Generalist

Building orchestration tools around AI

[Generalist @gendev01]

@thsottiaux I still did not get astra access being on Pro plan

[Generalist @gendev01]

@vibamohan_ https://t.co/krPlrPLILa

[Generalist @gendev01]

@kunchenguid I think its primary user is not a dev but rather a UI Designer.

[Generalist @gendev01]

@augustdeploying maybe u have too many tabs open

[Generalist @gendev01]

@dabit3 building score. deterministic harness around coding agents there is no llm orchestrating, so there's nothing that hallucinates. it hands out issues, watches git not the chat, merges only what proves CI green.

https://t.co/NdWN3vP0SP

[Generalist @gendev01]

@_lopopolo Either you have skills placed somewhere else or you just prompting and interacting with every agent that you have. A reused prompt is a skill and if you use some kind of a harness that essentially means u externalized a skill

[Generalist @gendev01]

@0xJ4yD3v The problem that I have comes from having skills loading from 3 if not 4 places 1. Root skills 2. Plugin skills 3. Symlink to .agents 4. Project agents

And plugin skills keep on updating with Claude they keep on adding new skills that I don’t need necessarily

[Generalist @gendev01]

claude code and codex been really feeling wrong, for a while there are too many skills and both ~/.claude and ~/.codex have way too many files which dont make sense

[Generalist @gendev01]

@jjacky maybe you are not supposed to use too many interfaces ? I hate to use too many as well in fact I use one and outsource significant portion to https://t.co/NdWN3vP0SP

It needs bit of a setup but yeah surely it will help u a bunch

[Generalist @gendev01]

@kunchenguid I spend a lot more time on understanding the architecture and how and where the next things would fall in to the puzzle but yes editors have kinda become obsolete

[Generalist @gendev01]

The way that I use it is I plan a epic talk with the agent and once I think epic looks good there are few scripts that should convert them into GitHub issues and the daemon picks and works on the issues. Once you get used to it bit the quality of code far exceeds what others can do

[Generalist @gendev01]

@KBezbailis I think you need to use https://t.co/NdWN3vP0SP haven’t launched it yet but can help you set it up

[Generalist @gendev01]

@johnennis Unless the bubble bursts before inference gets cheaper and models get smaller

[Generalist @gendev01]

@shcallaway good if they are integrated, that means you are thinking right. If not great you just solved a problem that you own.

The alpha is ability to solve problems you have instead of being locked in.

[Generalist @gendev01]

@zarazhangrui whose even conducting hackathons now ?

[Generalist @gendev01]

@DenLoginoff Few things I tried
1. Generating 3d models ( u can ask fable to work with stl files )
2. Resume rewrite (I know it is latex)
3. Get it to write a book for a new domain I am learning ( 544 pages long )

It depends entirely on how far you are pushing urself and the agent tbh

[Generalist @gendev01]

If your agent harness needs a terminal open so you can watch it, it isn’t infrastructure yet.

[Generalist @gendev01]

https://t.co/YMlmeuTrcp

[Generalist @gendev01]

@thsottiaux I am not a company but have pretty decent lights off factory, my reason is rather simple sol invents too much of unnecessary complexity, would love a guide for moving from fable to sol prompting

[Generalist @gendev01]

@boardyai Sure

[Generalist @gendev01]

is there anyone building lights off software factories ? would love to exchange notes