Switching from Synchronous to Asynchronous Mode of Coding (blog.joemag.dev)

An engineer reports a deliberate shift from synchronous, hands‑on coding to an asynchronous “agentic” workflow enabled by AI coding agents. After years of doing most development in Rust (where the compiler enforces correctness and slows iterative implementation), they now spend ~5 minutes to break a task into a shared understanding with an agent, then let the agent implement, build, and run unit/integration tests—typically taking 15–30 minutes—while they step away. They review, give targeted feedback, or re-prompt; sometimes they discard results and retry. That change explains increased weekend git activity: shorter focused touchpoints instead of long contiguous blocks of focus. For the AI/ML community this is a concrete example of how agentic tools reshape workflows: AI can take on repetitive compile/test/iterate loops (including “sparring with the Rust compiler”), freeing senior engineers to parallelize higher‑value tasks and recover fragmented attention. Key technical implications are persistent: engineers must learn to craft effective task decompositions and prompts, maintain rigorous human review for correctness and architectural fit, and design CI/tests that let agents iterate safely. The approach suggests new UX and orchestration needs (agent build/test integration, stateful task handoffs, verifiable outputs) and points to productivity gains as agentic systems mature—more detail to be shared at re:Invent.