Managed Research weekly: Tag beta, Factory proof, GameBench pilot

TL;DR

• SMR platform - reliability-focused launch path for terminal run receipts and artifact links.
• Synth Tag beta - delegate one bounded task, steer the same active run, and retrieve the SMR receipt via SDK/MCP.
• Research Factory - proof-tier Factory control plane for multi-run efforts; not the full always-on programme UI.
• GameBench pilot - a five-lane local smoke snapshot is in the launch blog; the 10-lane Grok Build table lands after terminal summaries complete.

Synth Tag Beta

Install:

pip install "synth-ai[research]==0.12.0"

SDK flow:

from synth_ai import SynthClient
 
client = SynthClient()
session = client.research.tag.create_session(
    "Investigate a failing benchmark lane and summarize the smallest fix.",
    definition_of_done="Return a root-cause note with evidence and next action.",
)
receipt = client.research.tag.get_session(session.session_id).receipt

Beta scope: SDK and MCP only. Not Slack, bundles, memory, routines, or Factory auto-link.

Research Factory

Factory is the proof-tier path for durable programmes. The current public proof is an Open Research run with receipt and artifact bundle:

The Factory proof run was retired with the Open Research product.

GameBench Pilot

Current local smoke snapshot:

The launch run uses Grok Build for orchestrator, worker, and reviewer roles across 10 lanes:

These runs are launch evidence, not benchmark certification.

Learn More

• Launch blog
• Synth Tag v1
• GameBench docs
• Managed Research docs