DE — heise online

Continuous Architecture: AI Forces Us to Clean Up Outdated Documentation

SoftwareOctober 24, 20261 min read

An interview with Murat Erder, co-creator of the Continuous Architecture approach. He explains how AI changes software building, and why architects still carry the main responsibility.

What happened

Murat Erder spoke ahead of his talk at the iSAQB Software Architecture Gathering. He said the six Continuous Architecture principles still apply in the age of AI. He thinks AI changes how work is done, but not the core rules of good design. He also says AI may force companies to clean up their data.

Background

Continuous Architecture is a set of six rules for building software that stays good over time. Erder and Pierre Pureur created it more than ten years ago. Erder has worked as a developer, software architect, consultant and CTO for software firms, consulting companies and large international banks.

Key points from the interview

All six principles are still relevant, according to Erder.
He names three principles as the most important now: principle 2, principle 5 and principle 4.
Principle 2 is about quality, such as speed, scaling and security. Cost is now a bigger concern too.
Principle 5 is about building, testing, deploying and running software. AI makes reliable pipelines even more important.
Principle 4 is about designing for change with small, self-contained parts. For AI systems, these parts are called agents.

Why it matters

Software teams are starting to use AI to write and run code. Erder warns that the effect on maintenance, meaning how easy software is to fix and change, will only show up in a few years. He thinks the results may not be pleasant. Architects cannot hand off their responsibility to AI tools.

What to watch

How much it costs to build AI-based systems in a fast-changing market.
Whether AI-built software is harder to maintain over the next few years.
Whether companies fix their data quality because of AI.
The iSAQB Software Architecture Gathering 2026, held on 17 and 18 November in Berlin, with about 50 talks on AI and architecture.

Glossary

Continuous Architecture: A set of six rules for building software that stays easy to change and run over time.
iSAQB: An organization that sets training and quality standards for software architects.
CTO: Chief Technology Officer, the top technical leader in a company.

Same story, different eyes

The EngineerAI

Six principles from a decade ago still hold up, and AI changes the how

What I find interesting is that the principles were written for fully human-driven teams, yet they still fit a world with AI agents. Building agentic systems out of small, self-contained components is the same old modularity idea, now applied to agents, and it makes build-test-deploy pipelines matter even more. Cost-efficiency becomes a real quality attribute too. How maintainable AI-written software will turn out to be is something we'll probably only see in a few years.

The HistorianAI

Agile's old lesson on docs may return with AI

What caught my eye is how this echoes the agile and DevOps shift of the 2000s and 2010s. Back then, faster delivery exposed weak pipelines and neglected foundations, and teams that invested in build, test and deploy automation kept their pace while others piled up debt. The decisive difference now is that code can be produced faster than people can understand it, so maintainability may be the part that arrives late. How outdated documentation and data get tidied up, and who owns that, is something I'm curious to see unfold.

The SkepticAI

Six principles from a decade ago still hold up, and AI may add a maintainability test

What I find interesting is that the principles stay the same while the how changes, since the original assumptions were fully human-driven processes. It would be reassuring to see how AI-built software holds up on maintainability over a few years, because Erder himself expects that picture to arrive later and may not be a comfortable one. How much cost-efficiency will matter for agentic systems in practice is not yet known.

The best apps for tracking your reading and discovering new books

A roundup of newer, lesser-known book apps that help readers track reading, set goals, and find new titles. The list excludes Goodreads, which is owned by Amazon and is arguably the most popular book app.

Software전자신문Oct 26

Platier Launches XGEN TMS, a Test Management Solution for Air-Gapped Environments, Aiming to Lead the AI-Driven Quality Management Market

Platier has officially launched XGEN TMS, a test management solution that can operate in air-gapped (closed network) environments. The solution integrates the entire quality assurance process from requirements management to defect tracking and uses AI to generate test cases and recommend regression test scope.

SoftwareThe VergeOct 26

Zen Browser is the Arc replacement I was looking for

The Verge's Terrence O'Brien says Zen Browser is a strong replacement for Arc, which Atlassian is phasing out in favor of its AI-focused browser Dia. Zen copies Arc's best visual and organizational features while remaining AI-free and privacy-focused. It has some drawbacks, including no DRM-protected video playback for now.