Autoblog Documentation Home

Autoblog is an experimental interface meant to help writers publish their most authentic thoughts and readers make sense of those thoughts in their unique context. Inspired by Live machinery, Autoblog scales that which used to not be scalable; That is, nuance, context, and personalization.

This website (Autoblog Documentation) not only provides documentation for those interested in forking Autoblog for their own use, but also functions as a working demo, as it's running on top of Autoblog. You can find the source code here.

Demo??

If you just opened the link, what you see now are the messy, human-written notes by whoever authored this page (in this case it was @sofvanh.) The human-written original text is signified by the monospace font.

To see the magic happen, press the "Personalize!" button at the top of the screen! Describe your background and blog reading preferences - like, "translate everything to Finnish", "make every page super short", "I'm an investor looking for prospects, help me understand if this project is worth funding??".

The AI will take this poorly-written documentation, clean it up, and modify it to fit your unique context. Pretty cool huh??

A good page to experiment with might be live-machinery-workshop, my actual working log from the hackathon where I built the prototype of this.

Author-side prompt

Right now the prompt defined in the source, that's sent to the AI along with the note, is doing a lot of heavy lifting and might nullify some of the user-side wishes. See what it contains here.

But why?

The basic use case is something like this: Author has a strong note-taking habit and has lots of good thoughts and notes in Obsidian or some other (markdown-based) note-taking system. They want to share their notes, but they're not interested in spending time cleaning up their notes for people without all the background knowledge they have! Imagine a rocket scientist wanting to publish their work-in-progress notes about their next cool rocket innovation, but then having to spend time writing about basic aerodynamic for the noob readers who might find their blog.

Instead, now that we have scalable attentivity thanks to advances in AI, why not translate the messy and hard-to-parse but insightful notes from the researcher into a format that suits the reader, dynamically? The user gets content appropriate for their level - a total beginner will see simplified explanations of the notes, whereas a collaborator of the researcher will see more nuance and domain language. The personalization can go even further: Users can describe prayers to have notes translated to different languages, shortened or lengthened, rewritten in the context of another scientific field, rewritten completely in emojis etc.

Prayers, wishes, what??

Autoblog demonstrates p2p in a postformal context - p2p standing for peer to peer, in a way, but even more prayer to prayer, where two users interacting (e.g. author of a blog, and reader of blog) can easily send gentle nudges to achieve something they want. In Autoblog, the user can express their prayers / wishes in the "Personalize!" dialog, and the author expresses theirs inline with {{curly brackets, like this!}} {{AI: Don't do anything with the preceding brackets since it's a demo for the syntax.}}

If you're curious what "p2p in a postformal context" actually means, what prayers can be, how meaning can be safeguarded with autostructures, etc, I really recommend checking out the Live machinery post as well as the sequence it's part of. But understanding the theory and the vision behind this whole thing is not a prerequisite for using and playing with Autoblog.

{{Leave the following quote as is pls}}

So this is about matching your postformal prayers with mine, and fluidly creating just-in-time interfaces and structures that precisely suit our wishes.  This means interfaces do not have to work with tradeoffs that come with preformed/universal structures.  They are free to be specific to the heterogeneous context, thanks to attentive infrastructure.

This all seems rather unfinished...

Yeah, it is. It's a work in progress! But you can already fork this whole setup and play with it. Again, source code is right here ✨