Navigating the AI Wave: Insights from ThoughtWorks' Technology Radar Volume 34

ThoughtWorks has just released the 34th edition of its Technology Radar, a biannual snapshot of the tools, techniques, platforms, and languages that have shaped the software development landscape. This volume includes 118 entries, each offering a brief but insightful take on a specific element of modern tech. As expected, artificial intelligence dominates the conversation — but the radar also reveals a surprising countercurrent: a renewed emphasis on the fundamentals of software craftsmanship.

The AI Dominance and a Return to Fundamentals

AI-oriented topics once again take center stage, with many entries exploring the implications of large language models (LLMs) for development workflows. Yet, the radar goes beyond the usual hype. It highlights how AI is forcing teams not only to look ahead but also to look back at the principles that underpin reliable software.

Navigating the AI Wave: Insights from ThoughtWorks' Technology Radar Volume 34 — Source: martinfowler.com

Revisiting Core Practices

While assembling this edition, the ThoughtWorks team found themselves gravitating toward established techniques. Pair programming, zero trust architecture, mutation testing, and DORA metrics all appear alongside newer AI-powered methods. Core principles such as clean code, deliberate design, testability, and accessibility are being re-evaluated as first-class concerns. This is not mere nostalgia; it is a necessary counterweight to the complexity that AI tools can generate at speed. Without a strong foundation, rapid AI-generated output can quickly become unmanageable.

The Command Line Resurgence

Another notable trend is the resurgence of the command line. After years of being abstracted away in the name of user friendliness, agentic tools are bringing developers back to the terminal as a primary interface. This shift underscores the need for precise control and scriptability — qualities that graphical interfaces often lack.

Security in the Age of Permission-Hungry Agents

Security remains a critical theme, especially given the serious concerns around LLM usage. The radar welcomes Jim Gumbley to the writing team — a security expert who has contributed to this site’s Threat Modeling Guide. His presence underscores the importance of weaving security into the radar’s analysis.

The Dilemma of Broad Access

One of the radar’s key concepts is the “permission hungry” nature of today’s agents. Agents worth building — like OpenClaw, Claude Cowork, or Gas Town — require extensive access to private data, external communication, and real systems. Each application argues that the payoff justifies the risk. But as the radar notes, this appetite for access collides with unsolved problems.

Unsolved Safeguards

The safeguards have not kept pace with ambition. Prompt injection attacks mean models still cannot reliably distinguish trusted instructions from untrusted input. The analogy in the radar is apt: it’s like a skier who has just learned to turn and confidently points themselves at the hardest black run. The ambition is there, but the protective gear is not. This is why the radar places a strong emphasis on harness engineering as a way to build safe guardrails for AI agents.

The Rise of Harness Engineering

With all this in mind, many of the radar’s blips focus on harness engineering — a discipline that Birgitta explores in depth in her excellent article on the subject. The radar includes several entries that suggest the guides and sensors necessary for a well-fitting harness. These range from monitoring tools to policy frameworks that can help teams deploy agents safely without stifling innovation.

Essential Guides and Sensors

For a harness to work, it needs both guides that steer the agent’s behavior and sensors that detect anomalies. The radar’s recommendations cover areas like rate limiting, audit logging, and behavioral constraints. These are not new concepts, but they are being adapted to the unique challenges posed by LLM-driven agents.

Looking Ahead

The ThoughtWorks team expects that when the next radar appears in six months, the list of harness engineering suggestions will grow significantly. As AI agents become more capable, the need for robust, well-designed harnesses will only increase. The current volume serves as a crucial first step in mapping out this emerging territory.

In summary, Technology Radar Volume 34 offers a balanced view: it celebrates AI’s potential while grounding progress in time-tested practices and security awareness. For anyone navigating today’s tech landscape, it provides a valuable compass — pointing both to the horizon and to the tools we already have in hand.

Tags: