Post

@augmentcode I work with customers who have codebases that are millions and tens of millions of lines. They need an OSS AI memory and context layer like @MemMachine_ai
github.com/MemMachine/Mem…
English

@augmentcode Been saying this.
We don't need AGI, we need ACI.
Artificial Contextual Intelligence.
To be AGI it requires ACI, imo.
English

@augmentcode maybe the Evolutive aspect of the Context is Needed to perform Solving
English

@augmentcode ELON MUSK, THIS IS WANDA PITTS GARY. I HAVE YOUR # 618. AND I PURCHASED A TESLA MEET AND GREET. NEED ADVICE DO YOU DEAL WITH WHATSAPP. I HAVE MORE INFO
English

@augmentcode And log files, don’t forget to tell augment to tail your logs and debug your app for you :)
English

@augmentcode 100%! Once you move from POC and simple code, you’re forced to break your code base up into modules that fit the context window. You can also use AI to generate a framework for assembling the modularity. It works!
English

@augmentcode we've been saying this for years but everyone kept throwing bigger models at the problem. turns out rag and context windows matter more than another trillion parameters who knew
English

@augmentcode I created a `fid` command that uses a little-known ability of `find` to create a digest of source files, ignoring binary files to speed up grep. It searches the whole codebase as if it was a single mmap'd document.
English

@augmentcode Context management + semantic/symbolic retrieval is key for high quality agents
English
@augmentcode Interesting. But how are you handling codebase/data security?
English

@augmentcode it's not impressive at coding tasks. engineers spend 20% more time on code generated openai. give this shit a rest
English

@augmentcode I've built a system context file into my code base - I generally ask the AI to run a new thread and run as long as I can before it maxes the context window and update the context. What's the benefit of using your app ? Also 600 msgs seems low?
English




