Justine Moore of a16z demonstrates an OpenAI Codex-powered utility that converts code repositories into visual diagrams

@ianneo_ai By default, it will generate in Chinese - but you can easily ask to make it in English (or another language) instead.

Link below, just give it to Codex and it will handle the rest.

H/T to @ianneo_ai for making this!

https://github.com/helloianneo/ian-xiaohei-illustrations/tree/main

this is awesome, but also i remain convinced that everyone is still sleeping on raw gpt-image-2.

it's both 1) a reasoning model that just happens to output images, and 2) will accept up to 32k character inputs.

here are two outputs i generated, using this prompt: "<a long chatgpt summary of the x recommendation algo> teach me this visually, a very simple illustration with a cute blob character"

@venturetwins @ianneo_ai https://github.com/helloianneo/ian-xiaohei-illustrations

Found the skill for the people interested:

@venturetwins So happy to see this. Thank you for sharing it！

mind blowing🤯

Thanks for sharing this @venturetwins

I dropped this repo link in my codex folder on ChatGPT iOS app. A small prompt and this explains way better than I could!

H/T @ianneo_ai for that repo and the Codex mobile app crew (cc @Dimillian et al)

Learn more —> http://click2track.app Install via free TestFlight —> http://testflight.apple.com/join/TmjuE1ja

@venturetwins @ianneo_ai nice one - added to https://skillvault.md/helloianneo/ian-xiaohei-illustrations

@ianneo_ai Thank you for making it!

@venturetwins @ianneo_ai the "dwell" signal being buried in the middle of that conveyor while "like" leads the action list is doing more honest work than most algo explainers written by humans

@venturetwins @ianneo_ai This is incredibly charming and looks fun to try and that also makes me wonder if it is the perfect vector for a prompt attack. (It seems to be clean, however.)

@venturetwins @ianneo_ai Fun use case. It also makes Codex feel a lot less like an assistant and a lot more like a canvas for communication.

@venturetwins @ianneo_ai ok this is actually brilliant. kinda need this cute little blob guy to explain how the Reddit algorithm works for our client pitches.

@venturetwins @ianneo_ai You 'inject' your follows into the hopper It adulterates with discovery There's a process The result is a "moulded" rank list

An injection moulding analogy

@venturetwins @ianneo_ai The X algo as a stress test was the right call. Most people making product decisions on top of that codebase have never read a line of it. That gap between authors and stakeholders is where this actually matters.

@venturetwins @ianneo_ai the cute output is burying the point here. a skill is a capability you teach once and reuse, not a prompt you rewrite every run, and that compounds

@venturetwins @ianneo_ai link please

@venturetwins @ianneo_ai This is the kind of AI use case that sneaks up on you. "Feed it code and get a visual story" sounds niche until you realize how many people struggle to communicate technical concepts to non-technical stakeholders.

Explainability as a skill is underrated in product building.

@venturetwins @ianneo_ai the blob looks like it just went through its own explanation and came out confused

@markerdmann @venturetwins @ianneo_ai this is so adorable, minor glitches aside, gpt-image 2 is really powerful

@venturetwins @ianneo_ai this style is weirdly charming for explaining a recommendation algo

would actually read docs if they came with these little blob characters?

@venturetwins @ianneo_ai This is the strongest use case for visual skills: turning invisible structure into something a non-owner can inspect. The X algo is a good test because everyone has opinions and almost nobody has read the system.