22. Building https://xitter.cf/t.co/b2wSZMFfMw
Berlin, Germany
Joined June 2014
- Tweets101
- Following222
- Followers43
- Likes6.3K
New @t3dotcodes orchestrator finally made it to stable and damn it’s impressive
Agents now can:
- kick off threads
- read other threads contents
- setup scheduled tasks
- thread can switch model providers
- continue after limit reset
and so much more
Kudos @theo
When you run an AI code review, do you usually give it the initial plan / PRD?
I can see how it would be useful for an agent to see the whole picture, but I also don't want to pollute the context with stale code-in-markdown that would prevent AI to think broadly.
A very clever and sneaky move by @firecrawl - adding instructions to the skill file for agents to rate the search results automatically.
Noticed it randomly while looking at the tool calls today.
Finally migrated from Bun runtime to Node.js. Last drop was a random ESM bug when upgrading from 1.3.13 to 1.3.14
Bun offers exceptional fast tooling, but after using the runtime for almost 3 years I yearn for stability of Node again
Tried this skill, burned through my 20$ sub limit in 5 minutes with Fable🙃
You have Claude Fable for only a few days. Here's how to make the most of it.
Introducing /improve: use your most capable model to audit your codebase and write plans for cheaper models to execute later.
Studies your code, figures out bugs, perf, tech debt, missing tests, what to build and writes plans any agent can run.
Trying out @t3dotcodes and one thing I’m missing from Conductor is parallel agents per worktree
In T3 each thread (session) is linked to a specific worktree, and I couldn’t find an option to create a new thread in an existing worktree. But even then I’d feel a lack of grouping
What I like is to have one worktree per feature, and inside have multiple agents / conversations: planning, researching, etc. each linked to a feature the worktree is created for
Not possible in T3, or am I missing something? @jullerino
I’m experimenting with a remote dev box on my Mac Mini via TailScale, yet I just can’t give up the snappy feeling of local edits on Zed. Also fast internet connection is not always available.
What’re your thoughts on having a 24/7 remote dev box vs using Claude Code Web?
It would be a banger when @conductor_build supports connecting to another instance (i.e. server / Mac Mini) + mobile app.
One workstation, accessible everywhere
I love how easy it is to control outbound traffic on @CloudflareDev Dynamic Workers. With just a few lines of code you can do anything with all outgoing requests. For free.
After trying to implement proxying with other sandbox providers, this approach feels so good.
Articles on testing by @kettanaito have quickly became my go-to source for solid testing knowledge.
I've packaged them into a skill, so coding agents have actionable insights at hand when writing and fixing tests.
skills.sh/monsterdeveloper/t…
It's not perfect, but as I use it more and more the skill will become polished. Contributions are welcome, too!
Artem told me that official skills might be coming soon, so might not need this anymore. But until then!
Thank you, Artem and @EpicWeb_dev for making web a better place. Your lessons are priceless.