Synth Blog
Product launches, release notes, workflows, and the engineering behind Synth.
An open Rust workflow substrate for research engineers: fan out model work into seeded sandboxes, capture artifacts, and replay the orchestration path without pretending future model calls are deterministic.
Managed Research is live. ReportBench scores vs Codex standalone, how the message queue works, and how to get started via MCP, SynthClient().research, or the frontend.
Proof-gated product updates, releases, and fixes.
Research Factory turns a bounded objective and budget into research cycles, typed maintenance decisions, immutable candidates, and benchmark-owned grading.
Jul 13, 2026
Monitors, gardener agents, sidekick guidance, and experimental handoffs for long-running coding agents.
Jun 30, 2026
A scoped launch update on evaluating Managed Research through GameBench: Harbor-native game-environment tasks, code-policy work, policy puzzles, cybernetic control, and run receipts.
This is historical material. Use the current Managed Research, Research Factory, and GEPA/GELO pages for product decisions.
Jun 26, 2026
Synth's Open Research Factory now has production proof: public research work can be published with a receipt, bundle, and private-work firewall.
This is historical material. Use the current Managed Research, Research Factory, and GEPA/GELO pages for product decisions.
Jun 18, 2026
Attach a repo, run launch preflight, trigger a directed run, and validate submission/optimizer.py with Nanoprogram's official harness.
This is historical material. Use the current Managed Research, Research Factory, and GEPA/GELO pages for product decisions.
Apr 6, 2026