Make the web the best platform it can be! Currently searching for the best way to deploy my stuff. Maintainer @openclaw
Leipzig
Joined October 2021
- Tweets2.3K
- Following202
- Followers267
- Likes9.7K
CodeMonument retweeted
OpenClaw deleted around 400k LOC of its own tests without much change in code coverage. Modern models just love writing tests for every tiny change, even if they aren't useful. This skill helped. github.com/openclaw/openclaw…
Me too! Zed is not the full answer, it is fast, but sometimes the ux is confusing. Also: I want the ability to add commands to the command palette via plugins!
VSCode is slower, but has lots more customizability...
maybe I should make a minimal vscode config for my needs ...
Wow, these words from Ryan are kinda special to me. I trust Ryan's judgement a long shot and hearing this increases my opinion of fable drastically! The only thing I'd like to know would be how this experience compares to 5.6 sol ultra for you! @RyanCarniato
Fable has made me distrust every other model I've used before. It isn't perfect, but when I have a nagging suspicion something won't work completely Fable identifies it without prompting. Every other model while thorough spews complete confidence even when pushed.
A while back @theo asked for problems that certain models couldn't solve. I was working on some hard problems but nothing I threw at them they couldn't solve if spec'd well enough(ie tests). In some cases it wouldn't be as good as I could do manually but it was respectable.
But when creating something completely new you don't know what you don't know. You work with the agent to spec it. You test hypotheses. You benchmark structures. You look for gaps. I let frontier models talk me into a place of comfort, even though I had concerns.
In the past week Fable confirmed my instincts and drastically tightened up the work I had been doing. Poking huge holes in architectures that I'd been repeatedly told were streamlined, stable, hardened even why I posed the same questions.
Having the back and forth swapping between it and GPT 5.5/Opus 4.8 to save credits for different tasks has made the difference even more evident. Even on simpler tasks like PR merges other models leave so much on the table.
It reminds me of when I first started.. going from GPT 5.0 to Opus 4.5. Maybe even more so. The difference is so profound it makes it really difficult to use anything else at the moment.
If I've learned anything in the last while I haven't been ambitious enough. The problems that I know have solutions but have been too hard to do are on the table. I fear financial pressures are going to eventually be the barrier here. It's too easy to make a tool like this a crutch when it is the only seemingly feasible way to close the gap to what has been thought previously not possible.
Yeahhh thanks for merging the glimpse stuff, Makes me really happy! 🙌🥰
Will I have to install glimpse globally on my system for this to work or do you deploy it alongside when installing plannotator? Second option was my intended way of doing things, but we could disagree here
Because WLS simply makes headaches everywhere it pops up. It makes the straight forward Linux way of things wrapped in a windows adapter and therefore introduces complexity
Sucks for people not using Claude. Might work for internal company sharing of skills, when company arranges on a set of supported harnesses, but is not flexible enough for distributing skills generally. Also, postinstall urkes me, as it is the source of so many non exploits.
I actually like it 😂 but my eye constantly sees it from the side and is like: wtf is glowing there on my Homescreen 🤣
I can tell you how I manage my skills:
I made me a GitHub gist that contains a "registry" of my preferred skills. They are split into different groups, like global, and per project skills.
I then tell pi my url and say: install my global core skills.
After that, I have my
This took way to fkin long - released today for iOS
ALT GitHub app changelog on iOS, stating that creating new repos from the app is now possible.
Because flutter has some serious flaws. I looked into it for a long time, but:
- on mobile iOS it's too buggy and has slight to severe differences to native behavior
- on web it's too heavy due to webassembly, coding tooling is awkward
- you have to use dart, which simply lost to typescript
Either
build full web and adapt it to mobile and desktop with electron or ionic/capacitor
Or
Build react native/ expo on iOS and android, web for the rest of
Or
Build everything native
I will copy this idea for my dotfiles management! I started moving my dotfiles repo to an agents based approach anyway, instead of fiddling with brittle symlinking tools and changing configurations. This would take it one step further: point your clanker to the git repo and let it execute the extension which teaches it anything it needs to know to setup your machine :)
Yes, definitely! Openclaw has some x cli for that, but since x has these bit protection measures the code for accessing x might break often, which is annoying
@PaulSolt
I wrote a blogpost anyway :D
Have to redo my blog at some point, but its good enough for a quick read!
blog.codemonument.com/posts/…
Haha, saw this question late, answered already on your new post! 😄 basically: offloading distractions into the agent.
