{"id":"456c51f07dd9d996fff5a03b95bfd7459f53e54da470e9d74650ca27195892b3","title":"MIT Technology Review reports Google is changing Gemini’s responses and adding an agent","argument":"On November 18, 2025, MIT Technology Review reported that Google unveiled Gemini 3, which Google says improves reasoning and multimodal capabilities. A multimodal model works across kinds of input or output, such as text, images and audio. The article’s central shift is from returning plain text by default to “generative interfaces”: Gemini 3 can choose a visual layout, diagram or animation for a response. For example, a travel request may produce an interactive, website-like view with follow-up prompts, rather than a block of text.\n\nGoogle also introduced Gemini Agent, an experimental feature that can connect to services including Calendar and Gmail to handle multi-step tasks. It displays progress and pauses for user approval before continuing. Google described it as a step toward “a true generalist agent”; the article says it would be available on the web to Google AI Ultra subscribers in the US starting November 18. Google also announced changes to Search, shopping recommendations and a developer platform for creating and managing code, tools and workflows from a prompt.\n\nFlowith CEO Derek Nee said Gemini 3 Pro addresses gaps in earlier models and that his company was integrating it, while stressing that deeper testing was needed to understand its limits. The article does not report results from such testing, so how well these capabilities work in practice remains unknown.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on November 3, 2026; originally published November 18, 2025. The article reports Google’s descriptions and Nee’s assessment, but does not provide results from deeper testing of the capabilities described.","speaker":{"name":"MIT Technology Review","roleAtSource":"Publisher"},"source":{"title":"Google’s new Gemini 3 “vibe-codes” responses and comes with its own agent","publisher":"MIT Technology Review","format":"Article","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","publishedAt":"2025-11-18T16:00:07Z","dateBasis":"Date from the publisher’s own records, through its API."},"evidence":[{"text":"MIT Technology Review reports that Google unveiled Gemini 3, which Google says improves reasoning and multimodal capabilities.","excerpt":"Google today unveiled Gemini 3, a major upgrade to its flagship multimodal model. The firm says the new model is better at reasoning","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","relationship":"Source-stated"},{"text":"The article says Gemini 3 introduces “generative interfaces,” allowing the model to choose an output format and assemble visual layouts or dynamic views.","excerpt":"Gemini 3 introduces what Google calls “generative interfaces,” which allow the model to make its own choices about what kind of output fits the prompt best","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","relationship":"Source-stated"},{"text":"Google introduced Gemini Agent as an experimental feature for multi-step tasks; it can connect to services and pauses for user approval before continuing.","excerpt":"With Gemini 3, Google is also introducing Gemini Agent, an experimental feature designed to handle multi-step tasks directly inside the app.","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","relationship":"Source-stated"},{"text":"Google described Gemini Agent as a step toward “a true generalist agent,” and the article said it would be available on the web to Google AI Ultra subscribers in the US starting November 18.","excerpt":"Google describes the feature as a step toward “a true generalist agent.” It will be available on the web for Google AI Ultra subscribers in the US starting November 18.","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","relationship":"Source-stated"},{"text":"Derek Nee said Flowith was integrating Gemini 3 Pro but needed deeper testing to understand how far it could go.","excerpt":"“We’re optimistic about its potential, but we need deeper testing to understand how far it can go.”","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","relationship":"Source-stated"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How will Gemini 3’s visual responses and multi-step tasks perform in deeper testing, and what limits will that testing find?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"MIT Technology Review reports Google is changing Gemini’s responses and adding an agent","summary":"On November 18, 2025, MIT Technology Review reported that Google unveiled Gemini 3, which Google says improves reasoning and multimodal capabilities. A multimodal model works across kinds of input or output, such as text, images and audio. The article’s central shift is from returning plain text by default to “generative interfaces”: Gemini 3 can choose a visual layout, diagram or animation for a response. For example, a travel request may produce an interactive, website-like view with follow-up prompts, rather than a block of text.\n\nGoogle also introduced Gemini Agent, an experimental feature that can connect to services including Calendar and Gmail to handle multi-step tasks. It displays progress and pauses for user approval before continuing. Google described it as a step toward “a true generalist agent”; the article says it would be available on the web to Google AI Ultra subscribers in the US starting November 18. Google also announced changes to Search, shopping recommendations and a developer platform for creating and managing code, tools and workflows from a prompt.