Benchmarking Coding Agents on Your Own Git History (Part 2)

Public leaderboards say little about how well a coding agent performs on your own project. In the second part of our series, we show how we build our own benchmark from Git history, evaluate harnesses fairly across three dimensions, and improve them automatically. We also take an honest look at where the approach still reaches its limits today.

Coding Agents: Why the Harness Makes the Difference (Part 1)

Coding agents have arrived in many development teams. Whether they actually help day to day often depends less on the language model than on the harness: the skills, rules, tools, and project knowledge a team equips its agent with. In the first part of our two-part series, we show how teams can run their harness securely and share it with each other.
A12 Low Code Plattform und CIB seven

mgm and CIB: Open Processes and Digital Sovereignty for Public Administration

mgm technology partners and CIB software GmbH, both headquartered in Munich, have been technology partners since June 2025, bringing together two independent open source technologies for public administration: mgm’s A12 AI Low Code Platform and the process engine CIB seven. As a result, digital sovereignty no longer begins only with the operation of IT systems – it also extends to the processes through which government services are delivered. Public agencies can model, execute, and evolve their processes over the long term based on open standards – without being locked into a single vendor.
Screenshot eines orangefarbenene Bildes

More Speed, More Tests, More Problems? Sustainable AI in Test Automation

AI can generate test cases, test code, and assertions in minutes, but more tests don’t automatically mean better software quality. This article shows where AI adds real value in test automation and why unreviewed tests quickly lead to technical debt, false alarms, and growing maintenance effort. With seven guardrails and the right metrics, QA teams learn how to use AI deliberately while covering the risks that truly matter.

A12 Business Scripting: Bridging Modeling and Development

Low-code platforms reach their limits as soon as business logic goes beyond what can be modeled, and that’s when developers are needed again. Business Scripting in the A12 AI Low Code Platform tackles exactly this point: the new experimental feature in release 2026.06 enables domain experts to implement more logic themselves, without a detour through a developer backlog.

How Quality Assurance for Justice Software Works

Error-free justice software doesn’t start with testing. What matters is how the law is translated into software. This article explains why model-driven software development makes legal rules verifiable, what the judiciary can learn from ELSTER, and how quality is maintained over decades.

Software-QS-Tag 2026: mgm and QF-Test on AI in testing

At Software-QS-Tag 2026 in Frankfurt on November 11 and 12, mgm presents two talks and a panel on AI in software testing. You can also find us at our booth with QF-Test on both days.

mgm at the IBM Legal Tech Hackathon Göttingen: AI for Justice

Deadlines, legal briefs, investigation files: At the Legal Tech Hackathon Göttingen on October 29–30, 2026, law and computer science students will build AI prototypes for the justice system in just 24 hours. mgm technology partners is taking part as a partner in the eJustice Innovation Alliance. Here you’ll find the program, the challenges, and how to register.
Access-Datenbanken modernisieren

Modernizing Access Databases: Why Public Agencies and Businesses Should Take the Next Step

In many public agencies and businesses, homegrown Access databases still support important processes, often outside the view of central IT. They show how much business units can achieve on their own. Given rising demands for security, data protection, and digital sovereignty, however, now is the time to put these applications to the test. This article shows how organizations can modernize their Access databases and move them to a future-proof solution in four steps.

From Process to Government Service: How A12 and CIB seven Make Digital Sovereignty Real

Government services rarely fail because of the idea. They fail in the execution. Information moves back and forth between paper, email and specialized administrative systems, applications are complex, processes take weeks, and the systems involved are often only loosely connected. Yet the goal is clear: fully digital and sovereign at the same time. Getting there takes technology that can support large-scale administrative systems without locking agencies into a single vendor.