As developers start to embrace generative AI coding tools, we need to know how robust they are. We also need to know how capable they are of showing relevant links to existing software code bases and to real world business logic. Describing its software as a generative-AI powered coding assistant for contextualized code understanding, Tel Aviv-based Swimm has now launched /ask Swimm. This is a tool for software development teams that combines an AI-powered chat alongside human input to provide personalised and contextualised code knowledge.
Challenging to type with its slash-space-lowercase intro, /ask Swimm is an AI-powered coding assistant that provides developers with a multilayered contextual understanding of code, enabling them to significantly boost productivity across the entire development lifecycle.
Contextualised code understanding
Software developers are continuously trying to ship code faster yet often struggle due to a lack of good code documentation hygiene and limited contextualized code understanding, which can result in answers that are too generic and unhelpful.
But what is contextualized code understanding?
This term is meant to cover all the elements of supporting information a developer might need to be able to operate, run and extend a typically pre-existing software project. It might include initial software requirements gathering information, initial architecture designs, UML plans to detail procedural steps in the software development lifecycle, notes on team member competencies past and present, details relating to debugging tasks already carried out which may provide insights into any future fragilities that surface and so on and so on.
“Code can be understood at a tactical level e.g. ‘what does this function do?’ might be a question we could ask,” explained Oren Toledano, CEO and co-founder at Swimm. “Many LLM-based tools now excel at this task and can come up with explanations or comments quickly. However, the main challenge for developers is understanding how complex flows and processes work as a whole, and why certain decisions were made. Swimm specialises in helping developers understand the story behind the code in all its complexity. Context here refers to the context in which a particular part of the code was written, and not only what it does. It may also refer to the business or technical context underlying the way the code was implemented.”
Swimm suggests that software developers spend hours a week looking for information to understand code or to answer questions about it. This highlights the current gap that exists between the need for effective code understanding and documentation, which is often required when debugging issues with code or when onboarding new developers.
In the context of /ask Swimm, contextualized code understanding takes the form of a generative AI GenAI coding assistant that promises to provide the next step in the evolution of code understanding by pairing human input with AI for multi-layered contextual awareness that utilizes knowledge from the codebase and non-code related information.
Personalised dynamic chat
The chat is dynamic and personalized to an organization’s specific codebase, documentation, user interactions and other 3rd party tools. A contextualized conversational chat within the IDE instantly enables developers to answer any questions about documentation, code, files, repositories or even entire software ecosystems.
To gain context, /ask Swimm incorporates factors that are not evident in the code itself, such as business decisions, product design considerations, limitations that were the basis for roads not taken, etc. /ask Swimm automatically captures and updates code-related knowledge in the process, while improving over time with continuous feedback and user-generated documents.
“Some of the biggest names in tech, including Microsoft (Copilot), Google (Duet AI), Amazon (CodeWhisperer), are building code assistance tools. They’re competing on the merits of their analysis capabilities and the LLMs they’re using, but no matter how good they are, a tool doesn’t know what it doesn’t know because it’s lacking the relevant context,” said Toledano. “To provide the most complete picture to engineers, we’ve developed /ask Swimm which is the logical next step in the evolution of code understanding. Context is invaluable especially when refactoring legacy code or trying to understand complex flows and processes in a codebase. By keeping documentation and context up to date with existing code, processing it and feeding it in time to the LLMs, Swimm has solved one of the biggest challenges developers face today.”
The Swimm leadership team says it understands software engineers and their pain points intimately as they’ve lived it themselves. They’ve taken the historically manual and outdated process of code documentation and completely reimagined it using AI. Swimm promises to help engineering teams transform the developer experience and improve code efficiency, reliability and consistency, while continuously increasing their documentation coverage, all without the upfront investment and productivity time loss.