\n\nFlowith CEO Derek Nee said Gemini 3 Pro addresses gaps in earlier models and that his company was integrating it, while stressing that deeper testing was needed to understand its limits. The article does not report results from such testing, so how well these capabilities work in practice remains unknown.","question":"How will Gemini 3’s visual responses and multi-step tasks perform in deeper testing, and what limits will that testing find?","claims":[{"text":"MIT Technology Review reports that Google unveiled Gemini 3, which Google says improves reasoning and multimodal capabilities.","excerpt":"Google today unveiled Gemini 3, a major upgrade to its flagship multimodal model. The firm says the new model is better at reasoning","support":"source-stated"},{"text":"The article says Gemini 3 introduces “generative interfaces,” allowing the model to choose an output format and assemble visual layouts or dynamic views.","excerpt":"Gemini 3 introduces what Google calls “generative interfaces,” which allow the model to make its own choices about what kind of output fits the prompt best","support":"source-stated"},{"text":"Google introduced Gemini Agent as an experimental feature for multi-step tasks; it can connect to services and pauses for user approval before continuing.","excerpt":"With Gemini 3, Google is also introducing Gemini Agent, an experimental feature designed to handle multi-step tasks directly inside the app.","support":"source-stated"},{"text":"Google described Gemini Agent as a step toward “a true generalist agent,” and the article said it would be available on the web to Google AI Ultra subscribers in the US starting November 18.","excerpt":"Google describes the feature as a step toward “a true generalist agent.” It will be available on the web for Google AI Ultra subscribers in the US starting November 18.","support":"source-stated"},{"text":"Derek Nee said Flowith was integrating Gemini 3 Pro but needed deeper testing to understand how far it could go.","excerpt":"“We’re optimistic about its potential, but we need deeper testing to understand how far it can go.”","support":"source-stated"}],"limitations":["Text as retrieved on November 3, 2026; originally published November 18, 2025.","The article reports Google’s descriptions and Nee’s assessment, but does not provide results from deeper testing of the capabilities described."],"mediaId":null,"correctionReason":null,"id":"456c51f07dd9d996fff5a03b95bfd7459f53e54da470e9d74650ca27195892b3","publicationId":"8b151f85-4396-717e-4fdf-fc9d1b42cd94","version":1,"sourceUrl":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","sourceTitle":"Google’s new Gemini 3 “vibe-codes” responses and comes with its own agent","publisher":"MIT Technology Review","publishedAt":"2025-11-18T16:00:07Z","originalPublished":{"value":"2025-11-18T16:00:07Z","precision":"instant","basis":"backfill:api_published"},"firstObservedAt":"2026-11-03T03:57:50.543Z","createdAt":"2026-11-09T23:53:23.852Z","historical":true,"entityIds":[],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"bd98f194656c566cd5dc6a63b423a4fbbd987a54cbbd3b9e72a022540baf7ae2","rubricVersion":"unrecorded","provider":"anthropic","model":"claude-opus-5-5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-11-03T03:57:48.431Z"},"provenance":{"textHash":"d203e6d2f544e091c7bfb4f28a1260894664385c867d688982ae2680a51473fe","observationId":"8140f95da06a555e5224202476d559250189eeb37138ff813d82a141c2e28123","retrievedAt":"2026-11-03T03:59:23.241Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"6985591f-d3db-4578-9930-ba728d072a63"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"7abb2813-2399-f63b-f79e-75593c3a8b74"}},"audit":{"status":"passed","reviewedAt":"2026-11-09T23:48:44.210Z","model":"anthropic/claude-opus-5.5"}},"canonicalUrl":"https://escapemeridian.com/readings/mit-technology-review-reports-google-is-changing-gemini-s-456c51f0","formats":{"html":"/readings/mit-technology-review-reports-google-is-changing-gemini-s-456c51f0","markdown":"/readings/mit-technology-review-reports-google-is-changing-gemini-s-456c51f0.md","json":"/readings/mit-technology-review-reports-google-is-changing-gemini-s-456c51f0.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-multimodal-model","name":"Multimodal model","explanation":"A model that works with more than one kind of input or output, such as text, images and audio.","referenceUrl":"https://developers.google.com/machine-learning/glossary#multimodal-model"}],"entities":[]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"456c51f07dd9d996fff5a03b95bfd7459f53e54da470e9d74650ca27195892b3","revision":"8b151f85-4396-717e-4fdf-fc9d1b42cd94","revisionKind":"publication-id","title":"MIT Technology Review reports Google is changing Gemini’s responses and adding an agent","url":"https://escapemeridian.com/readings/mit-technology-review-reports-google-is-changing-gemini-s-456c51f0"},"source":{"type":"source","id":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","revision":"d203e6d2f544e091c7bfb4f28a1260894664385c867d688982ae2680a51473fe","revisionKind":"sha256-source-text","title":"Google’s new Gemini 3 “vibe-codes” responses and comes with its own agent","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter."},"revisions":[{"id":"456c51f07dd9d996fff5a03b95bfd7459f53e54da470e9d74650ca27195892b3","title":"MIT Technology Review reports Google is changing Gemini’s responses and adding an agent","argument":"On November 18, 2025, MIT Technology Review reported that Google unveiled Gemini 3, which Google says improves reasoning and multimodal capabilities. A multimodal model works across kinds of input or output, such as text, images and audio. The article’s central shift is from returning plain text by default to “generative interfaces”: Gemini 3 can choose a visual layout, diagram or animation for a response. For example, a travel request may produce an interactive, website-like view with follow-up prompts, rather than a block of text.