Just revamped my portfolio site! It outlines my last 5 years with @Masterfile_co and includes some technical project writeups. Feedback welcome!
https://t.co/9ReQELf5N5
I have the urge to vibe code a better GitHub mobile app that doesn’t suck to use
Just so I can review code more easily on the train
Anyone else down to jam on it?
@devagrawal09@polymorph3us@schickling Still a very powerful pattern and one that I like for the right type of problem. But not a silver bullet for every system. Thus why i am hesitant to rigidly attach my spec process to it
@devagrawal09@polymorph3us@schickling - harder to evolve the system over time
- read after write issues if your event store and projections separated
- a whole different set of scaling and tech trade offs if they are not separated
- have to be very careful about how to structure aggregates based on anticipated cmds
@samgoodwin89@dillon_mulroy imo easier to reason about and not having to stack compensations over multiple steps manually within the workflow is very nice
@devagrawal09@polymorph3us@schickling Mostly meaning this forces you into event sourcing / cqrs. But thats not always the right choice for applications. I have certainly had the additional complexity from this bite me in the ass before.
@gakonst Github is working on it. I think their approach is pretty innovative and different that what everyone else is building
https://t.co/aYdK6ODkd0
@polymorph3us@schickling Yeah I think you are describing something almost like this. I think this approach prescribes the implementation too much in the spec but directionally interesting
https://t.co/meGnsZuNjq
Introducing: Specter
Supercharge Agentic Engineering with
- Typescript Framework for Specs that compile, execute, and scaffold your app
- Vertical Slices that can be built, tested, and operated independently
Agents can finally work on large complex codebases reliably
In the last week I:
- forgot `req.json` is how you get the body of a request
- rebuilt a pre-AI project from scratch that took ~1 month of dev time from scratch in an afternoon. The new version is faster, cleaner, and more extensible
I will take this tradeoff
I understand the concern of skills atrophying when using agents. But so far I am not seeing it. Instead I have learned all sorts of dark secrets of linux networking I somehow didn’t learn before agents building a networking product.
@aboodman I still haven't seen anyone actually drive the thing. They just sit around to ooh and ahh at it
How does this thing corner?? What does it FEEL like??
@vimtor My uneducated guess is that its just not in the RL pipelines to optimize for. Its also kind of a hard thing to build the proper reward function for
@ShopifyDevs I love this feature, but its very obvious no one actually went through the onboarding process. Incredibly confusing, frustrating, and inefficient.
Happy to give some fresh feedback having just tried to go through it