Wed, Nov 11 · 3:06 PM–3:19 PM · Continental 4–6 (Mainstage)
Brian Kennedy, ZoomInfo
Don't trust an LLM judge until you've measured it: run a second check in shadow mode and look where the two disagree.
None captured.
None filed.
My Voice Memos from the room. The public livestream: YouTube, around 6:45:53 (approximate).
The scrolling talk page is built.
Cross-references across the 6 conference themes, contested debates, and connected talks.
Trust in agent-generated systems comes from owning the verification loop—deterministic rules, stakes-scaled credentials, replayable execution traces, hardened sandboxes, and continuous failure diagnosis—rather than line-by-line diff review.
Effective agentic software factories are built on timeless engineering practices—balancing automated forward pressure and rigorous back pres…
Agents render traditional software org charts, enterprise sales checklists, and third-party SaaS tools obsolete by enabling tiny, high-equit…
Post-training Gemini 4 (Carbon and Hydrogen) for multi-day software engineering requires continuously sourcing non-saturated hard tasks, rep…
Software factories fail when applied to open-ended greenfield discovery, but excel on brownfield systems where existing customers and runtim…
Building an ahead-of-time compiler for Bun to cut Claude Code's memory usage and startup latency required five rewrites over a month, succee…
Brian Kennedy, Principal Engineer, Agentic AI Platform, ZoomInfo
ZoomInfo: "ZoomInfo is the #1 GTM Platform: Marketing and sales AI plus the best B2B data." (from the site)
Machine transcript (qwen3-asr-1.7b). Garbled words are kept as heard.
A 28-second fragment: the last lines of the talk only, recorded 15:22:39–15:23:07, after the slot's scheduled end (15:06–15:19); the talk ran over. Filed on the operator's attribution, not on a name in the audio: the folder title names the speaker, and the closing lines answer the abstract (a judge measured against a second check run in shadow mode). Neither transcript names the talk, the speaker or his company.
Transcribed locally with Qwen3-ASR 1.7B (timings per ~5 s segment), prompted with the speaker names and affiliations of overlapping sessions. Qwen's text ends at 00:00:18; the closing "Thank you." is only in the Voice Memos transcript. The Voice Memos transcript is kept beside it as
transcript.txt.
Then read the cases, see where the two disagree, and don't overthink it. Just start small. We entrusted our judge until we measured it. Measure yours. Run a second check. Run in shadow mode, seeing where it disagrees. You may be surprised at what you find.
Generated Nov 13, 2026 at 5:56 AM · bun scripts/render-pages.ts