A source-level study of three coding-agent harnesses finds five converged elements and one that nobody has built
The authors read LangChain's deepagents, Earendil's pi and DeepSeek's dsh at pinned commits and followed their commit histories. deepagents has been subtracting authored scaffolding while pi has been accreting durable infrastructure, yet both arrived at the same five parts: a commoditised loop, an append-only replayable session record, model quirks stored as data, progressive disclosure of context, and explicit extension seams. A held-out third harness shows all five and in one seam reuses another's implementation verbatim, so the paper splits the convergence into parallel discovery, diffusion and literal reuse rather than claiming independent invention. The gap they name is external verifiability: no harness has a tamper-evident record an outside party can check without trusting the runtime.
↳ Follow the thread