CtrlK
BlogDocsLog inGet started
Tessl Logo
PluginPublic

ainativedev/aidevcon-2026-nyc

AI Native DevCon 2026 New York — practical agent skills grounded in conference talks

Quality

86%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

Overview
Quality
Evals
Security
Files

outline.mdskills/talk-singhal-operating-agents-at-scale/references/

Transcript outline

Use the original cue numbers in the transcript. Promotional cues 219–229 are omitted there. This navigation aid adds no independent evidence.

TimeCuesTopic
00:00:45,540–00:03:21,70013–50Heterogeneous setups and agent-use layers
00:03:26,980–00:06:06,06051–91Four operating questions and adoption funnel
00:06:10,540–00:08:10,06092–118Speaker-reported external productivity research
00:08:17,060–00:11:42,620119–168Activity, delayed outcomes and attribution problems
00:11:54,700–00:14:50,580170–210Cost factors and default-setting tradeoffs
00:16:05,140–00:18:19,900230–263Evolution and complementary visibility layers
00:18:25,460–00:19:53,500264–285Gateway identity, telemetry and blind spot
00:19:58,840–00:21:44,540287–311Router pilot, model choices and caching
00:22:12,020–00:23:39,100315–337Common session traces and content limits
00:23:44,820–00:28:23,020338–398Internal repeated evaluations and pattern-aware grading
00:28:27,740–00:29:19,020399–412Compression averages concealing task regressions
00:29:21,740–00:31:03,940413–437Versioned rollouts and ambiguous outcome wording
00:31:06,780–00:34:28,660438–481Combined checkpoints, central controls and rollback

Exact short quotations

  • “The default is a product decision.” — 00:13:31,660, cue 193.
  • “It's not what engineers do day to day.” — 00:29:19,020, cue 412.

Evidence limits

  • Reported router, compression and model-switching results apply to the speaker’s described workloads and experiments.
  • External study figures are the speaker’s reports and have not been independently checked.
  • The spoken cost-factor explanation references a visual without fully specifying a consistent formula.
  • The one-out-of-ten outcome passage has an unclear denominator and cannot establish a precise PR failure rate.
  • Traces capture structure and timing by default in the described setup, not prompts or code.
  • The transcript does not supply full APIs, implementation details, current prices or the resource-slide URLs.

skills

talk-singhal-operating-agents-at-scale

tile.json