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-gruenbaum-reverse-graph-engineering/references/

Source outline

Source: Benjamin Gruenbaum, Graph Engineering in Reverse, AI Native DevCon New York 2026. Recording. All anchors refer to original numbered cues in the transcript. Short quotations below preserve the supplied transcription, including spoken wording.

TopicSource anchorExact short quoteApplication and scope
Company-brain pipeline00:02:02–00:04:12, cues 35–65“clean and deduplicate and normalize”Forward retrieval relies on extraction and entity resolution.
Answer-key gap00:04:16–00:05:22, cues 66–81“a real company has no answer key.”Complex real data makes evaluation difficult.
Reverse construction00:05:25–00:07:30, cues 82–109“questions and tasks defined by human experts”Start with domain questions rather than arbitrary synthetic documents.
Planted graph and corruption00:07:32–00:09:11, cues 110–132“we still know what the answers are at every stage.”Known mutations preserve the evaluation oracle.
Realistic mess00:09:16–00:10:17, cues 133–149“if the data looks too good, it's obviously fake.”Incomplete and inconsistent formatting can matter for realism.
Digital twins00:10:21–00:12:01, cues 150–175“hyper optimized for read only tasks”The described twins are targeted test systems.
Interface behavior00:12:09–00:14:50, cues 176–213“It needs to have real pagination”Compatibility includes relevant protocol behavior.
Statistical sketches00:19:07–00:21:19, cues 278–308“shape counts and distribution histogram”Reported metadata distributions guide generation.
Validation loop00:21:39–00:24:00, cues 313–348“deterministic checks and rules about the world.”Distribution and judge checks supplement deterministic validation.
World invariants00:24:12–00:25:06, cues 351–363“nobody acts before they were hired”The source uses explicit constraints to validate synthetic records.
Generation economics00:25:11–00:26:23, cues 364–380“We do a lot of caching.”Expensive generation motivates artifact reuse.
Execution-based evaluation00:27:45–00:29:41, cues 403–427“A lot of stuff you only see when you eval”Execution reveals weaknesses beyond textual skill review.

Source limitations

  • Aera scale, compatibility, generation cost, and experimental success are speaker-reported; the transcript supplies no independently inspected artifacts or reproducible score.
  • The speaker says some post-training work is unpublished and declines detail at cues 334–339. Do not invent or expose details absent from the source.
  • Statistical sketching is not by itself a privacy guarantee, and a deterministic synthetic answer does not prove real-world representativeness.
  • Early statements that the system is free are later qualified by enterprise monetization discussions; do not promise perpetual pricing or current availability.
  • Promotional interlude cues 236–246 are omitted from the bundled transcript; the remaining cue numbers retain their source positions.
  • This is a spoken transcript; unseen slides, linked repositories, and third-party research were not independently inspected. No outcome beyond the described talk is established.

skills

talk-gruenbaum-reverse-graph-engineering

tile.json