We humans need to make a ton of decisions each day. Each decision chips away more or less energy. Example: if I have a todo, I have to open my todo app, find the right category, maybe add tags and reminders and priorities or whatnot and then comes the hard part: Not getting distracted by other Todos along the way! 😳 esp. hard for an ADHD brain.
The solution: teaching an openclaw agent the structure of my task tracking / how I think about tasks. This allows me to simply give it the task at hand and it places it into the correct structure automatically.
For example:
I have around 1 big focus-topics for the next 7 days, and around 1 to 3 big focused topics for 1 to 3 months.
If I teach an agent this prioritization it starts being able to
- warn me if I overbook myself (countering adhd time/workload blindness)
- help me with consistency because it keeps my North Star in mind
- helps me to juggle general life tasks alongside my projects to not miss stuff.
At least, this is the vision. It's not there yet, you have to teach the agent a lot of workflows and attach tools like cron, memory logic etc. so that it definitely finds what you need.
For example I made it note things in nb-cli, a git-backed cli for keeping notes.
But they won't be read automatically (understandable, how should claw know why and when to read them?)
So I told him in Agents.md: if you need info about my current projects, use the nb cli and query the notebook named "c"
Next issue: keeping the internal structure of each notebook I use in agents.md clutters the main context heavily. So I added an Agents.md into each nb notebook and told the agent in my main agents.md: if you open a notebook, read the included Agents.md file once per session to understand the notes structure.
As you see, there is still a long way to go to reach the universally usable agent companion without config 😄 but the little wins along the way are extremely exciting, so much so to motivate me to keep going!
This seems like an easy question but it gets deep quickly 😄
First, I simply toyed with openclaw and wanted some fixes in there for the gateway setup and cli.
I made PRs and then was asked by Pete if I want to become a maintainer. Sich an honor! But also a lot of pressure and self doubt. There are SO many AMAZING people maintaining this that I really feel not capable enough sometimes. But all are so friendly and supportive, it's really nice! Also, the vibe is extra fine, chatting with Molty (steipetes bot) is hilarious and everyone is soooo excited to build this!
Now, let's talk about coding in this environment. As I started, I only knew the classic GitHub OSS contribution flow: make an issue, get approval / thought from maintainers, work on it, make a pr, get comments, fix stuff, merge.
But when the founder of this project works with 10 or more clankers at the same time, the first question is: how do I even start to work at a bug without stepping on his toes or working on something image finishes while I'm working?
You have to learn agentic engineering. You do anything with your clanker: learning the codebase, analyzing PRs and issues, making plans and architecture decisions. Even shipping/merging mostly avoids the GitHub Ui and automated everything via skills and gh cli.
And to wrap this story up, I'm really thankful for that. Pete showed me the way to agenting engineering and it was a crashcourse at the start of the year, but I'm feeling great in the game right now and able to incrementally improve my workflows and processes.
His blogpost "Shipping at inference speed" basically changed may life this year, literally. I feel like I'm on the way to pivot my whole career, calling me "Agentic engineer" now.
Right now I'm moving in with my girlfriend and finishing driving school, therefore I'm not able to contribute much right now as maintainer, which I find very sad, since I'm very interested in this onboarding Flow stuff and sanding off the rough edges you're experiencing right now.
I can't wait to come back into it after the stress is over ! 🙌
Btw: thanks for following! I feel honored, since I normally read your posts, find useful insights for iOS and macOS development, not the other way around! 😄✨ hope you find gems for you in the things I post/repost!