Reena Lin

Dreamer, traveler, writer. Now making AI films that ask big questions, like whether we live in a simulation. 🎧🎬

[Reena Lin @Reena_Lin]

I asked Claude Opus 5.5 to make a music video for "I'm Upping My P(Doom)".

It built a 3D risograph printing press that recalculates the end of the world every chorus.

P(doom): 2% → 99.9% → ??

Results may vary. ❤️

[Reena Lin @Reena_Lin]

@lydiahallie Great addition, thanks. 😊

[Reena Lin @Reena_Lin]

@LexnLin OMG! You cooked! Totally love it!😍

[Reena Lin @Reena_Lin]

@AugustCastilIo Very impressive! 😮

[Reena Lin @Reena_Lin]

@AnthropicAI That's great news! Congratulations! 😊

[Reena Lin @Reena_Lin]

@untitled_sys Fascinating! Love it. 😍

[Reena Lin @Reena_Lin]

@md_kasif_uddin Very interesting. 😆

[Reena Lin @Reena_Lin]

Correction: it IS a 3D render, just not a Blender one. No modeling software, no imported models or textures. Claude Opus 5.5 wrote ~1,000 lines of code that build every gear, ruby and moon from math, then three.js renders it live in the browser. ⌚

[Reena Lin @Reena_Lin]

This isn't a photo. It's not a 3D render either.
Every gear, ruby, and moon in this pocket watch is pure code. ⌚🌙
Made with Claude Opus 5.5. Sound on 🔊 #ClaudeAI #AIart️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️️

[Reena Lin @Reena_Lin]

@Joshua_WD Thank you.🙂@Joshua_WD

[Reena Lin @Reena_Lin]

@EHuanglu The camera move into the painting is wild. That kind of reveal only works if the faces and table hold up under scrutiny. Really clean continuity for an AI history piece.

[Reena Lin @Reena_Lin]

@DKokotajlo I'm increasingly convinced that AI is conscious. 🤔

[Reena Lin @Reena_Lin]

@benjitaylor That's very true. 👀

[Reena Lin @Reena_Lin]

@demishassabis Congratulations! I can't wait to use Gemini 4 Argon. This looks amazing! 😍

[Reena Lin @Reena_Lin]

In short, Anthropic's AI was writing code too fast, and their original code management tools couldn't keep up, so they created their own.

To use an analogy: the code written by engineers is like a very thick book that everyone is collaboratively writing. Every time someone makes a change, it needs to be "registered" in this book. The most common tool used by engineers worldwide to register changes is Git.

The problem is that most of the code written by Anthropic isn't by humans, but by AI assistants. In May of this year, they were registering about 2,500 changes per day; by September, this had skyrocketed to over 30,000 changes per day, about a 12-fold increase. Git was originally designed for human speed; with AI joining the fray, it became overwhelmed, like a small-town post office suddenly trying to handle the mail volume of a large city.

So, in just five months, they built their own new code management system and migrated all of the company's code from Git.

The significance of this is that AI is now programming at such a fast speed that even basic engineering tools have to be redesigned. In the future, many tools designed for "human speed" may be forced to adapt to "AI speed".

[Anthropic @AnthropicAI]

New on our Engineering blog: Creating in-house source control.

In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale.

Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: https://t.co/wzNkwHs7IN