DeepLearning.AI posts new course

DeepLearning.AI has launched a new short course on spec-driven development, a method that has developers write detailed plans before handing work to an artificial intelligence coding agent. (learn.deeplearning.ai) The course is called “Spec-Driven Development with Coding Agents,” is taught by JetBrains developer advocate Paul Everitt, and lists 15 video lessons on DeepLearning.AI’s learning platform. It says the class was built in partnership with JetBrains. (learn.deeplearning.ai) DeepLearning.AI’s course page says students learn to write a “project constitution” and feature specifications, then run a plan-implement-verify loop for both new projects and legacy codebases. The lesson list includes modules on feature validation, project replanning, building a minimum viable product, and “agent replaceability.” (learn.deeplearning.ai, learn.deeplearning.ai) Spec-driven development treats a written specification like a blueprint before construction starts, instead of relying on back-and-forth prompting after code is already being generated. DeepLearning.AI describes it as an alternative to “vibe coding,” which its page says is fast but often produces code that does not match the original request. (learn.deeplearning.ai) The hands-on project is not just a toy example. DeepLearning.AI’s community forum opened a seven-day learner challenge on April 14, 2026 asking students to build a “Tiny Tamagotchi MVP” tied to the new course. (community.deeplearning.ai) The companion GitHub repository shows how the course is structured in practice. It includes lesson folders, example specifications, reusable “skills,” and an “AgentClinic” project that moves from an empty scaffold to specs, implementation, and validation files across successive lessons. (github.com) JetBrains has been pushing the same idea across its artificial intelligence tooling. Its recent materials describe spec-driven workflows as a way to turn requirements, tickets, and notes into implementation plans, generated code, tests, and validation inside JetBrains environments. (blog.jetbrains.com, info.jetbrains.com) That pitch lands at a moment when coding agents are moving into mainstream developer workflows. JetBrains said in March 2026 that 22 percent of developers in its January 2026 AI Pulse survey already use artificial intelligence coding agents, and 66 percent of surveyed companies planned to adopt them within 12 months. (blog.jetbrains.com) DeepLearning.AI, which says more than 7 million people use its courses, has recently added more short classes aimed at working developers rather than only model researchers. This new course fits that pattern by focusing on how to keep agent-generated software aligned with requirements, tests, and maintenance work after the first draft of code. (deeplearning.ai, learn.deeplearning.ai)