Nearly a year into the agentic coding explosion, many teams now use AI to write software as their de facto approach. As a result, Anthropic’s engineers are shipping eight times as much code as before, and some engineers are shipping thousands of PRs per month.
Management teams expecting their engineering capacity to explode will find that AI-written code doesn’t necessarily compress the software development lifecycle (SDLC) on its own.
The critical bottleneck in software creation has shifted from writing code to a later SDLC stage. After coding agents spit out new code, it still needs testing, and traditionally that meant humans writing the tests. That work includes deciding which user flows need testing, scripting tests for them, and rejiggering them when those flows change.
We’ve made it easy to write code and, in the process, increased the burden on humans to test new code. If we want to slash software development timelines, something has to change.
The observability and monitoring giant Datadog bought Propolis earlier this year, adding agent-written, goal- and output-oriented test writing to its platform. The approach lets agents explore applications, follow user flows on their own, and generate tests from plain-language prompts.
Add Datadog’s observability heft, and you get a system that can test code while providing the logs and traces it needs to vet system performance without the fragility hardcoded testing brings.
You have successfully registered for the webinar.
The New Stack will sit down with Ruxanda Lueck and Matthew White of Datadog on Wednesday, November 4, at 9 a.m. Pacific / 12 p.m. Eastern.
Lueck is a group product manager working on goal-based testing of AI features. White co-founded and served as CTO of Propolis. These are precisely the people to help us understand the current state of the art and what comes next.
We’ll ask how much test-writing teams can offload today and whether agentic testing has reached its own Claude Code moment. We’ll also ask whether plain-language testing instructions expand the number of staffers who can help test new code, and how long it will be before agents can read agent-written code and handle testing without humans lifting a finger.
Agent-written code has created tensions across the SDLC that no single tool can settle on its own. If testing can speed up code writing, though, teams clear one more obstacle between an idea and working software in production.
You’ll learn:
- How agents can test user flows that no one thought of
- How to write plain-language testing prompts
- How goal-based testing persists even when user flows change
- How to test AI features that return probabilistic, instead of deterministic, outputs
- Where scripted tests still have their place
The post Scripted tests can’t keep up with AI-written code. Datadog is betting agents can. appeared first on The New Stack.