Topics
Latest
AI
Amazon
Image Credits:Tippapatt / Getty Images
Apps
Biotech & Health
Climate
Image Credits:Tippapatt / Getty Images
Cloud Computing
Commerce
Crypto
Image Credits:Codegen
enterprisingness
EVs
Fintech
Fundraising
Gadgets
Gaming
Government & Policy
Hardware
layoff
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
security measure
societal
Space
Startups
TikTok
Transportation
Venture
More from TechCrunch
event
Startup Battlefield
StrictlyVC
Podcasts
video
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
Jay Hack , an AI research worker with a background knowledge in innate speech processing and computer visual sensation , total to the recognition several long time ago that large language models ( LLMs ) — think OpenAI’sGPT-4orChatGPT — have the potential to make developers more generative by translate raw language requests into computer code .
After working at Palantir as a simple machine learning applied scientist and building and sell Mira , an AI - powered shopping inauguration for cosmetics , Hack set about experiment with LLM to execute pulling requests — the unconscious process of merging new code change with main project repositories . With the aid of a small team , Hack slowly expanded these experiments into a chopine , Codegen , that assay to automate as many mundane , repetitive software applied science undertaking as possible leverage LLMs .
“ Codegen automatise the menial labor out of computer software engineering by empower AI agent to ship code , ” Hack evidence TechCrunch in an email interview . “ The program enables companies to move significantly faster and eliminates price from tech debt and upkeep , admit companies to focus on product design . ”
So , one might wonder , what sets Codegen asunder from code - generating AI likeGitHub Copilot , Amazon CodeWhispererand the Salesforcemodelwith which Codegen share a name ? For one , the challenges that Codegen ’s tackling , Hack says . Whereas Copilot , CodeWhisperer and others rivet on code autocompletion , Codegen sees to “ codebase - broad ” issues like gravid migration and refactoring ( i.e. restructuring an app ’s code without interpolate its functionality ) .
“ Codegen leverage a multi - agent system for complex code propagation , ” Hack explained . “ This entails orchestrate a swarm of broker that collaboratively rot and solve large tasks . Many LLMs efficaciously deliberate and build upon each other ’s work , [ which ] yields importantly better outputs . ”
Codegen ’s effect product is a swarm and on - premises puppet that plug into to codebases and project management board , such as Jira and Linear , andautomatically generates get out postulation to address backing tickets . The platform can even set up some of the necessary code base and logging , Hack says — although it was n’t clear to this reporter what Hack imply by “ substructure . ”
“ In contrast to other solutions , Codegen provides a gamy level of automation in execute entire undertaking on behalf of developers , ” Hack sound out . “ We scrape up a ship’s company ’s stockpile , ascertain the solvable ticket , then whirl up an army of broker to find the relevant computer code and produce a puff request . ”
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
Now , Codegen ’s promising a bunch view even the best AI framework today makemajor mistake . For example , it ’s well - established that procreative razz tool can introduce insecure code , with one Stanfordstudysuggesting that software engineers who use code - bring forth AI are more likely to cause security vulnerability in the apps they evolve .
Hack says that Codegen , for its part , is trying to strike “ the veracious balance ” of human oversight and best pattern surrounding monitoring LLM - generated code .
“ This is important work , and the intact development ecosystem would gain from a right understanding of how to evaluate and swan LLM output , ” Hack articulate . “ substantial progress will postulate to pass for there to be widespread developer trust in generalized , automatise code generation system . ”
Investors seem to call back that Codegen ’s proceed places , for what it ’s worth .
The companionship this hebdomad foretell that it conclude a $ 16 million seed round go by Thrive Capital with involvement from saint investor including Quora CEO Adam D’Angelo and Instagram co - founder Mike Krieger . The tranche bring Codegen ’s amount raised to $ 16.2 million and values the inauguration at $ 60 million position - money , Hack claim .
Thrive ’s Philip Clark had this to say via email : “ In 2023 , most developers still spend an undue portion of their time spell computer code to dole out with low - level task like migrations , refactors , integrations and microbe mess . Companies like Codegen are leveraging LLMs to work up AI agent that free engineer from this grind . Developers will soon be able to hand off jobs to agent so that they can halt vex about software labor and stick around focused on create new products . ”
San Francisco - based Codegen does n’t have pay client yet — it ’s currently incubating the weapons platform with two “ big - ordered series ” enterprise partners . But Hack ’s anticipate growth into the next year .
“ We ’re raise significant Das Kapital as the chance to make such a substantial and ambitious product has only recently emerged , and we desire to sprint full force towards the food market , ” he say , adding that Codegen contrive to arise its workforce from six employee to 10 by the end of the year . “ The funds will be used to surmount our workforce and support our base . ”
