Stas Bekman: top coding agents all fail at concurrent multi-tasking — orchestrator plus sub-agents is the fix
StasBekman · x · 2026-10-08
ML engineer Stas Bekman spent weeks on a task requiring testing and fixing dozens of configs, each with dozens of slow tests. Every top coding agent he tried failed at concurrent multi-tasking — dropping the ball every 20 minutes and wasting resources; planning mode didn't help.
His working recipe: instruct the main agent to do no work itself and act purely as an orchestrator, delegating all actual tasks to sub-agents. Only then did the work move forward. A directly actionable pattern for anyone running long, parallel agent workloads.
More from coding & agent
- Handwritten text, AI-built interactions: a new golden age of technical explainers — srush_nlp · 2026-10-08
- Coinbase Institute paper: stablecoins fit agent-to-agent micropayments in AiFi era — MurrLincoln · 2026-10-08
- Depth2Depth fuses DepthAnything with RealSense for dense, metric depth maps — chrismatthieu · 2026-10-08
- 12 things engineers miss since coding died: faster models, busier teams, lost satisfaction — rseroter · 2026-10-08
- Microsoft's MAI-Code-1.1-Flash goes local in GitHub Copilot with free on-device calls — flngr · 2026-10-08
- Game dev pushes back on one-shot AI hype: real winners still iterate for months — IanArawjo · 2026-10-08