I'm finding that I wish conversation threads in LLM tools worked more like tabs in arc, and just auto deleted after say 3 days.

Anyone else noticing that search engine results are increasingly terrible?

Mirroring code-switching -- side-by-side reading was thinking more of each pair of book pages being connected. I was able to get into a good flow, but maybe the English should only be visible on highlight.

Interlinear code-switching -- more fully translation-focused, seeing both languages at the same time. When trying it, I find I need to focus more on not defaulting to the English, which feels like cheating.

Progressive code-switching -- translation progressed in complexity based on word frequency and book length. This still feels like the most appealing learning curve when trying it.

Strange Harvests has been a wild collection of rabbit holes. Would recommend.

石の上にも三年 (ishi no ue ni mo sannen) Stick with things even when they’re uncomfortable; steady patience turns “this kinda sucks” into real progress with habits, skills, or long-term projects.

Anyone come across research on Variation Theory for dyslexic learners?

Been thinking a lot about interfaces that promote learning in new ways what if you started reading a book in one language and finished it in another?

202603 - How to use AI Updated my workflow for feature and v0 product creation. www.gndclouds.earth/journal/how...

just noticed that Mac Pro is gone rip

Starting to play with github.com/collaborato... its one the nicer solutions I have found for tracking agents across multiple projects at once.

love how old school html pages feels almost impossiably fast, modern web stack are so heavy 😅

I have been finding the discussions around design being "cooked" with recent AI model changes to be a little exaggerated. Curiouse if this resonate with others. www.gndclouds.earth/journal/des...

friendly reminder space things are happening today 🚀🌕 www.youtube.com/watch?v=vr4t...

Looking for 30 people to test tiny.garden. If you ever used small victories back in the day this is very much similar but instead of dropbox for now its all based on are.na channels.

tiny.garden is ready for testing! If you have an arena channel you want to make into a mini website give it a try and lmk what could work better!

😱 me frantically trying to find an outlet before the agents loose power.

The electronics faire schedule (keynote, workshops, and more) is now live! Excellent time for a visit to Philly <3 sites.temple.edu/efaire/sched...

Claude helping me clean up years of inconsistent front matter properties in Obsidian!

Trying to incorporate more natural systems thinking into my work. And realize there's a lot of functionality in using a tree ring to visualize data overtime.

does anyone use or want a tea tracker? I'm thinking like post.cv but for a tea community.

Hot take: "thinking" isn't the right word for what LLMs do. Humans think. Models reason. He feels important that we keep these two things separate until machines do actually think.

Any scientists using Cursor for their work? I’d love to talk with any scientists using Cursor about an MCP service I’ve been working on. DM me please!

Sometimes I wonder how my plants feel about me drinking tea in front them.

Over-applying LLMs is like driving to the corner store: you spend longer parking than walking. They’re great for teaching and cold starts, less great as an unmonitored end-to-end pipeline. www.gndclouds.earth/journal/whe...

Would be cool if Cursor's .md preview was a TipTap-style rich text editor. I always get a little frustrated when I try to edit the preview and it just opens the raw markdown instead.

After a few weeks with Claude Cowork, I've landed on three patterns that changed how I work: solving the cold start problem, building to learn instead of reading docs, and rapid visual exploration at scale. www.gndclouds.earth/journal/tho...

It's pretty wild how much I have learned in the last year just by talking to cursor.

We have protocols for giving agents tools (MCP). We have protocols for giving agents tasks (A2A). What we don't have is a protocol for giving agents presence. www.gndclouds.earth/journal/obs...

Noticed a drone pod on the roof of my local fire department and got curious. No public flight logs from SFFD yet, but SFPD is required to log all drone flights and retain records for two years. #crafted-with-cursor

"The chatbot is in the desert and dying. Why are you crying and helping it?"

just doodling on some things

sensors.planetary.software is an early experiment making it easy to explore just how much of the physical world is already being measured, quietly and constantly, all around us. #crafted-with-cursor gndclouds.earth/project/sen...

pretty happy with how this is evolving, there are now about 22k public sensors available as an MCP service, with all the data now normalized across the external APIs. Its starting to feel like a substrate for physical world intelligence.

exploring sensor data visualizations

just more globes

The tabs state still feels a little off, but I'm finally happy with the loading state for the post title.

When you finally have docs for project and need to tell Cursor about it. 🙂 iwg-atlas-docs.vercel.app/

Testing something #crafted-with-cursor

I may regret this, but hoping to learn a lot either way. Crafting personal tools has become an obsession. With Cursor, I keep making software instead of buying it. And I want to hold myself to that for this year. www.gndclouds.earth/journal/curs...

I've been wanting to explore this idea for a while. Instead of using a dedicated chat app, what does a chat overlay feel like? One where I can quickly ask questions or call local tools without leaving what I'm working on. But with memory then say a Raycast.

progress submit button is a vibe.

For those using Cursor with Xcode, do you have any tips or integrations that make it easier to edit interfaces and work with Xcode all via Cursor?

planetary boundaries are a compelling idea to me, the data is so scattered across NOAA, EPA, USGS, university sensors, satellite feeds. no unified layer to ask "is this system losing resilience?" How do we track these when the data is so spread out. stockholmresilience.org

the globe experimenting continues

There's a gap between frontier AI and the physical world. Claude and GPT can discuss climate science abstractly but can't tell you if the creek behind your house is rising. The data exists. The infrastructure to connect it doesn't.

I don't have a clear use for this yet but it was fun to make.

what's your favorite weird nature dataset? I'm collecting earth APIs for a project and want the niche stuff. plants? fungi? ocean buoys? soil? send them my way!