ThoughtWorks Technology Radar v32

I remain a big fan of the Thoughtworks TechRadar, and v32 (pdf) was just released.

Below are some of the things that stood out for me. Mostly copy/pasted/reworded with a few comments.

Themes

Supervised agents in coding assistants

More and more tool accelerate the capabilities of coding assistants by allowing developers to drive implementation directly from an AI chat within their IDE. A mode also called “agentic”, “prompt-to-code” or “chat-oriented programming (CHOP).” In this approach, AI assistants go beyond answering questions or generating small snippets; they navigate and modify code, and update tests, fix issues etc. Developers still guide and oversee the agent’s actions. Cursor, Cline and Windsurf are leading this trend in the IDE-integrated tools space, with GitHub Copilot also advancing. Agentic assistants such as aider, goose and Claude Code are terminal-based alternatives. But use caution: there is still a need for steering and vigilance in the code review. With great power…

Evolving observability

One emerging focus is LLM observability, a critical piece in operationalizing AI. We’ve seen a surge in tools for monitoring and evaluating LLM performance, including Weights & Biases Weave, Arize Phoenix, Helicone and HumanLoop.

Another trend is the integration of AI-assisted observability, where tools leverage AI to enhance analysis and insights. (But I wish they shared more details here!)

Additionally, the increasing adoption of OpenTelemetry is fostering a more standardized observability landscape, enabling teams to remain vendor-agnostic and more flexible
in their tooling choices. Many leading observability tools — such as Alloy, Tempo and Loki — now support OpenTelemetry.

Techniques

Data product thinking

Data product thinking is an approach that treats data as a product with its own lifecycle, quality standards and a focus on meeting consumer needs. The data as a product mindset is one of the four principles of data mesh. We recommend it regardless of whether organizations choose architectures like data mesh or lakehouse. Design data products by working backward from use cases. Consider capturing and managing both business and technical metadata using data catalogs like DataHub, Collibra, Atlan and Informatica to improve data discoverability and usability.

API request collection as API product artifact

Treating APIs as a product means prioritizing developer experience. While OpenAPI (Swagger) specifications can effectively document API interfaces, onboarding remains a challenge. Developers need rapid access to working examples, with preconfigured authentication and realistic test data. With the maturing of API client tools (such as
Postman, Bruno and Insomnia), we recommend treating API request collection as API product artifact.

API request collections should be thoughtfully designed to guide developers through key workflows, helping them to understand the API’s domain language and functionality with minimal effort. To keep collections up to date, we recommend storing them in a repository and integrating them into the API’s release pipeline.

Architecture advice process

One of the persistent challenges in large software teams is determining who makes the architectural decisions that shape the evolution of systems. The State of DevOps report reveals that the traditional approach of Architecture Review Boards is counterproductive, often hindering workflow and correlating with low organizational performance. A compelling alternative is an architectural advice process — a decentralized approach where anyone can make any architectural decision, provided they seek advice from those affected and those with relevant expertise.

This method enables teams to optimize for flow without compromising architectural quality, both at small and large scales. At first glance, this approach may seem controversial, but practices such as Architecture Decision Records and advisory forums ensure that decisions remain informed, while empowering those closest to the work to make decisions.

Model distillation

Model distillation transfers knowledge from a larger, more powerful model (teacher) to a smaller, cost-efficient model (student), to allow it to be used on smallwe consumer hardware. A related topic is SLMs…

Small language models

Small language models (SLMs) continue to be interesting. These smaller models can be run on much more modest hardware.

Using GenAI to understand legacy codebases

GenAI can be a useful tool to understand legacy codebases. While tools such as GitHub Copilot can be used to modernize codebases, tools such as Sourcegraph’s Cody are making it easier for developers to understand legacy codebases. These tools use a multitude of GenAI techniques to provide contextual help, simplifying work with complex legacy systems.

Tools

Renovate

Renovate has become the tool of choice for dependency version management. While Dependabot remains a safe default choice for GitHub-
hosted repositories, Renovate is a more comprehensive and customizable solution.  To reduce developer fatigue, consider automatic merging of dependency update PRs.

Vite

Vite is a high-performance front-end build tool with fast hot-reloading. It’s becoming a default choice in many front-end frameworks, including Vue, SvelteKit and React.

Claude Sonnet

Claude Sonnet is an advanced language model for coding, writing, analysis and visual processing. It’s available in the browser, terminal, most major IDEs and even integrates with GitHub Copilot.  It excels in greenfield projects, particularly in collaborative software design and architectural discussions. Claude Sonnet is among the most reliable coding assistance model we’ve worked with.

At time of writing, 3.7 is the latest model.

Cursor

Cursor is an AI-first code editor, a leader in the competitive AI coding assistance space. Its code context orchestration is very effective, and it supports a wide range of models, including the option to use a custom API key. The Cursor team often comes up with innovative user experience features before the other vendors, and they include an extensive list of context providers in their chat, such as the referencing of git diffs, previous AI conversations, web search, library documentation and MCP integration. Alongside tools like Cline and Windsurf, Cursor also stands out for its strong agentic coding mode. This mode allows developers to guide their implementation directly from an AI chat interface, with the tool autonomously reading and modifying files, as well as executing commands. Additionally, we appreciate Cursor’s ability to detect linting and compilation errors in generated code and proactively correct them.

D2

D2 is an open-source diagrams-as-code tool that helps users create and customize diagrams from text. It introduces the D2 diagram scripting language, which prioritizes readability over compactness with a simple, declarative syntax. It is designed for software documentation and architecture diagrams.

Software engineering agents

There is still no agreed definition of the term “agent.” However, a major development has emerged in supervised agentic modes within the IDE. These modes allow developers to drive implementation via chat, with tools not only modifying code in multiple files but also executing commands, running tests and responding to IDE feedback like
linting or compile errors.

This approach, sometimes called chat-oriented programming (CHOP) or prompt-to-code, keeps developers in control while shifting more responsibility to AI than traditional coding assistants like auto-suggestions. Leading tools in this space include Cursor, Cline and Windsurf, with GitHub Copilot slightly behind but catching up. Claude’s Sonnet is the state of the art model to use with these agentic modes

They can provide a notable increase in coding speed, but keeping problem scopes small helps developers better review AI-generated changes.

Languages and Frameworks

OpenTelemetry

OpenTelemetry is the industry standard for observability. The release of the OpenTelemetry protocol (OTLP) specification established a standardized way to handle traces,
metrics and logs, reducing the need for multiple integrations or major rewrites as monitoring distributed solutions and interoperability requirements grow. As OpenTelemetry expands to support logs and profiling, OTLP ensures a consistent transport format across all telemetry data, simplifying instrumentation.

Adopted by vendors like Datadog, New Relic and Grafana, OTLP enables organizations to build flexible, vendor-agnostic observability stacks without being locked into proprietary solutions.

Tags: radar, techradar, thoughtworks, thoughtworkstechradar