--- schema: 1 type: video title: >- How Google DeepMind Runs Agents at Scale — KP Sawhney & Ian Ballantyne, Google DeepMind description: >- Google DeepMind employees have worse token quotas than paying customers. That is not a mistake. KP Sawhney explains: customers get priority, and if an internal team spikes usage on a cluster someone monitoring 24/7 will just call and ask them to stop. This panel covers how DeepMind thinks about agents at scale from the inside: managing quota across thousands of power users, a Darwinian skills library where only the strongest skills survive as engineers contribute en masse, and where the Deep Research pipeline is going next. KP's current focus is replacing the pipeline's giant context blobs with a shared file system so each research component can collaborate the way human researchers would, and produce supporting artifacts that the current architecture cannot. Speaker info: - https://linkedin.com/in/ianballantyne - https://github.com/irbg - https://linkedin.com/in/kyle-sawhney - https://github.com/KPSawhney Timestamps: 0:00 Introduction of KP Sawhney and Ian Ballantyne 0:57 Demo of the anti-gravity agentic platform 4:32 Discussion on KP's work with the deep research agent 5:46 Using skills libraries and managing agent sprawl 7:52 Addressing scalability and token quotas at Google 9:44 Audience Q&A: Managing per-user agent behaviors 13:12 Audience Q&A: Observability and agent trajectory stores 14:46 Audience Q&A: Future of deep research pipelines 16:10 Audience Q&A: Handling multi-agent systems 17:45 Audience Q&A: Perspectives on skills vs. MCP 19:12 Audience Q&A: Evaluating agentic workflows 20:25 Audience Q&A: Handling model limits and quota management 22:48 Audience Q&A: Automated code review processes --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/7gujZrJ9L5I-how-google-deepmind-runs-agents-at-scale-kp-sawhney-ian-ballantyne-google-deepmind Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- canonical: 'https://liquidgold.cc/watch/7gujZrJ9L5I' markdown: 'https://liquidgold.cc/watch/7gujZrJ9L5I.md' llms: 'https://liquidgold.cc/watch/7gujZrJ9L5I/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T13:20:27.982Z' language: null children: [] videoId: 7gujZrJ9L5I durationSeconds: 1513 publishedAt: '2026-05-24T18:00:06.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 71 endSeconds: 274 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=71' heldReason: rights - ordinal: 1 text: null startSeconds: 71 endSeconds: 274 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=71' heldReason: rights - ordinal: 2 text: null startSeconds: 71 endSeconds: 274 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=71' heldReason: rights - ordinal: 3 text: null startSeconds: 71 endSeconds: 274 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=71' heldReason: rights - ordinal: 4 text: null startSeconds: 284 endSeconds: 350 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=284' heldReason: rights - ordinal: 5 text: null startSeconds: 363 endSeconds: 475 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=363' heldReason: rights - ordinal: 6 text: null startSeconds: 486 endSeconds: 587 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=486' heldReason: rights - ordinal: 7 text: null startSeconds: 486 endSeconds: 587 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=486' heldReason: rights - ordinal: 8 text: null startSeconds: 486 endSeconds: 587 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=486' heldReason: rights - ordinal: 9 text: null startSeconds: 597 endSeconds: 792 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=597' heldReason: rights - ordinal: 10 text: null startSeconds: 803 endSeconds: 887 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=803' heldReason: rights - ordinal: 11 text: null startSeconds: 803 endSeconds: 887 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=803' heldReason: rights - ordinal: 12 text: null startSeconds: 898 endSeconds: 979 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=898' heldReason: rights - ordinal: 13 text: null startSeconds: 991 endSeconds: 1065 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=991' heldReason: rights - ordinal: 14 text: null startSeconds: 1076 endSeconds: 1152 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=1076' heldReason: rights - ordinal: 15 text: null startSeconds: 1162 endSeconds: 1226 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=1162' heldReason: rights - ordinal: 16 text: null startSeconds: 1249 endSeconds: 1368 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=1249' heldReason: rights - ordinal: 17 text: null startSeconds: 1379 endSeconds: 1513 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=1379' heldReason: rights - ordinal: 18 text: null startSeconds: 1379 endSeconds: 1513 watch: 'https://liquidgold.cc/watch/7gujZrJ9L5I?t=1379' heldReason: rights --- # How Google DeepMind Runs Agents at Scale — KP Sawhney & Ian Ballantyne, Google DeepMind Google DeepMind employees have worse token quotas than paying customers. That is not a mistake. KP Sawhney explains: customers get priority, and if an internal team spikes usage on a cluster someone monitoring 24/7 will just call and ask them to stop. This panel covers how DeepMind thinks about agents at scale from the inside: managing quota across thousands of power users, a Darwinian skills library where only the strongest skills survive as engineers contribute en masse, and where the Deep Research pipeline is going next. KP's current focus is replacing the pipeline's giant context blobs with a shared file system so each research component can collaborate the way human researchers would, and produce supporting artifacts that the current architecture cannot. Speaker info: - https://linkedin.com/in/ianballantyne - https://github.com/irbg - https://linkedin.com/in/kyle-sawhney - https://github.com/KPSawhney Timestamps: 0:00 Introduction of KP Sawhney and Ian Ballantyne 0:57 Demo of the anti-gravity agentic platform 4:32 Discussion on KP's work with the deep research agent 5:46 Using skills libraries and managing agent sprawl 7:52 Addressing scalability and token quotas at Google 9:44 Audience Q\&A: Managing per-user agent behaviors 13:12 Audience Q\&A: Observability and agent trajectory stores 14:46 Audience Q\&A: Future of deep research pipelines 16:10 Audience Q\&A: Handling multi-agent systems 17:45 Audience Q\&A: Perspectives on skills vs. MCP 19:12 Audience Q\&A: Evaluating agentic workflows 20:25 Audience Q\&A: Handling model limits and quota management 22:48 Audience Q\&A: Automated code review processes --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/7gujZrJ9L5I-how-google-deepmind-runs-agents-at-scale-kp-sawhney-ian-ballantyne-google-deepmind Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- Canonical: https://liquidgold.cc/watch/7gujZrJ9L5I · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.