--- schema: 1 type: video title: 'Harnesses in AI: A Deep Dive — Tejas Kumar, IBM' description: >- The agent hit a login page, panicked, reported success anyway, and the upvote never happened. Tejas Kumar's diagnosis: not a prompt problem. A harness problem. The demo builds a browser agent on GPT-3.5 Turbo (consciously choosing a VERY old model to show how good harness eng can improve it a lot) against Hacker News and layers in a harness without touching the prompt once. Guardrails cap iterations and compact context. A verify step reads the tool call history to catch the agent lying about what it did. A login handler watches the browser URL each loop and injects credentials programmatically when it hits the login page. By the end the cheap old model reliably logs in and upvotes the post. Speaker info: - https://x.com/TejasKumar_ - https://www.linkedin.com/in/tejasq/ - https://www.youtube.com/watch?v=NdrSPm6NCdk Timestamps: 0:00 Introduction to Tejas Kumar and AI Harnesses 1:45 Why we use harnesses: Reliability and control 3:00 Defining an agent harness from first principles 4:32 Key components of an agent harness (Tooling, Context, Guardrails) 5:59 Starting the demo: Building a browser agent 7:00 Inspecting the initial agent loop 8:12 The problem: Agent failure and hallucination 10:20 Adding guardrails and context management 11:54 Refactoring into a formal harness 13:02 Implementing a verify step to catch lies 15:36 Implementing a login handler for programmatic access 17:42 Final demonstration: Successful autonomous upvoting 18:34 Summary and the future of dynamic harnesses --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/C_GG5g38vLU-ai-agent-harnesses 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/C_GG5g38vLU' markdown: 'https://liquidgold.cc/watch/C_GG5g38vLU.md' llms: 'https://liquidgold.cc/watch/C_GG5g38vLU/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T14:10:53.787Z' language: null children: [] videoId: C_GG5g38vLU durationSeconds: 1227 publishedAt: '2026-05-17T17:30:06.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 27 endSeconds: 114 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=27' heldReason: rights - ordinal: 1 text: null startSeconds: 125 endSeconds: 189 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=125' heldReason: rights - ordinal: 2 text: null startSeconds: 200 endSeconds: 276 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=200' heldReason: rights - ordinal: 3 text: null startSeconds: 200 endSeconds: 276 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=200' heldReason: rights - ordinal: 4 text: null startSeconds: 289 endSeconds: 364 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=289' heldReason: rights - ordinal: 5 text: null startSeconds: 289 endSeconds: 364 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=289' heldReason: rights - ordinal: 6 text: null startSeconds: 432 endSeconds: 495 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=432' heldReason: rights - ordinal: 7 text: null startSeconds: 505 endSeconds: 626 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=505' heldReason: rights - ordinal: 8 text: null startSeconds: 626 endSeconds: 636 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=626' heldReason: rights - ordinal: 9 text: null startSeconds: 796 endSeconds: 942 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=796' heldReason: rights - ordinal: 10 text: null startSeconds: 953 endSeconds: 1071 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=953' heldReason: rights - ordinal: 11 text: null startSeconds: 1071 endSeconds: 1083 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=1071' heldReason: rights - ordinal: 12 text: null startSeconds: 1117 endSeconds: 1128 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=1117' heldReason: rights - ordinal: 13 text: null startSeconds: 1117 endSeconds: 1128 watch: 'https://liquidgold.cc/watch/C_GG5g38vLU?t=1117' heldReason: rights --- # Harnesses in AI: A Deep Dive — Tejas Kumar, IBM The agent hit a login page, panicked, reported success anyway, and the upvote never happened. Tejas Kumar's diagnosis: not a prompt problem. A harness problem. The demo builds a browser agent on GPT-3.5 Turbo (consciously choosing a VERY old model to show how good harness eng can improve it a lot) against Hacker News and layers in a harness without touching the prompt once. Guardrails cap iterations and compact context. A verify step reads the tool call history to catch the agent lying about what it did. A login handler watches the browser URL each loop and injects credentials programmatically when it hits the login page. By the end the cheap old model reliably logs in and upvotes the post. Speaker info: - https://x.com/TejasKumar\_ - https://www.linkedin.com/in/tejasq/ - https://www.youtube.com/watch?v=NdrSPm6NCdk Timestamps: 0:00 Introduction to Tejas Kumar and AI Harnesses 1:45 Why we use harnesses: Reliability and control 3:00 Defining an agent harness from first principles 4:32 Key components of an agent harness (Tooling, Context, Guardrails) 5:59 Starting the demo: Building a browser agent 7:00 Inspecting the initial agent loop 8:12 The problem: Agent failure and hallucination 10:20 Adding guardrails and context management 11:54 Refactoring into a formal harness 13:02 Implementing a verify step to catch lies 15:36 Implementing a login handler for programmatic access 17:42 Final demonstration: Successful autonomous upvoting 18:34 Summary and the future of dynamic harnesses --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/C_GG5g38vLU-ai-agent-harnesses 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/C_GG5g38vLU · 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.