--- schema: 1 type: video title: 'The 100-Tool Agent Is a Trap - Sohail Shaikh & Ankush Rastogi, Prosodica' description: >- The common “Fat Agent” architecture loads a large catalog of tools directly into the system prompt. This often creates latency, cost, and reliability problems in production agent systems. As tool schemas take up more of the context window, agents can become slower and more likely to choose the wrong tool. This session takes a practical look at the Semantic Tool Router pattern, a deterministic layer that reduces the amount of context shown to the model in real time. The talk will share benchmarks across frontier models, including GPT-4o and Gemini 2.0, showing how the number of available tools affects Time-to-First-Token latency and tool-selection accuracy. Attendees will learn how to move from static tool loading to Just-in-Time Context Injection, where only the most relevant tools are added to the prompt for each request. In high-tool-density benchmark scenarios, this approach can reduce response latency by up to 90%, reduce cross-tool confusion, and improve agent reliability. The session will end with a practical framework for building tool routers that can scale to hundreds of capabilities without sacrificing speed or predictability. Speakers: - Sohail Shaikh (Prosodica): Sohail Shaikh is a data scientist with nearly a decade of experience across AI, data science, analytics, marketing, and software-oriented work, focused on building practical, reliable, and scalable AI systems using NLP, RAG, conversational intelligence, and LLM workflows. LinkedIn: https://www.linkedin.com/in/sohail-shaikh/ GitHub: https://github.com/Sohail-Sh - Ankush Rastogi (Prosodica): Ankush Rastogi is a Senior Data Solutions Engineer with over a decade of experience building scalable data, analytics, and machine learning platforms, with a focus on turning AI models into reliable, production-ready enterprise systems through strong evaluation, inference performance, cost optimization, and operational design. LinkedIn: https://www.linkedin.com/in/ankushrastogi/ GitHub: https://github.com/ankushrastogi04 canonical: >- https://liquidgold.cc/channel/aidotengineer/explore/video/the-100-tool-agent-is-a-trap-sohail-shaikh-ankush-r-fc53af99f3af markdown: >- https://liquidgold.cc/channel/aidotengineer/video/the-100-tool-agent-is-a-trap-sohail-shaikh-ankush-r-fc53af99f3af.md llms: null parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T09:19:35.403Z' language: null children: [] videoId: vh2VGuQ3zhY durationSeconds: 1707 publishedAt: '2026-06-28T15:15:15.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 13 endSeconds: 118 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=13' heldReason: rights - ordinal: 1 text: null startSeconds: 129 endSeconds: 268 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=129' heldReason: rights - ordinal: 2 text: null startSeconds: 268 endSeconds: 281 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=268' heldReason: rights - ordinal: 3 text: null startSeconds: 281 endSeconds: 380 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=281' heldReason: rights - ordinal: 4 text: null startSeconds: 281 endSeconds: 380 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=281' heldReason: rights - ordinal: 5 text: null startSeconds: 380 endSeconds: 392 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=380' heldReason: rights - ordinal: 6 text: null startSeconds: 392 endSeconds: 495 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=392' heldReason: rights - ordinal: 7 text: null startSeconds: 392 endSeconds: 495 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=392' heldReason: rights - ordinal: 8 text: null startSeconds: 506 endSeconds: 608 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=506' heldReason: rights - ordinal: 9 text: null startSeconds: 608 endSeconds: 619 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=608' heldReason: rights - ordinal: 10 text: null startSeconds: 619 endSeconds: 708 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=619' heldReason: rights - ordinal: 11 text: null startSeconds: 721 endSeconds: 858 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=721' heldReason: rights - ordinal: 12 text: null startSeconds: 721 endSeconds: 858 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=721' heldReason: rights - ordinal: 13 text: null startSeconds: 721 endSeconds: 858 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=721' heldReason: rights - ordinal: 14 text: null startSeconds: 869 endSeconds: 993 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=869' heldReason: rights - ordinal: 15 text: null startSeconds: 993 endSeconds: 1004 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=993' heldReason: rights - ordinal: 16 text: null startSeconds: 1004 endSeconds: 1136 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=1004' heldReason: rights - ordinal: 17 text: null startSeconds: 1147 endSeconds: 1228 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=1147' heldReason: rights - ordinal: 18 text: null startSeconds: 1240 endSeconds: 1329 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=1240' heldReason: rights - ordinal: 19 text: null startSeconds: 1329 endSeconds: 1340 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=1329' heldReason: rights - ordinal: 20 text: null startSeconds: 1469 endSeconds: 1570 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=1469' heldReason: rights - ordinal: 21 text: null startSeconds: 1469 endSeconds: 1570 watch: 'https://liquidgold.cc/watch/vh2VGuQ3zhY?t=1469' heldReason: rights --- # The 100-Tool Agent Is a Trap - Sohail Shaikh & Ankush Rastogi, Prosodica The common “Fat Agent” architecture loads a large catalog of tools directly into the system prompt. This often creates latency, cost, and reliability problems in production agent systems. As tool schemas take up more of the context window, agents can become slower and more likely to choose the wrong tool. This session takes a practical look at the Semantic Tool Router pattern, a deterministic layer that reduces the amount of context shown to the model in real time. The talk will share benchmarks across frontier models, including GPT-4o and Gemini 2.0, showing how the number of available tools affects Time-to-First-Token latency and tool-selection accuracy. Attendees will learn how to move from static tool loading to Just-in-Time Context Injection, where only the most relevant tools are added to the prompt for each request. In high-tool-density benchmark scenarios, this approach can reduce response latency by up to 90%, reduce cross-tool confusion, and improve agent reliability. The session will end with a practical framework for building tool routers that can scale to hundreds of capabilities without sacrificing speed or predictability. Speakers: - Sohail Shaikh (Prosodica): Sohail Shaikh is a data scientist with nearly a decade of experience across AI, data science, analytics, marketing, and software-oriented work, focused on building practical, reliable, and scalable AI systems using NLP, RAG, conversational intelligence, and LLM workflows. LinkedIn: https://www.linkedin.com/in/sohail-shaikh/ GitHub: https://github.com/Sohail-Sh - Ankush Rastogi (Prosodica): Ankush Rastogi is a Senior Data Solutions Engineer with over a decade of experience building scalable data, analytics, and machine learning platforms, with a focus on turning AI models into reliable, production-ready enterprise systems through strong evaluation, inference performance, cost optimization, and operational design. LinkedIn: https://www.linkedin.com/in/ankushrastogi/ GitHub: https://github.com/ankushrastogi04 Canonical: https://liquidgold.cc/channel/aidotengineer/explore/video/the-100-tool-agent-is-a-trap-sohail-shaikh-ankush-r-fc53af99f3af · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks --- #### Chunk 0: Document excerpt (0:01 to 2:09) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 1: Document excerpt (2:09 to 4:28) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 2: Document excerpt (4:28 to 6:20) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 3: Document excerpt (6:20 to 8:15) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 4: Document excerpt (8:15 to 10:08) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 5: Document excerpt (10:08 to 11:48) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 6: Document excerpt (11:48 to 14:18) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 7: Document excerpt (14:18 to 16:33) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 8: Document excerpt (16:33 to 18:56) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 9: Document excerpt (18:56 to 20:28) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 10: Document excerpt (20:28 to 22:09) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 11: Document excerpt (22:09 to 24:18) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 12: Document excerpt (24:18 to 26:10) > Held: transcript rights do not permit this chunk's text. --- #### Chunk 13: Document excerpt (26:10 to 28:27) > Held: transcript rights do not permit this chunk's text.