DEVOURED
Verifying Agentic Development at Scale
Cognition's Ido Pesok revealed that Devin now handles more asynchronous, verified-before-merge testing sessions than interactive ones, thanks to running 10-20 Devins in parallel.
What: Ido Pesok from Cognition highlighted that Devin, their AI agent, is increasingly used for autonomous end-to-end testing, with asynchronous sessions now outnumbering interactive ones. This shift was enabled by engineers running multiple Devin instances (10-20) in parallel with dedicated dev servers, a capability added roughly six months ago.
Why it matters: The move towards massively parallel, asynchronous execution for AI agents like Devin demonstrates how scaling infrastructure can unlock entirely new modes of operation for agentic development, making "verified-before-merge" a practical reality rather than an aspirational goal.
Takeaway: If building or integrating AI agents for development tasks, consider how parallel execution and dedicated ephemeral environments could drastically improve throughput and enable automated verification workflows.
Decoder
- Devin: An AI software engineer, developed by Cognition, designed to autonomously plan and execute complex engineering tasks.
Original article
Cognition's Ido Pesok shares lessons from building autonomous end-to-end testing into Devin, noting that for the first time, more Devin sessions are now triggered asynchronously than interactively, making verified-before-merge results a hard requirement rather than a nicety. Devin's harness gained computer-use tools roughly six months ago, and the breakthrough came when engineers started running 10-20 Devins in parallel, each with its own dev server, something impossible on a single laptop.
Explore all 68 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi