bersabel

As models and agents get more capable, it seems like the workflows around them keep getting more complex. That feels backwards. Workflows should get simpler as technology gets better. Moss was built on that tenet: mossnotes.app

@insights_infra @sawyerhood i’ll dm you!

@brsabel @sawyerhood Which wework is this?

“trust me bro, we all need to get one”

@sawyerhood i did ask you if i could post this, but maybe i should've also shown it to you

@brsabel why are you framemogging me like this

i hear faint ASMR whispering then turn around to this

new release with these changes (and a ton more) just went out! mossnotes.app/release-notes

Been iterating a lot on Moss' design recently... • Docked the header & revamped its layout • Moved note title to canvas • Replaced left rail with a footer • Added themes for code blocks + overhauled the chart palette. Planning on making these customizable

@umang this is so immensely helpful thank you!! will work through all of this today/tomorrow. also about to get out a release where i refactored the comment system to fix some of the issues you mention, so hoping that will help soon!

some early feedback: 1. hard to understate how excited i am about this :) 2. i got very nervous when i realized it's an agent running on my computer - some q&a via cmd-k helped, but would love upfront content in Getting Started about this + security model 3. commenting feels like the most powerful add-on vs other md editors, but ran into bugs (text not following cursor, cmd-enter not saving comment) 4. would love to use my own agent with it 5. sidebar chat defaults to editing the doc, so it's often responding by editing the doc rather than in the chat, which is ... unnatural. sometimes i want doc edits, sometimes i just want to discuss. 6. cannot continue a sidebar chat currently, which feels like it would be useful to do 7. would be nice to be able to see recent changes made by the agent (file was in a git repo, so that might be easy to pull up) 8. i added a whole git repo folder to the app, and it froze for ~minutes, then added every .md file in node_modules to the notes list (readme, changelog, etc for every node package!) 9. clicking on cmd-f results take me to the note, not the actual result inside the note sorry if this comes across as overly critical - i really really like everything about it - the premise, the UI, everything else. this might replace obsidian for me.

As models and agents get more capable, it seems like the workflows around them keep getting more complex. That feels backwards. Workflows should get simpler as technology gets better. Moss was built on that tenet: mossnotes.app

@iamcoreyg getting this out in a release this evening, along with a ton of other stuff!

@brsabel could it do this step automatically? (Ask the user to make it the default .md app)

It’s a beautiful day, so naturally I kicked off work with Claude and Codex and went for a run. Had them use Moss as a workspace to document and present their work. Came back to: • interaction design research • QA notes • PostHog error investigation

@iamcoreyg let me look into this!!

Out now for anyone to try: mossnotes.app

@iamcoreyg thank you!! feel free to share any/all feedback

@brsabel I was about to say I'm excited for you to release this — then checked the site and saw its out, congrats!

working on something fun

@trq212 i went from always being surprised at how quickly i’d get to 60% to being shocked at how long i stay at 95%

i think we might have undersold 1M context tbh, the performance is so so good, I really just don't clear the context window much these days

fucked around and kinda made an art tool

@wzono1 try the latest release! mossnotes.app/download existing apps will also auto-update

@brsabel cmd + k does not support Claude CLI API mode.

Seems like Anthropic’s compute hasn’t caught up with recent demand surge. Classic utility pricing move: incentivize off-peak usage. Does this become the new normal? Peak hours also line up with East Coast working hours. Finance, legal, and consulting prob dominate use cases.

I always have a subagent review a plan before I implement it, and this is also recommended by the Claude Code team. Why isn't this an option?

@sawyerhood i’m afraid you’re lost in the sauce

no one told me this is what coding would become

@trq212 @tylerangert just hit this a few times. seems like the agent is unable to use the ExitPlanMode tool when it happens. manually tab'ing out of plan mode fixes this

@tylerangert like this happens every time for you?

hey @trq212 there's a bug in CC where after approving a plan claude will keep asking you to reapprove the plan or it mistakenly thinks you want to make edits to the plan