Home » Codegen raises new cash to automate software engineering tasks.

Codegen raises new cash to automate software engineering tasks.

by Alex Turner
Image Credits: Tippapatt / Getty Images

Jay Hack, an AI researcher with a background in natural language processing and computer vision, came to the realization a few years ago that large language models (LLMs) — think OpenAI’s GPT-4 or ChatGPT — have the potential to make developers more productive by translating natural language requests into code. Hack came to this conclusion after working in natural language processing and computer vision.

Hack started exploring with LLMs to execute pull requests after working as a machine learning engineer at Palantir and founding and selling Mira, an AI-powered e-commerce firm for cosmetics. Pull requests are the process of integrating new code modifications with primary project repositories. These experiments gradually evolved into a platform called Codegen by Hack with the assistance of a small team. Codegen uses LLMs to automate as many menial and repetitive software engineering jobs as possible.

Get Posts Like This Sent to your Email
Iterative approaches to corporate strategy foster collaborative thinking to further the overall value.
Get Posts Like This Sent to your Email
Iterative approaches to corporate strategy foster collaborative thinking to further the overall value.

“Codegen automates the menial labor out of software engineering by empowering AI agents to ship code,” Hack said in an email interview with TechCrunch. “Codegen” “The platform enables companies to move significantly quicker and eliminates costs from tech debt and maintenance, allowing companies to focus instead on product innovation.”

Consequently, one may speculate what distinguishes Codegen from other AIs that generate code, such as GitHub Copilot, Amazon CodeWhisperer, and the Salesforce model with which Codegen has a name. First, according to Hack, some problems need to be solved by Codegen. Codegen addresses “codebase-wide” concerns such as massive migrations and refactoring, whereas Copilot, CodeWhisperer, and similar tools concentrate on code autocompletion. Refactoring refers to reorganizing an application’s code without affecting its functioning.

Hack explains that “Codegen leverages a multi-agent system to generate complex code.” “This requires directing a swarm of agents that can deconstruct and solve huge tasks collaboratively. A significant number of LLMs successfully deliberate and build upon each other’s work, [which] results in much-improved outcomes.

The fundamental product that Codegen offers is a tool that may be hosted in the cloud or on-premises. This tool can link to codebases and project management boards such as Jira and Linear and produce pull requests to resolve support problems automatically. According to Hack, the platform is even capable of putting in place part of the required code infrastructure and logging; however, it was unclear to this reporter exactly what Hack meant when he referred to “infrastructure.”

“In contrast to other solutions, Codegen provides a higher level of automation in executing entire tasks on behalf of developers,” said Hack. “We scrape a company’s backlog, find the tickets that can be solved, and then spin up an army of agents to find the relevant code and produce a pull request,” we explain.

Given that even the most advanced AI models available today are prone to making significant errors, Codegen has a lot of promise. For instance, it is common knowledge that generative coding tools can introduce unsafe code. Furthermore, research conducted at Stanford University suggests that software developers who use AI that generates code are more likely to introduce security flaws into the applications they create.

According to Hack, Codegen is doing its bit by attempting to find “the right balance” between human inspection and best practices for monitoring LLM-generated code.

“This is important work, and the entire development ecosystem would benefit from a better understanding of how to evaluate and verify LLM output,” said Hack. “This is work that we need to do.” “Significant advances will need to occur in order for there to be widespread developer trust in generalized, automated code generation systems.”

Investors appear to believe that Codegen has a bright future.

This Monday, the firm announced that it had completed a seed funding round of $16 million, headed by Thrive Capital, with participation from angel investors such as Adam D’Angelo, CEO of Quora, and Mike Krieger, co-founder of Instagram. According to Hack, the new sum raises Codegen’s total funding to $16.2 million and places a post-money valuation of the firm at $60 million.

Philip Clark from Thrive sent an email with the following statement: “In the year 2023, most developers still spend an unreasonable share of their time writing code to deal with low-level tasks like migrations, refactoring, integrations, and bug fixes.” Companies such as Codegen are utilizing LLMs to construct artificial intelligence agents that remove programmers from the monotony of their jobs. Soon, developers can delegate tasks to agents, allowing them to stop worrying about the tedious work of maintaining software and instead concentrate on developing new solutions.

Located in San Francisco Since Codegen is still incubating its platform with two “large-scale” business partners, the company does not yet have any paying clients. But Hack is optimistic about its prospects for expansion in the coming year.

“We’re raising significant capital as the opportunity to make such a substantial and ambitious product has only recently emerged, and we want to sprint full force towards the market,” he said, adding that Codegen intends to increase its staff from six people to ten employees by the end of the year. This money will be used to expand our personnel and strengthen our infrastructure.

You may also like

Leave a Comment