{"event":{"id":"evt-openai-instructgpt-20220127","dedupe_key":"model_deployment_shift:2022-01-27:openai-makes-instructgpt-style-models-the-default-api-language-models","event_type":"model_deployment_shift","title":"OpenAI makes InstructGPT-style models the default API language models","summary":"OpenAI said models fine-tuned with human demonstrations and preference rankings were now the default models on its API, operationalizing instruction-following post-training in a production service.","occurred_at":"2022-01-27T00:00:00.000Z","published_at":"2022-01-27T00:00:00.000Z","observed_at":"2026-10-01T17:04:00.000Z","reconstructed_at":"2026-10-01T17:34:32.000Z","ingest_type":"live_world_scan","locations":null,"status":"verified","confidence":0.99,"metadata":{"actors":["OpenAI"],"what_happened_at_time":"RLHF-tuned instruction-following models became OpenAI API defaults.","evidence_at_time":"OpenAI's Jan 27 publication documented deployment and method; a fuller paper followed in Q1.","retrospective_inference":"The model stack was shifting from pretraining alone toward post-training shaped by human feedback and user intent.","uncertainty":"This was maturation/operationalization, not the invention of RLHF; beta versions had existed earlier."},"created_at":"2026-10-01T18:05:36.364Z","updated_at":"2026-10-01T18:05:36.364Z"},"artifacts":[{"id":"art-openai-instructgpt-20220127","canonical_url":"https://openai.com/index/instruction-following/","artifact_type":"research_product_release","title":"InstructGPT default API deployment","summary":"OpenAI reported that instruction-following models trained with reinforcement learning from human feedback were deployed as the default language models on its API.","creator_entities":["OpenAI"],"released_at":"2022-01-27T00:00:00.000Z","content_hash":null,"metadata":{"historical_scope":"2022-Q1"},"created_at":"2026-10-01T18:05:36.358Z","updated_at":"2026-10-01T18:05:36.358Z","relation":"documented_by"}],"signals":[{"id":"sig-instruction-following-production-2022q1","statement":"Human-feedback post-training moved from research technique into default production model behavior in a major commercial API.","signal_type":"deployment_shift","direction":"increasing","confidence":0.98,"epistemic_status":"supported_inference","mapping_mode":"retrospective","reconstructed_at":"2026-10-01T17:34:32.000Z","created_at":"2026-10-01T18:05:36.373Z","updated_at":"2026-10-01T18:05:36.373Z"},{"id":"sig-preagent-stack-coupling-2022q1","statement":"By the end of Q1 2022, the clearest stack coupling ran through research, models, software, compute, chips, networks, manufacturing and governance; the selected evidence does not yet support persistent agents or agent permissioning as independent operational layers.","signal_type":"civilization_stack_coupling","direction":"increasing","confidence":0.86,"epistemic_status":"supported_inference","mapping_mode":"retrospective","reconstructed_at":"2026-10-01T17:34:32.000Z","created_at":"2026-10-01T18:05:36.382Z","updated_at":"2026-10-01T18:05:36.382Z"}]}