GraphWorkflow paper: compile-once engine with proofs, cost model, and open benchmarks
KyeGomezB · x · 2026-09-08
The Swarms team published the full systems paper behind GraphWorkflow (Kye Gomez et al.): a formal execution model with proofs, a cost model for agent orchestration, engine design details, and an open benchmark suite against LangGraph 1.0.4 (5 topologies, 10–200 nodes, 15 configurations).
- Compiled-graph execution: 7.0x geometric-mean speedup, 62.5x on 200-node chains
- 21.6x–31.3x faster compilation; 7.9x faster cold path
- All numbers are medians of 9 samples with 95% CIs; harness, raw data, and reproduction scripts are open
- A companion 14-minute blog post walks through the cost taxonomy, compile-once architecture, and how to reproduce every number
More from coding & agent
- OpenAI Veteran Counters Codex Origin Story: It Was Built as an Internal Infra Tool — gabrielchua · 2026-09-08
- GLM-5.3-Flash builds a 3D kitchen in Blender; author argues Blender-RL is not GPU-expensive — bookwormengr · 2026-09-08
- Factory's /missions pre-build planning wins users over: architecture, milestones, and a HTML-review trick — matanSF · 2026-09-08
- Rauchg launches OSS grants v2: $1,000 each to 35 contributors across agents, local AI, performance — iamsahaj_xyz · 2026-09-08
- MCP tooling explodes on PyPI: mcp-types up 425%, OpenHands doubles 4 months straight — rajistics · 2026-09-08
- Muse Spark 1.3 builds a 16 km² browser 3D driving demo in Three.js — alexandr_wang · 2026-09-08