{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreiczvnrgcmek5bm7zekhkmabs5jfquclz7aznyvzonjykpmxkkdqny",
"uri": "at://did:plc:2gbt2dlwaqovtnmxkat3tyke/app.bsky.feed.post/3mklyki54zif2"
},
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreigkftinjphtmhja7pvej56um4bs7cgj6rzwiyjsqtne4rakiw63fu"
},
"mimeType": "image/jpeg",
"size": 84849
},
"path": "/articles/d41586-026-00820-5",
"publishedAt": "2026-04-29T01:54:29.850Z",
"site": "https://www.nature.com",
"tags": [
"doi:10.1038/d41586-026-00820-5"
],
"textContent": "Nature, Published online: 28 April 2026; doi:10.1038/d41586-026-00820-5\n\nTraining AI world models on data about physical environments could improve their real-world capabilities in technologies such as robotics.",
"title": "‘World models’ are AI’s latest sensation: what are they and what can they do?",
"updatedAt": "2026-04-28T00:00:00.000Z"
}