Zencoder omitted Solver from their comparisons. Given how new Solver is, that’s comprehensible. Nonetheless, Solver is much more promising than Zencoder, because it not solely reads complete repositories, it may possibly modify complete repos. You possibly can learn my evaluate of Solver right here.

On the left, you may see Zencoder’s clarification of a C++ sorting operate; on the proper, you may see GitHub Copilot’s clarification of the identical code utilizing the o1-preview mannequin. On the high you may see the Zencoder capabilities obtainable in VS Code within the Zencoder v0.5.1 plugin. General, I like the reason on the suitable higher, however I’m keen on a documentation fashion that mixes phrases and code.
IDG
An bold try
Zencoder is, in my view, an bold try and enter the code technology and restore market, however it’s nonetheless moist behind the ears. Utilizing a pipeline with error correction to keep away from hallucinations, as an alternative of straight code technology with an LLM, sounds promising, however it isn’t ipso facto higher than a mannequin that generates right code proper out of the field and suppresses hallucinations internally. Equally, “grokking” complete repositories ought to present higher context than simply no matter pages are open within the editor, however the time it takes to digest massive repos is critical, and it seems to be like Zencoder redoes that course of from scratch everytime you refresh the repo.
I’m seeing fairly a little bit of convergence on this market. Zencoder’s two improvements are grokking a complete repo for context and utilizing an error-corrected inference pipeline. Solver and brokers from Amazon and GitHub can modify complete repos. It’ll be fascinating to see what occurs as all these efforts cherry-pick one another’s methods.