\n\nGoogle also introduced Gemini Agent, an experimental feature that can connect to services including Calendar and Gmail to handle multi-step tasks. It displays progress and pauses for user approval before continuing. Google described it as a step toward “a true generalist agent”; the article says it would be available on the web to Google AI Ultra subscribers in the US starting November 18. Google also announced changes to Search, shopping recommendations and a developer platform for creating and managing code, tools and workflows from a prompt.\n\nFlowith CEO Derek Nee said Gemini 3 Pro addresses gaps in earlier models and that his company was integrating it, while stressing that deeper testing was needed to understand its limits. The article does not report results from such testing, so how well these capabilities work in practice remains unknown.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on November 3, 2026; originally published November 18, 2025. The article reports Google’s descriptions and Nee’s assessment, but does not provide results from deeper testing of the capabilities described.","speaker":{"name":"MIT Technology Review","roleAtSource":"Publisher"},"source":{"title":"Google’s new Gemini 3 “vibe-codes” responses and comes with its own agent","publisher":"MIT Technology Review","format":"Article","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","publishedAt":"2025-11-18T16:00:07Z","dateBasis":"Date from the publisher’s own records, through its API."},"evidence":[{"text":"MIT Technology Review reports that Google unveiled Gemini 3, which Google says improves reasoning and multimodal capabilities.","excerpt":"Google today unveiled Gemini 3, a major upgrade to its flagship multimodal model. The firm says the new model is better at reasoning","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","relationship":"Source-stated"},{"text":"The article says Gemini 3 introduces “generative interfaces,” allowing the model to choose an output format and assemble visual layouts or dynamic views.","excerpt":"Gemini 3 introduces what Google calls “generative interfaces,” which allow the model to make its own choices about what kind of output fits the prompt best","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","relationship":"Source-stated"},{"text":"Google introduced Gemini Agent as an experimental feature for multi-step tasks; it can connect to services and pauses for user approval before continuing.","excerpt":"With Gemini 3, Google is also introducing Gemini Agent, an experimental feature designed to handle multi-step tasks directly inside the app.","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","relationship":"Source-stated"},{"text":"Google described Gemini Agent as a step toward “a true generalist agent,” and the article said it would be available on the web to Google AI Ultra subscribers in the US starting November 18.","excerpt":"Google describes the feature as a step toward “a true generalist agent.” It will be available on the web for Google AI Ultra subscribers in the US starting November 18.","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","relationship":"Source-stated"},{"text":"Derek Nee said Flowith was integrating Gemini 3 Pro but needed deeper testing to understand how far it could go.","excerpt":"“We’re optimistic about its potential, but we need deeper testing to understand how far it can go.”","url":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","relationship":"Source-stated"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How will Gemini 3’s visual responses and multi-step tasks perform in deeper testing, and what limits will that testing find?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"MIT Technology Review reports Google is changing Gemini’s responses and adding an agent","summary":"On November 18, 2025, MIT Technology Review reported that Google unveiled Gemini 3, which Google says improves reasoning and multimodal capabilities. A multimodal model works across kinds of input or output, such as text, images and audio. The article’s central shift is from returning plain text by default to “generative interfaces”: Gemini 3 can choose a visual layout, diagram or animation for a response. For example, a travel request may produce an interactive, website-like view with follow-up prompts, rather than a block of text.\n\nGoogle also introduced Gemini Agent, an experimental feature that can connect to services including Calendar and Gmail to handle multi-step tasks. It displays progress and pauses for user approval before continuing. Google described it as a step toward “a true generalist agent”; the article says it would be available on the web to Google AI Ultra subscribers in the US starting November 18. Google also announced changes to Search, shopping recommendations and a developer platform for creating and managing code, tools and workflows from a prompt.