How AI generated code accelerates technical debt
A clear signal from the engineering frontlines: when AI velocity outpaces human review, UX must step in to govern intent and ensure system comprehensibility.
carlsz.dev Main Feed
The main feed for everything on carlsz.dev
Latest posts
Welcome to the age of the Slop Fork
As AI makes cloning software nearly free, UX leverage moves entirely to authoring the specifications and verification suites that guide the machine.
The Hidden Costs of AI-Generated Software
Research reveals that pairing AI development with rigid architectural constraints is the key to preventing technical debt and unlocking true long-term velocity.
The Mythical Agent Month
Throwing a swarm of AI agents at a late project won’t save it, it just trades the communication bottleneck for a verification bottleneck. The solution is Intent Architecture.
Eigenquestions: The Art of Framing Problems
To effectively direct autonomous agents, UX leaders can master the art of finding the eigenquestion—the single framing choice that dictates all subsequent systemic decisions.
Writing AI coding agent context files is easy. Keeping them accurate isn't.
As the agentic workforce scales, the true art of context engineering shifts from file creation to fighting metadata drift.
Emerging AI Antipatterns
The latest Thoughtworks Radar highlights the importance of the Middle Loop by identifying the need to balance rapid AI generation with curated architectural guidance.
Onboarding the Machine
To orchestrate an autonomous workforce, UX and engineering leaders have the opportunity to transition from writing static documentation to actively engineering the agent subconscious.
Magentic-UI: Towards Human-in-the-loop Agentic Systems
Microsoft’s release of Magentic-UI validates the need for Openable Box interfaces and explicit Action Guards to safely collaborate with agents in the Middle Loop.
AI Doesn’t Reduce Work—It Intensifies It
As autonomous agents accelerate production, the cognitive load of verification highlights a massive opportunity to evolve our workflows.
The risks of agentic chaos
As autonomous AI agents proliferate, robust infrastructure governance and access controls become the defining requirements of the Middle Loop.
The Approval Interface
As generation ceases to be the bottleneck, UX leaders have the opportunity to design human-in-the-loop Approval Interfaces that effortlessly govern agentic execution.
How an AI-enabled software product development life cycle will fuel innovation
McKinsey research validates that as AI collapses traditional SDLC roles, embedding automated quality and compliance guardrails up front becomes essential.
Diamond Prompting in UX Work
Adapting the classic double-diamond design process to AI workflows provides a structured framework for systematic context engineering.
The Factory Model: How Coding Agents Changed Software Engineering
When software development shifts from writing code to orchestrating fleets of autonomous agents, the spec becomes your ultimate point of leverage.
The Kaizen of Context
When AI allows us to prototype in production at machine speed, Test-Driven Design becomes our most effective form of prompt engineering.
Design as Infrastructure: Bridging the Vibe Coding Gap
As coding agents gain visual awareness via the Model Context Protocol (MCP), UX design transitions from static artifacts to live, queryable databases.
MCP Apps: Bringing UI Capabilities To Agents
The Model Context Protocol now allows agents to render interactive interfaces, creating a new frontier for UX designers building orchestration tools.
Introducing A2UI: Safe, Agent-Driven Interfaces
Google’s A2UI protocol validates the shift toward declarative Generative UI, prioritizing structural intent over raw code generation.
Say What You Mean: Design Adaptation for an Agentic World
To thrive in the agentic era, UX professionals can stop handing AI visual puzzles and start providing explicit, machine-readable answer keys.
Vibe Coding vs. Agentic Coding: The Taxonomy of Intent
A new academic taxonomy validates that as we scale to managing multiple autonomous agents, UX must focus on the interface of orchestration.
Designing the Catalog
To unlock the full potential of autonomous agents, teams can shift from exporting static images to curating programmable component catalogs.
Understanding builder intent in the AI era
DORA research reveals that safely delegating tasks to AI requires UX designers to architect the Openable Box for maximum transparency.
English will become the most popular development language in 6 years
As software creation moves to spec-centric paradigms, UX professionals must master Context Engineering to mitigate the ambiguity of natural language.
The Software Development Lifecycle Is Dead
As AI collapses the SDLC, UX leverage shifts from manual design to defining strict verification constraints and observability.
The Evolution of Craft
As AI shifts from an assistant to an autonomous workforce, the UX profession must transition from manual artifact production to supervisory intent architecture.
Something Big Is Happening
As AI models automate the technical execution of software development, UX leaders must pivot to governing intent in the Middle Loop.
Hello World
As AI transforms into an autonomous workforce, UX and engineering must evolve from manual artifact production to Intent Architecture.