What frustrates me about many projects is it's never been easier to just make a quick video or image or gif and put it in the README to show what the value is. This was at least somewhat complex before AI but now you can just paste an image and have the AI put it in your README.
I guess the frustrating bit is that the readme is 100s of lines long. AI is already saving you so much time on a project like this but you still make me clone and spin it up if i want to see it for real.
I get that open source is a gift and you’re not obliged to do anything I say but please consider taking the extra few minutes on your next project.
I've had a fairly thorough exploration of how I can give my agent a whiteboard so we can work on architectures together. To my surprise, current solutions (including Excalidraw) were not good enough and didn't deliver what I wanted. I ended up finding Mermaid to be the most agent-friendly medium and coded an Obsidian plugin for it. It works okay; we can work on the same doc while I bring my own AI agent, and we can brainstorm together.
https://community.obsidian.md/plugins/mermaid-relay
YMMV, but the value I get from producing a diagram is derived from the thinking. Thinking about what I'm trying to draw leads to understanding about where my/my team's knowledge is poorer, what assumptions we're making, etc.
I don't remember who said it, but someone mentioned that mathematics papers don't read well because the models don't know which parts to emphasize and which parts to gloss over. I would imagine similar tendencies here.
Yeah I tend to make my diagrams either as I go, as exploration, or after I'm done figuring things out as a summarizing artifact (and agents are great at using Mermaid to do that second one).
I’ve been trying to find the best practice to connect a person’s existing ChatGPT/Codex / Claude accounts so they can be used in an app for the agent in the app.
Any advice? Everything seems clunky, even the best MCP efforts
Posted this somewhere else aswell but I've had good success with https://whiteboard-mcp.com - it seems to work better than other free alternatives for things like architecture diagram creation
tldraw has a lot of programmatic access I have not compared it to excalidraw but there is still no equivalence between the model capabilities to build amazingly detailed infographics and diagraming tool use which is a gap I hope will narrow
What frustrates me about many projects is it's never been easier to just make a quick video or image or gif and put it in the README to show what the value is. This was at least somewhat complex before AI but now you can just paste an image and have the AI put it in your README.
I guess the frustrating bit is that the readme is 100s of lines long. AI is already saving you so much time on a project like this but you still make me clone and spin it up if i want to see it for real.
I get that open source is a gift and you’re not obliged to do anything I say but please consider taking the extra few minutes on your next project.
I've had a fairly thorough exploration of how I can give my agent a whiteboard so we can work on architectures together. To my surprise, current solutions (including Excalidraw) were not good enough and didn't deliver what I wanted. I ended up finding Mermaid to be the most agent-friendly medium and coded an Obsidian plugin for it. It works okay; we can work on the same doc while I bring my own AI agent, and we can brainstorm together. https://community.obsidian.md/plugins/mermaid-relay
YMMV, but the value I get from producing a diagram is derived from the thinking. Thinking about what I'm trying to draw leads to understanding about where my/my team's knowledge is poorer, what assumptions we're making, etc.
I don't remember who said it, but someone mentioned that mathematics papers don't read well because the models don't know which parts to emphasize and which parts to gloss over. I would imagine similar tendencies here.
Yeah I tend to make my diagrams either as I go, as exploration, or after I'm done figuring things out as a summarizing artifact (and agents are great at using Mermaid to do that second one).
Excalidraw offers their own open source first party MCP endpoint[0] and server[1]:
[0]https://mcp.excalidraw.com
[1]https://github.com/excalidraw/excalidraw-mcp
I’ve been trying to find the best practice to connect a person’s existing ChatGPT/Codex / Claude accounts so they can be used in an app for the agent in the app.
Any advice? Everything seems clunky, even the best MCP efforts
Posted this somewhere else aswell but I've had good success with https://whiteboard-mcp.com - it seems to work better than other free alternatives for things like architecture diagram creation
video https://lnkd.in/p/ez4EZede
tldraw has a lot of programmatic access I have not compared it to excalidraw but there is still no equivalence between the model capabilities to build amazingly detailed infographics and diagraming tool use which is a gap I hope will narrow