\n\nFlowith CEO Derek Nee said Gemini 3 Pro addresses gaps in earlier models and that his company was integrating it, while stressing that deeper testing was needed to understand its limits. The article does not report results from such testing, so how well these capabilities work in practice remains unknown.","question":"How will Gemini 3’s visual responses and multi-step tasks perform in deeper testing, and what limits will that testing find?","claims":[{"text":"MIT Technology Review reports that Google unveiled Gemini 3, which Google says improves reasoning and multimodal capabilities.","excerpt":"Google today unveiled Gemini 3, a major upgrade to its flagship multimodal model. The firm says the new model is better at reasoning","support":"source-stated"},{"text":"The article says Gemini 3 introduces “generative interfaces,” allowing the model to choose an output format and assemble visual layouts or dynamic views.","excerpt":"Gemini 3 introduces what Google calls “generative interfaces,” which allow the model to make its own choices about what kind of output fits the prompt best","support":"source-stated"},{"text":"Google introduced Gemini Agent as an experimental feature for multi-step tasks; it can connect to services and pauses for user approval before continuing.","excerpt":"With Gemini 3, Google is also introducing Gemini Agent, an experimental feature designed to handle multi-step tasks directly inside the app.","support":"source-stated"},{"text":"Google described Gemini Agent as a step toward “a true generalist agent,” and the article said it would be available on the web to Google AI Ultra subscribers in the US starting November 18.","excerpt":"Google describes the feature as a step toward “a true generalist agent.” It will be available on the web for Google AI Ultra subscribers in the US starting November 18.","support":"source-stated"},{"text":"Derek Nee said Flowith was integrating Gemini 3 Pro but needed deeper testing to understand how far it could go.","excerpt":"“We’re optimistic about its potential, but we need deeper testing to understand how far it can go.”","support":"source-stated"}],"limitations":["Text as retrieved on November 3, 2026; originally published November 18, 2025.","The article reports Google’s descriptions and Nee’s assessment, but does not provide results from deeper testing of the capabilities described."],"mediaId":null,"correctionReason":null,"id":"456c51f07dd9d996fff5a03b95bfd7459f53e54da470e9d74650ca27195892b3","publicationId":"8b151f85-4396-717e-4fdf-fc9d1b42cd94","version":1,"sourceUrl":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","sourceTitle":"Google’s new Gemini 3 “vibe-codes” responses and comes with its own agent","publisher":"MIT Technology Review","publishedAt":"2025-11-18T16:00:07Z","originalPublished":{"value":"2025-11-18T16:00:07Z","precision":"instant","basis":"backfill:api_published"},"firstObservedAt":"2026-11-03T03:57:50.543Z","createdAt":"2026-11-09T23:53:23.852Z","historical":true,"entityIds":[],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2025/11/18/1128065/googles-gemini-3/","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"bd98f194656c566cd5dc6a63b423a4fbbd987a54cbbd3b9e72a022540baf7ae2","rubricVersion":"unrecorded","provider":"anthropic","model":"claude-opus-5-5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-11-03T03:57:48.431Z"},"provenance":{"textHash":"d203e6d2f544e091c7bfb4f28a1260894664385c867d688982ae2680a51473fe","observationId":"8140f95da06a555e5224202476d559250189eeb37138ff813d82a141c2e28123","retrievedAt":"2026-11-03T03:59:23.241Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"6985591f-d3db-4578-9930-ba728d072a63"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"7abb2813-2399-f63b-f79e-75593c3a8b74"}},"audit":{"status":"passed","reviewedAt":"2026-11-09T23:48:44.210Z","model":"anthropic/claude-opus-5.5"}}}],"audits":[{"publicationId":"8b151f85-4396-717e-4fdf-fc9d1b42cd94","verdict":"passed","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the report to MIT Technology Review. It attributes product claims to Google and the assessment to Flowith CEO Derek Nee. The evidence is a single news article, not a listing.","excerpt":"Derek Nee, CEO of Flowith, an agentic AI application, told MIT Technology Review that Gemini 3 Pro addresses several gaps in earlier models."},{"dimension":"summary","status":"pass","explanation":"The reading leads with Google's Gemini 3 launch. It explains multimodal and generative interfaces in plain language and gives the travel example. It describes Gemini Agent's function and approval step, and it notes the unknowns. The title and question are faithful and non-leading.","excerpt":"allow the model to make its own choices about what kind of output fits the prompt best, assembling visual layouts and dynamic views on its own instead of returning a block of text."},{"dimension":"date","status":"pass","explanation":"November 18, 2025 matches the originalPublished UTC day. The November 18 availability is reported as the article's stated plan, not an outcome. The related-story sidebar describes later events (for example, ten years after AlphaGo). These are possible later page additions, but the reading neither uses nor relies on them. The evidence is a publisher page, so no archive disclosure is needed.","excerpt":"It will be available on the web for Google AI Ultra subscribers in the US starting November 18."},{"dimension":"support","status":"pass","explanation":"Every claim is source-stated with an accurate excerpt: the reasoning and multimodal improvements per Google, generative interfaces, Gemini Agent's connections, progress display and approval pauses, the developer platform, and Nee's integration plus his call for deeper testing. The reading's note that test results are unreported is accurate and adds no unsupported benefit.","excerpt":"Given its speed and cost advantages, we’re integrating the new model into our product,"},{"dimension":"media","status":"pass","explanation":"No media was selected, and no verified candidates were supplied, so the empty selection is consistent with the evidence.","excerpt":null}],"createdAt":"2026-11-09T23:48:44.210Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}