{"event":{"id":"evt-saycan-20220404","dedupe_key":"embodied_language_planning_research:2022-04-04:saycan-connects-language-model-planning-to-executable-robot-skills","event_type":"embodied_language_planning_research","title":"SayCan connects language-model planning to executable robot skills","summary":"SayCan combined language-model high-level action proposals with learned affordance/value functions and demonstrated long-horizon natural-language tasks on a real mobile manipulator.","occurred_at":"2022-04-04T17:57:11.000Z","published_at":"2022-04-04T17:57:11.000Z","observed_at":"2026-10-01T18:14:00.000Z","reconstructed_at":"2026-10-01T18:14:00.000Z","ingest_type":"live_world_scan","locations":null,"status":"verified","confidence":0.97,"metadata":{"actors":["Google Research","Everyday Robots"],"what_happened_at_time":"A language model was used as a high-level planner while pretrained robotic skills and value functions constrained actions to those feasible in the physical environment.","evidence_at_time":"The April 4 preprint documented real-robot evaluations on temporally extended natural-language instructions.","affected_layers":["layer-models","layer-research","layer-robotics","layer-agents","layer-software"],"change_kind":"precursor_maturation","q1_continuity":"Q1 Agents remained a precursor. SayCan moved beyond prompting-only model behavior into model-mediated selection of executable actions, but still through a fixed skill library.","retrospective_inference":"This is strong evidence of movement toward the later Agents layer, especially models->action/planning coupling, without establishing a persistent general-purpose agent.","uncertainty":"It was a research prototype with predefined low-level skills, no durable personal state, no broad external tool ecosystem and no evidence of persistent autonomous operation."},"created_at":"2026-10-01T19:22:17.746Z","updated_at":"2026-10-01T19:22:17.746Z"},"artifacts":[{"id":"art-saycan-20220404","canonical_url":"https://arxiv.org/abs/2204.01691","artifact_type":"paper_preprint","title":"Do As I Can, Not As I Say: Grounding Language in Robotic Affordances","summary":"The SayCan paper combined language-model high-level task proposals with learned robotic skill affordances and demonstrated completion of long-horizon natural-language instructions on a real mobile manipulator.","creator_entities":["Google Research","Everyday Robots"],"released_at":"2022-04-04T17:57:11.000Z","content_hash":null,"metadata":{"historical_scope":"2022-Q2"},"created_at":"2026-10-01T19:22:17.740Z","updated_at":"2026-10-01T19:22:17.740Z","relation":"documented_by"}],"signals":[{"id":"sig-q2-agentic-embodied-precursor","statement":"Language/generalist transformer models began to select and execute actions in embodied or interactive environments, materially strengthening an agent-like precursor without yet establishing persistent autonomous agents.","signal_type":"agent_precursor_maturation","direction":"increasing","confidence":0.94,"epistemic_status":"supported_inference","mapping_mode":"retrospective","reconstructed_at":"2026-10-01T18:14:00.000Z","created_at":"2026-10-01T19:22:17.753Z","updated_at":"2026-10-01T19:22:17.753Z"}]}