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-moss-concurrent-agent-risk/references/

Source outline

Source: James Moss, Dealing with Agentic Coding Risk: What Breaks Going From 1 Agent to 10, 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
Five risk categories00:01:08–00:02:32, cues 21–41“multiple coding agents sharing one delivery system.”The talk addresses coordination across a shared delivery system.
Adoption and measurement00:04:54–00:07:23, cues 77–116“PRs are not like, you know, a great definitive metric”PR volume is a directional observation, not a definitive productivity metric.
Event-driven flow00:07:31–00:10:14, cues 119–162“The human is responsible for the entire outcome.”Separate agent steps do not remove human outcome ownership.
Shared state00:10:25–00:12:51, cues 164–201“independent tickets do not kind of guarantee independent changes.”Indirect shared files can cause contention.
Staging and guidance00:13:04–00:14:16, cues 204–224“technology choices that you make can affect how”The team staged schema work and supplied conflict guidance.
Stale assumptions00:14:23–00:16:22, cues 225–257“it would just silently throw away the other work”Rebase resolution can erase valid work.
Review spiral00:17:32–00:19:12, cues 278–305“11 review rounds over 23 hours.”The reported incident involved overlapping reviewers and inconsistent standards.
Convergence controls00:19:57–00:21:24, cues 317–337“recognizing that repetition.”Aggregate repeated findings and agree on standards and intervention conditions.
Duplicate work00:21:29–00:24:16, cues 338–379“atomicity and idempotency”Single-worker tests missed a duplicate-event concurrency failure.
Counter failure00:24:21–00:27:08, cues 380–419“it looked like a human to our systems”Bot identity and human-intent assumptions undermined a hard-coded limit.
Budget proposal00:27:10–00:27:48, cues 420–431“we haven't done this yet”Work-item budgeting was planned, not completed.
Earned autonomy00:29:39–00:31:09, cues 462–488“We kind of started small, right?”Autonomy expanded as the team gained confidence.

Source limitations

  • The incidents and throughput figures are reported Tessl experience; PR count does not establish productivity, quality, or causality.
  • The work-item budget approach is explicitly future work in this recording. The described interaction limit is not a recommended universal threshold.
  • Drizzle, GitHub, and issue-tracker behavior is described as of the talk; do not infer present product capabilities or invent repair commands.
  • Formal verification of the rebuilt queue is reported secondhand by the speaker; no model, proof, or implementation is supplied.
  • Promotional interlude cues 306–316 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-moss-concurrent-agent-risk

tile.json