im in a weird state, where I dont even like writing code by hand anymore. its just too addicting to ship something working quick, the time I use to type out types or constructors feels wasted.
weird times
yo @linear why did you kill the "full screen" review mode
was really helpful on smaller-ish screens :)
I feel a bit dumb
I finally started using herdr, its worktrees and whatever properly and holy shit, I feel so much more productive than switching zed windows
whats the issue with opus doing everything it can to avoid @pidotdev edit tools? It will literally mount volumes to pipe python commands through it instead of calling edit()
what the hell
turns out you can just vibe things. wanted my agents to be able to verify their UI work better and make better PRs, here we are.
github.com/timzolleis/shot
please tell me if this is shit, needs improvements or can be deleted alltogether because this already exists.
Since I got some feedback that the skill is a bit unreadable for humans without context, I updated the skill and also used it to build this example implementation:
github.com/timzolleis/effect…
It also includes SPEC.md, which was the original draft I iterated on to build the skill.
okay, so I've written a few PRs with this and it worked out really well, both reviewing the plan and the final code (and agent verification), so im just gonna post my DDD findings here
github.com/timzolleis/.dotfi…
I think strong default patterns help agents write really nice code
okay, so I've written a few PRs with this and it worked out really well, both reviewing the plan and the final code (and agent verification), so im just gonna post my DDD findings here
github.com/timzolleis/.dotfi…
I think strong default patterns help agents write really nice code
read a lot about DDD today (code examples, concepts). I built some principles out of this with Opus and @EffectTS_ that should make it really hard for agents to slop, and really easy to verify correctness.
Still got to try it out a bit more, but really excited!
last repo. what a ride! actually okay-ish diff for the size of the repo
shoutout to @EffectTS_ for the migration path, lots of "breaking" apis but seamless agent documentation
im _really_ torn on agentic coding today. Are we accepting that „good enough“ is now fine (in terms of abstractions, coding style)?
I cant really justify spending hours cleaning up or refactoring ai code to myself. Idk man, this feels weird
if you used a library that interacts with users server side logic (like a sync engine that needs to publish events)
what would be your favorite way to ensure consistency (transactions)? implement an effect service? fire and forget?
kinda fiddling with the details here
the renames of Schema.TaggedError have come to an end
seriously, congrats. couldn't imagine building typescript software without it, effect has become such an amazing piece of software to work with
where can I find beautiful and somewhat clean effect code?
need to de-slop my brain a bit from viewing all the agent code the last few months








