Berlin, Germany – April 24, 2026 – Delivery Hero, a global leader in local delivery services, today announced the unveiling of Herogen, an advanced autonomous software delivery agent. Introduced to the company's product and engineering teams in February, Herogen has already demonstrated a remarkable capacity to enhance engineering output. Built in-house using leading Large Language Models (LLMs), Herogen is currently delivering an annual coding output equivalent to 130 senior engineers, with its capabilities expanding rapidly.
Herogen operates autonomously, freeing up human engineers to focus on higher-level tasks such as system architecture and feature ideation. The system allows engineers to delegate coding work in natural language, after which Herogen writes, tests, and iterates on the code before submitting it for a final human review. A sophisticated "council of agents," leveraging multiple LLMs from different providers, reviews the code from various perspectives to mitigate potential blind spots. Currently, Herogen boasts an 85% success rate, measured by the ratio of merged to rejected pull requests, with most tasks requiring minimal human intervention.
With an 18% rollout completion across Delivery Hero's technology teams, Herogen is already responsible for 9% of all code change requests. The agent autonomously merges over 100 code change requests daily, effectively freeing up an estimated 250,000 hours of manual coding annually. This increased efficiency is a key component of Delivery Hero's strategic transition into a global "Everyday App."
The development of Herogen underscores the growing trend of AI integration in software development, aiming to boost productivity and accelerate innovation within the tech industry. This advancement is expected to significantly impact how software is developed and deployed across the sector.