- I’m not clear what “just loading the project” even means here - if that’s how many tokens are consumed by system prompt plus Claude.md and MCP tools well that has nothing to do with the size of the project
- Can’t confirm or deny comparison with JS but I can second that it write decent elixir
The only problem I’ve ever had was on maybe 3 total occasions it’s added a return statement, I assume because of the syntax similarity with ruby
- Hah I’m only on the cutting edge part time on the side so my experience has been more like - start thinking about the problem and then 2 or 3 days later new tools come out that solve it for me
- i think you have an error there about haiku pricing
> For comparison, Sonnet 4.5 is $3/$15 and Haiku 4.5 is $4/$20.
i think haiku should be $1/$5
- I’ve been surprised at the lack of discussion about sourcegraph’s Amp here which I’m pretty sure you’re referring to - it started a bit rough but these days I find that it’s really good
- Yes that was the one point in here where I thought to myself oh yeah I’m going to go implement that immediately
- You are maybe confusing caching and context windows. Caching is mainly about keeping inference costs down
- the agent greps for the obvious term or terms, reads the resulting documents, discovers new terms to grep for, and the process repeats until its satisfied it has enough info to answer the question
> You could expand grep queries with synonyms, but now you're reimplementing query expansion, which is already part of modern RAG.
in this scenario "you" are not implementing anything - the agent will do this on its own
this is based on my experience using claude code in a codebase that definitely does not have consistent terminology
it doesn't always work but it seemed like you were thinking in terms of trying to get things right in a single grep when it's actually a series of greps that are informed by the results of previous ones
- Steve Yegge has a great story about Google HR doing an exercise with the eng hiring committee where they reviewed their own anonymized interview packets and they decided not to hire 40% of themselves.
- > I have gone from using Claude Code all day long since the day it was launched to only using the separate Claude app. In my mind that is a nice balance of using it, but not too much, not too fast.
I’m on a similar journey - I never used it all day long but definitely a lot during a brief honeymoon period and now I’m back to using it very sparingly but I put questions to the Claude app all the time
For me the sweet spot for Claude code is when I have a very clear and well documented thing to set up that I really don’t want to do for the umpteenth time - like webhook signature verification - just paste the docs and let it rip - or setting up the most basic crud forms for an admin dashboard - ezpz
But otherwise I’ve gone back to mostly writing everything by hand
- Not to trivialize the work being done here but isn’t this as simple as a hook on edit and write tool calls that commits to git? I’m not sure I see the need for a whole app around this vs just the standard git workflow
- yup was building it as a linear agent https://linear.app/developers/agents
- Very cool! I’ve been building an internal tool at work that’s very similar but primarily focused on automatically triaging bugs and tech support issues, with MCP tools to query logs, search for errors in bugsnag, query the db etc. also using linear for issue tracking. They’ve been launching some cool stuff for agent integrations.
And sorry I’m a light mode fan
- I think this is true about many different aspects of programming languages
There are many things that ultimately come down to personal taste more than some sort of objective pros and cons list, as much as people will attempt to argue otherwise
- Made me think about John Wilkins' "philosophical language" which I first heard about in Neal Stephenson's book Quicksilver
https://en.wikipedia.org/wiki/An_Essay_Towards_a_Real_Charac...
I'm sure there have been countless similar attempts at categorizing knowledge
one of the more successful ones being the dewey decimal system
I have my doubts about whether the thing the OP alleges we have "failed" at is even possible at all
- Source graph amp is pretty good as well albeit lacking a lot of the polish and features of Claude code
But I sometimes reach for it for code review in particular since it calls out to o3 via its “oracle” tool
- FWIW you can run Claude code async via GitHub actions and have it work on issues that you @ mention it from - there’s even a slash command in Claude code that will automatically set up your repository with the GitHub action config to do this
- love the ability to add tools to the mcp server - would expect nothing less from emacs :)
as a long time emacs user i've only recently started really writing my own elisp tools, but claude is pretty good at writing elisp so i've been doing more there (sometimes it loses track of parentheses and you need to fix that, but overall pretty good)
I'll def be trying this out alongside steve yegge's efrit which kicks the emacs up to 11 by letting the agent just write and evaluate arbitrary elisp expressions https://github.com/steveyegge/efrit
- With the caveat that I have not spent a serious amount of time trying to get RAG to work - my brief attempt to use it via AWS knowledge base to compare it vs agentic search resulted in me sticking with agentic search (via Claude code SDK)
My impression was there’s lots of knobs you can tune with RAG and it’s just more complex in general - so maybe there’s a point where the amount of text I have is large enough that that complexity pays off - but right now agentic search works very well and is significantly simpler to get started with
Even without AI you cannot do a tight 10 minute video on legacy code unless you have done a lot of work ahead of time to map it out and then what’s the point