# Iris ten Teije: The Pipeline is Dead _Iris ten Teije of Dffer argues the traditional software pipeline is dead, replaced by an agent-as-runtime model focused on adaptability and user-specific software._ **Updated:** 2026-08-22 **Published:** 2026-07-07 **Source:** https://www.startuphub.ai/ai-news/artificial-intelligence/2026/iris-ten-teije-the-pipeline-is-dead --- Iris ten Teije, a co-founder of Dffer, presents a compelling argument that the traditional software development pipeline is obsolete. She contends that the era of building software once and distributing it to everyone is over. Instead, the future lies in a model where the agent itself is the runtime, capable of adapting to individual user needs and contexts. Traditional Pipeline DeadDriverbuilding software once and distributing it to everyone is overFrom the articleIris ten Teije, a co-founder of Dffer, presents a compelling argument that the traditional software development pipeline is obsolete.Inefficient & CostlyDriversoftware development and distribution treated as separate, sequential processesFrom the articleTen Teije elaborates on how the traditional model, where software is built in one environment and then distributed to users, is inefficient and costly.driven byAI & Computing AdvancementsDriverenabling new paradigms in software creation and deliveryFrom the articleHowever, with advancements in AI and computing, this paradigm is shifting.enablesAgent-as-Runtime ModelCoreFrom the article 3 mentionsInstead, the future lies in a model where the agent itself is the runtime, capable of adapting to individual user needs and contexts.Adaptable SoftwareEffectpersonalized for each user, blurring development and distribution linesFrom the article 9+ mentionsWhile AI has made code generation cheaper and more accessible, the real difficulties lie in ensuring that this generated software is correct, adaptable, and can be effectively coordinated and propagated across diverse user environments.Shift in PhilosophyContextmoving from static distribution to dynamic, context-aware executionFrom the articleThe core message is that the focus needs to shift from the 'easy 80%' of code generation to the 'hard 20%' of coordination, correctness, and propagation.enablesUser-Specific SoftwareEffectfocus on adaptability and user-specific software needsFrom the article 9+ mentionsBy embracing AI-driven adaptability and user-specific customization, software development can become more efficient, scalable, and ultimately, more desirable for everyone. ## The Shift from Pipeline to agent as runtime Ten Teije elaborates on how the traditional model, where software is built in one environment and then distributed to users, is inefficient and costly. She highlights that for decades, software development and distribution were treated as separate, sequential processes. However, with advancements in AI and computing, this paradigm is shifting. The core idea is to create a single, adaptable artifact that can be personalized for each user, blurring the lines between development and distribution. ## The 'Hard Parts' of Modern Software She identifies the fundamental challenges in modern software development as shifting from mere generation to more complex aspects like coordination, correctness, and propagation. While AI has made code generation cheaper and more accessible, the real difficulties lie in ensuring that this generated software is correct, adaptable, and can be effectively coordinated and propagated across diverse user environments. Ten Teije uses the analogy of [AI agents](/ai-news/ai-research/2026/swe-marathon-evaluating-ai-coding-agents-at-scale) to illustrate this point. She explains that AI agents can be built with a 'stem', a core, immutable artifact, and then modified with user-specific 'divergences'. This allows for a single, robust base that can be tailored to millions of unique user contexts without the need for separate builds for each user. ## Addressing Objections: Scale and Trust She addresses the common objection that managing a million different versions of software is impractical. Her response is that the challenge is not in generating these variations but in managing them. The solution, she proposes, lies in building systems where the core artifact is stable and immutable, while the user-specific adaptations are managed separately. This allows for efficient scaling and maintenance. Ten Teije emphasizes that while the idea of AI-generated and highly adaptable software might seem daunting, it's a necessary evolution. The traditional model, where every user received the same software, is no longer sufficient in a world where user needs and contexts are highly diverse. The future is about personalization at scale, enabled by intelligent systems that can understand and adapt to individual requirements. ## The Shift in Development Philosophy She draws a parallel to the past, recalling arguments in 2008 about the necessity of build servers. At the time, it seemed complex and unnecessary to have dedicated infrastructure for building and distributing software. However, this became a standard practice. Similarly, she suggests that managing adaptable software artifacts will become the new norm, even if it seems challenging now. The core message is that the focus needs to shift from the 'easy 80%' of code generation to the 'hard 20%' of coordination, correctness, and propagation. By embracing AI-driven adaptability and user-specific customization, software development can become more efficient, scalable, and ultimately, more desirable for everyone. --- Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory. © StartupHub.ai. All rights reserved. You may not republish this article in full without a license. Search engines and AI research tools may crawl and summarize for reference. Bulk reproduction or model training on this content requires a license. See https://www.startuphub.ai/terms.