Introduction
Last week, Thoughtworks unveiled the 34th edition of its Technology Radar, a biannual pulse check on the tools, techniques, platforms, and languages that shape our industry. This volume packs 118 "blips"—concise assessments of each element based on hands-on experience. While AI topics dominate the landscape, the radar also offers a nuanced look at how emerging technologies are forcing us to revisit foundational practices and confront new security challenges.
AI Dominance and the Revisiting of Core Practices
Unsurprisingly, artificial intelligence takes center stage. But what's striking is how AI isn't just pushing us forward—it's also compelling us to look backward. In assembling this edition, the team returned to many established techniques, from pair programming and zero trust architecture to mutation testing and DORA metrics. They also revisited software craftsmanship principles like clean code, deliberate design, testability, and accessibility as first-class concerns.
This is not nostalgia; it's a necessary counterweight to the speed at which AI tools can generate complexity. As AI accelerates development, these core practices act as a stabilizing force, ensuring quality and maintainability. Another interesting trend: the resurgence of the command line. After years of abstracting it away for usability, agentic tools are bringing developers back to the terminal as a primary interface.
Strengthening Security Expertise
I was especially pleased to see my colleague Jim Gumbley join the writing team. He has been a trusted source of security insights, including contributions to our Threat Modeling Guide. Having strong security representation on the radar team is increasingly vital given the serious concerns around using large language models (LLMs).
"Permission Hungry" Agents
One of the radar's key themes is securing what it calls "permission hungry" agents. This phrase captures the central tension of the current agent moment: the agents worth building are precisely those that need extensive access. Tools like OpenClaw and Claude Cowork supervise real‑world tasks; Gas Town coordinates agent swarms across entire codebases. These agents require broad access to private data, external communication, and live systems—each arguing that the payoff justifies the risk.
However, like a skier who has just learned to turn and confidently points toward the hardest black run, the safeguards haven't caught up with that ambition. The appetite for access collides with unsolved problems, such as prompt injection, where models still cannot reliably distinguish trusted instructions from untrusted input.
Harness Engineering Takes Center Stage
Given these challenges, many blips in this radar focus on Harness Engineering. In fact, the radar meeting itself was a major source of ideas for Birgitta's excellent article on the subject. The radar includes several blips suggesting the guides and sensors necessary for a well‑fitting harness—mechanisms to control and monitor AI agents. I expect that when the next radar appears in six months, that list will have grown substantially.
This edition of the Technology Radar doesn't just highlight the latest innovations; it serves as a call to action: embrace AI's potential, but anchor it in time‑tested practices and robust security. The convergence of old and new is where the future of software development will be forged.