This blog is also a blyg now
I stumbled upon Venkatesh Rao’s Blygger protocol hackathon. I was curious because, as I recall, the announcement alluded to how he wrote Cognitive Mercantilism, which I found surprisingly readable for a heavily AI-assisted piece. That essay was the real reason I signed up, and I learned a lot more about what a blyg is in the process.
This post is my first since setting up my blyg.
Venkatesh Rao · Starter checklistWrite a hello world post
So what is a blyg? For newcomers: a website becomes a blyg once it meets at least Level 1 of the Blygger spec, where every post keeps its versions and can be pinned or withdrawn. A plain RSS feed counts only as Level 0, which is how blygs can still follow ordinary blogs. Most of the blygs listed on blygger.com so far are personal ones.
An earlier version of this post said any site with an RSS feed could become a blyg. Venkatesh replied:
Venkatesh RaoI defined RSS as Level 0 to grandfather the blogosphere in :) But it doesn’t become a blyg until at least Level 1 spec compliance.
A blyg starts as a site with a feed, but each post also publishes a small data file beside it. That file records the post’s permanent identity, every version it has gone through and a short note on why each one changed. Because of that record, another writer can quote a passage tied to the exact version they read, or declare their post a reply to yours, and software on both sides can check that the link is real. Versions you choose to pin are kept online for good. Blygger Studio is the reference app built on these rules: an editor, a reader for following other blygs, and a compose screen for quoting and replying, in one place.This passage was written by Claude Fable 5.2.
What interested me most: the Blygger Studio editor lets agents fill in the blanks. You put an instruction between TK marks wherever you need text, and the agent writes that passage. I tried something similar for the description of Studio above, and I’d already been filling gaps this way in a new blog post.
Venkatesh Rao · Starter checklistAdd an AI for [TK] prompt [/TK] generation and Changelog generation (recommended: Sonnet and Haiku or equivalent)
Given how good the Cognitive Mercantilism post was, I’d assumed Blygger used a fine-tuned model. Nothing that fancy was going on, and I was glad to be corrected: his approach turned out to resemble what I’ve been trying recently.
As I understood it from the call, Venkatesh’s special command inside TK changes how imported AI text appears on his blyg, making its provenance clearer for readers and authors who want that. I could be wrong; this way of writing is still maturing.
Venkatesh Rao · Starter checklistDo a [TK] impyrt = text [/TK] cut and paste AI section from an external LLM
I’m not using Studio, so I don’t expect to mark text as AI-written as formally as Studio does. Where it matters I’ll say so in a side note or a footnote, as I have in this post, and anyone who is curious can run a page through Pangram. I don’t see the lack of formal marking as a deal-breaker. Most of what the agent writes is there for me to understand when I review, or to clarify something I knew was unclear; I then discuss that text with the agent again and polish it. My old post, Claude finds Contradictions in My Thinking, was the first time I let Claude write in my Obsidian vault. There I kept my own writing in the margin notes, but no convention seems clearly better to me; I keep the friction in my final edit, so that a post says what I have learned even where agents drafted or researched parts of it.
Summarising another post from inside TK seems particularly cool: I could link to an old post, mine or someone else’s, and get a summary of about 100 words.
Venkatesh Rao · Starter checklistDo a [TK] prompt ![[id]] [/TK] TK-transclude to summarize another item (your own or somebody else's) with AI
But I could also do the same in my own setup, right? I do most of my writing in Obsidian and build the site with Jekyll. Much of my writing now involves Botference, a plugin I’ll submit to this hackathon. Botference lets me highlight passages as I read, take notes and discuss ideas with an agent when needed—like PDF comments, but for websites. I also built an export to Obsidian, in a set format, for when I finish taking notes on a site or blyg. The note lands in the same vault as my blog, so all the note needs is some final polishing and formatting before it can be published.
I’m mainly trying to solve an interface problem. I already have a quick path from writing to publication. What I want is for the whole path, from reading to writing to publication, to be just as quick: jot things down as I read, have them land in Obsidian with the quotes exact and the details in place, and publish soon after. That speed is what the AI-assisted plugin is for, and why I won’t be using Studio. Others are trying to close the same gap. I wasn’t a fan of the editor shown, though I’ve never been entirely happy with any editor: Obsidian works well for some things and frustrates me with others. Ultimately, I want to learn without friction, and agents help with that.
I also do things on my site with agent-made images and other content that a blyg may need to accommodate. The interactive lineage figure in A Vacuum of the Imagination is one: an agent made it, and in a feed it can only be a link back to my site. If Blygger is about agents, part of the challenge is making all of that content available to other people, their agents and mine.
I was glad to see a decentralized protocol that lets me keep my blog’s appearance while feeding its content elsewhere. I’d love my blog to flow into Substack automatically, but it doesn’t. Here, I can write on my site and have posts enter a feed where others read them in whatever format they prefer, while my site keeps its quirks, which suit some readers more than others.
I want to read more of other people’s work. It would be cool if everyone moved to blygs, though that seems unlikely. Blygger could offer a way out of the metrics-driven feeds of Substack and Twitter. Many people are thinking about how to write better with AI, and I’m glad I joined the call, even if what I’m building may only be useful to me.
Blygger · v4 · Changelog
Created on Blygger:
Most recent: , v4
-
v4 ·
Replaced the footnote about blyg eligibility with an embedded excerpt from Venkatesh Rao’s reply, immediately after the explanation of RSS and the Blygger spec. (AI-generated)
-
v3 ·
Corrected the blyg definition to require Level 1, clarified that RSS alone is Level 0, and added a footnote crediting Venkatesh Rao. (AI-generated)
-
v2 ·
Links to my own posts now point to their Blygger versions.
-
v1 ·
First published.