AI Dev 26 Field Card

The talk is pure Yegge-shape from the most discourse-shaping coding-agent vendor, with "Era" framing serving as the continuity move that smooths over discontinuity. The session is foil exhibit territory. Capture verbatim. Conflicts: Bain (Stage 2) is running "Autonomous AI Agents in the Wild," which is management-consultancy case-study framing and prime foil material for the AI Alibi paper. Mike Chambers (AWS, Stage 3) has "The Loop Was Never the Hard Part," which is Brooker-adjacent and possibly PSF-adjacent if he displaces the difficulty to evaluation or deployment rather than the loop itself. Default is Cursor for the durable foil. If you can peek into Bain and see data slides, flip there, since case-study evidence is harder to recover from a recording than vendor framing.

The title is PSF in plain language. Andela is a distributed talent platform with proprietary visibility into what happens when AI deploys across global engineering teams, so the underlying data may be evidence-grade rather than rhetoric-grade. KJ is almost certainly Kennith Jackson, SVP of Solutions at Andela. He published "Why hiring 'AI engineers' won't work" on April 23, three days before the conference. The article reads like a public companion to the talk and gives you a 3-day preview of his framing: a 75% fail rate on basic AI skills assessments, a missing variable he calls "technical taste," and a three-archetype decomposition (Prototypers, Builders, Scalers) plus Forward Deployed Engineers. The framing is PSF-adjacent rather than foil. He sees the proxy-substitution problem but treats technical taste as a static property to select for, missing the dynamic capacity-erosion mechanism. Q&A prompt if you get face time: "Your archetype decomposition reads as a fix to a static classification problem. What changes if technical taste is dynamic, eroding rather than fixed, when engineers engage AI systems heavily?" The question signals you have read the article, names a mechanism distinction without naming PSF, and tests whether his team has thought about the dynamic case.

Maloney was on the Day 1 panel, and this is his team's deeper take. Listen for what "production-grade" means in their account, since the phrase usually contains a proxy substitution. Useful for the LandingAI exhibit set.

Direct PSF target. Connect the talk to the Koren et al. Tailwind anchor (downloads up, revenue −80% under vibe coding) and to Catasta's framings from the Day 1 panel. The "master class" format is interesting because it is instruction-coded rhetoric, which makes the implicit evaluation criteria more visible than a keynote would. Capture verbatim.

AI reviewing AI is recursive proxy in pure form. Listen for what "review" means in their account and against what standard. The talk pairs with the Sonar session at 2:30 for an evaluator-vendor exhibit set.

Sonar is the code-quality vendor, so this is an evaluator-tooling company explicitly framing the evaluation question. High-yield evaluation discourse. Listen for whose definition of "quality" is operative and how it gets measured.

IDE vendor augmentation rail. With Cursor at 9:00 and JetBrains here, plus Elangovan from Day 1, you will have a complete vendor-exhibit set on the post-craft engineering mode. The "shift" framing is itself a continuity move worth flagging.

Observability at scale, the natural follow-on to Chase from Day 1. The "100 agents" framing presupposes scale and velocity as the success variable. Datadog is a heavyweight observability voice in enterprise, so the talk captures the enterprise side of what Chase represents on the developer-tools side.

Moura has been on your radar already. "Governed enterprise workflows" is governance discourse that pairs with the AI Alibi paper. Sample if you stayed on Stage 3 after Bishop, or skip and use the slot to consolidate.

Vibe coding rhetoric extended into ML training. A brief sample completes the vibe-engineering exhibit set across software, code review, and ML. If energy is gone by here, skip and head to Happy Hour to find the Andela team.

Approach the Andela team if you have not already. Try to find the Bain people if you missed them in the morning. Brooker if he is still around from Day 1. Have cards ready and be in follow-up mode.