already in the industry? The reality is that they remained successful. The audience only increases by # of people turn 18 each year, so all things considered attention simply spread thinner and became harder to hold. So what is the underlying dynamics? Well the oversupplied market treats average work the way the internet treats a slow page. The world moves on without ceremony on "average". Skill issue is a harsh synonym for not good enough. So who made money on onlyfans? Those who are oddly specific. Software is walking into similar equilibrium: It helps to be known for something oddly specific. And of course if you already have an audience, try to keep it close and give them a reason to stay. When creation is easy, skill stops being scarce. Being wanted and being in others desire does.

100刀pro初印象不佳 codex没有pro模型，故此我先让codex导出细节+数据为2个文本文件 然后提交到chatGPT窗口pro研究 等了半天，说finished reasoning。 但啥内容也没有，我就追问ta完成了吗？ 结果又开始跑reasoning了 感觉被奥特曼这家伙忽悠了

GPT-5.4 Thinking and GPT-5.4 Pro are rolling out now in ChatGPT. GPT-5.4 is also now available in the API and Codex. GPT-5.4 brings our advances in reasoning, coding, and agentic workflows into one frontier model.

OpenAI的100刀plan终于出来了 跟claude一样的，5x的plus用量 + 可以使用pro模型 现在促销期间到5月底，为10x用量 我订阅了