dearwolf/aider
1
0
Fork 0
mirror of https://github.com/Aider-AI/aider.git synced 2025-06-05 12:14:59 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

copy

Browse source
This commit is contained in:
Paul Gauthier 2024-07-01 16:37:16 -03:00
parent 1a6ea77976
commit e3d6bba6c1
1 changed files with 27 additions and 17 deletions
Download patch file Download diff file Expand all files Collapse all files

44
website/_posts/2024-07-01-sonnet-not-lazy.md
View file

@ -15,8 +15,11 @@ nav_exclude: true
Claude 3.5 Sonnet represents a step change Claude 3.5 Sonnet represents a step change
in AI coding. in AI coding.
It is so industrious, diligent and hard working that It is incredibly industrious, diligent and hard working.
it has caused multiple problems for aider. Unexpectedly,
this initially presented a few challenges
that prevented aider from taking maximum advantage of
Sonnet's capabilities.
It's been worth the effort to adapt aider to work well It's been worth the effort to adapt aider to work well
with Sonnet, with Sonnet,
@ -28,46 +31,53 @@ as a result of the changes discussed below.
This moved Sonnet into second place, ahead of GPT-4o and This moved Sonnet into second place, ahead of GPT-4o and
behind only Opus. behind only Opus.
## Problems ## Hitting the 4k token output limit
Sonnet's amazing work ethic caused a few problems: All LLMs have various token limits, the most familiar being their
context window size.
But they also have a limit on how many tokens they can output
in response to a single request.
Sonnet and the majority of other
models are limited to returning 4k tokens.
Sonnet's amazing work ethic caused it to
regularly hit this 4k output token
limit for a few reasons:
1. Sonnet is capable of outputting a very large amount of correct, 1. Sonnet is capable of outputting a very large amount of correct,
complete code in one response. complete code in one response.
So much that it can easily blow through the 4k output token limit 2. Similarly, Sonnet can specify long sequences of edits in one go,
on API responses, which truncates its coding in mid-stream.
2. Similarly, Sonnet can specify large sequences of edits in one go,
like changing a majority of lines while refactoring a large file. like changing a majority of lines while refactoring a large file.
Again, this regularly triggered the 4k output limit 3. Sonnet tends to quote large chunks of a
and resulted in failed edits. file when performing a SEARCH & REPLACE edits.
3. Sonnet is not shy about quoting large chunks of an Beyond token limits, this is very wasteful.
existing file to perform a SEARCH & REPLACE edit across
a long span of lines.
This can be wasteful and also trigger the 4k output limit.
## Good problems ## Good problems
Problems (1) and (2) are "good problems" Problems (1) and (2) are "good problems"
in the sense that Sonnet is in the sense that Sonnet is
able to write more high quality code than any other model! able to write more high quality code than any other model!
We just don't want it to be interrupted prematurely
by the 4k output limit.
Aider now allows Sonnet to return code in multiple 4k token Aider now allows Sonnet to return code in multiple 4k token
responses. responses.
Aider seamlessly combines them so that Sonnet can return arbitrarily
long responses.
This gets all the upsides of Sonnet's prolific coding skills, This gets all the upsides of Sonnet's prolific coding skills,
without being constrained by the 4k output token limit. without being constrained by the 4k output token limit.
## Wasting tokens ## Wasting tokens
Problem (3) does cause some real downsides. Problem (3) is more complicated, as Sonnet isn't just
being stopped early -- it's actually wasting a lot
of tokens, time and money.
Faced with a few small changes spread far apart in Faced with a few small changes spread far apart in
a source file, a source file,
Sonnet would often prefer to do one giant SEARCH/REPLACE Sonnet would often prefer to do one giant SEARCH/REPLACE
operation of almost the entire file. operation of almost the entire file.
This wastes a tremendous amount of tokens,
time and money -- and risks hitting the 4k output limit.
It would be far faster and less expensive to instead It would be far faster and less expensive to instead
do a few surgical edits. do a few surgical edits.