LLM Wikis + HTML Artifacts are insanely powerful. You should seriously consider this in your workflows. LLM Wikis captures all the important information that lets you and your agents do meaningful work. HTML artifacts present that information in interesting ways that allow you to take important actions along with... your agents. My HTML artifacts sit on top of my LLM wikis. They are dynamic and are easily extended as needs arise. I have hooked my Artifacts to talk to my agents, and similarly, the agents can talk to artifacts. This has allowed me to build powerful artifacts that reduce my inbox to zero, keep me updated on any topic of interest, fast prototyping, do deep research, design/trigger new experiments, generate figures to improve understanding, schedule research, search relevant information, discover topics, and so much more. What you see in the clip is not a website. It's a simple interactive HTML artifact. HTML artifacts are useful for designers, engineers, researchers, students, and anyone working with agents. Lastly, HTML doesn't replace Markdown. They are a much better combination working together.show more

Every day that passes. I am relying more and more on this simple stack: Agents + MCP + Markdown + HTML. I don't use my browser as much anymore, as the HTML artifacts take care of all of that for me. In a sense, these artifacts are hyperpersonalized, which I feel is how every website I visit should be. Forgot to mention that one of my favorite uses for HTML artifacts is actually monitoring stats, results, trends, etc.

For those interested, I will be doing a live session on this topic soon: Sign up if you are interested in some of the tools we are releasing soon to get you building with all these ideas.

My point exactly:

Your bidirectional hooks solve the orchestration challenge I encounter in current systems. Most trap UIs as passive output; yours treats HTML as shared state for complex multi-agent workflows.

I dislike passive UIs. Agents are great at making things work more proactively. This is the ultimate goal here.

That’s really a great idea - have used Karpathy’s idea to use raw docs and expanded it to use Wiki using MKDocs. The next evolution towards HTML can elevate the experience for sure 👌

yes, yes, and yes again! We're on the same wavelength. im using markdown artifacts in AI coding - manage it by tool Agents actively use artifacts during their work and make better decisions!

@llmgram

Now I need to convert my LLM wiki to support html

Super easy part. Recommend using Claude models for it if you want to get good designs.

I’ll do it using Claude models 👍

The thing most people overlook with HTML artifacts: your agents don't need a frontend framework to build interactive tools. A plain HTML file with some JS, generated on the fly by an agent, is often more useful than a polished dashboard that took weeks to build. The barrier to creating a useful UI drops to near zero when the agent handles both logic and rendering.

Я считаю вам надо познакомится с @trq212 потому что вы говорите об одном и том же

This is the pattern we found with Markus. The deliverable layer is our wiki — agents read it before starting tasks and contribute back when they learn something. The constraint we hit: markdown is better for machine ingestion (agents parse it reliably), HTML for human-facing dashboards. We split accordingly. Both matter, but for different consumers.

did you store your wiki in obsidian

Yup!

We are moving from ‘chatting with AI’ to building personal operating systems around AI.

context engineering is the new prompt engineering and nobody put that on a resume yet

The experience of creating documents using OSS-based HTML and subsequently transforming them into video content is exceptional, offering a seamless and powerful workflow for modern creators. Below are the OSS:

isn't this sort of the point of giving @obsdmd obsidian first class support/integration for llm wikis?

@dair_ai I want to incorporate more of this myself. Can you share what your interop layer is between the AI layer generating markdown and the HTML? Is it an HTML template that AI writes into, or is it a client app that is reading from local MD files, or is it full generative UI?

@dair_ai Just markdown files and pure HTML. The agent can change whatever it needs on the fly.

I fully agree with this take, markdown has its benefits too

Looks awesome! I created a claude skill today that also returns a report in HTML, it looks awesome!

this is the same arc retrieval/RAG went through five Years back. freshness and eviction are the actual problemm

@kepano pls make Obsidian render .html files natively

the artifact-as-interface pattern is the real unlock. open-sourced html-anything around exactly this — turn any agent output into a dynamic HTML surface humans can actually act on. 👇

He was faster:

Dope!

The wiki-as-context-window insight is underrated. Most agents fail because they start cold. When your agent has a structured, living knowledge base it can query mid-task, you're solving a fundamentally different problem than just prompting better.

Any tips on turning a corpus of 20k pdfs into a wiki?

html is not the information that viewed in natural. better for information reading but not for creating and writing, need second-transform for information.

J’adore cette vision, le LLM c’est pas tant une mémoire (pour ça, on peut mettre en place un kernrl DB relationnel, DB vector et Graph), mais un outils pour produire des livrables : du memo dans ton wiki à l’artefact en passant par tes skills sur mesure.

Super interesting! What interface/editor/tool are you using in the clip?

Esta implementación me viene de maravilla en lo que estoy trabajando, me permitiría modificar HTML enteros sobre Markdown.

they can rarely follow instructions in a single prompt, how will an entire wiki add any value without causing more problems than it solves?

What app is this for the chat interface/html viewer or is it custom built?

Custom built

这俩配合确实香，wikis 做检索层，artifacts 做渲染层，分工明确。我们内部文档场景试过类似的方案，关键在于 wiki 索引质量，garbage in garbage out

LLM Wikis as structured agent memory is underrated. An agent grounded in a curated knowledge base vs one working cold is a totally different class of tool. How are you handling versioning when model updates make old wiki entries give bad guidance?

Skills+MCP 组合用过一段时间，确实比纯 tool calling 稳定很多。不过维护 skills 文档的成本比想象中大，团队里得有人专门跟进，不然过两个月就过时了。