A full day aimed at technical leaders and builders designing, building and scaling production-ready agentic systems. Keynotes on the foundations, then architecture deep dives, then labs in which every attendee builds in their own cloud environment rather than a shared sandbox.

The published list:

• Build agentic applications hands-on with Amazon Bedrock and AgentCore.
• Hear from customers running agents in production today.
• Architectural patterns for agent orchestration, governance and scaling.
• Connect with others working on similar problems.
• Accelerate initiatives through the AWS AI-DLC program.

The direct AWS/Dynatrace invitation and the AICamp listing advertise different days. Worth recording before attending, because the difference is the interesting part.

Two things follow.

The failure claim appears in one channel only. The AICamp listing keeps the customer stories and drops "what failed", which is the usual direction for that edit. It may simply be a shorter blurb. It is still the first thing to check.

And observability is explicit on the listing — "improve AI observability and operational visibility" is a stated learning outcome, not an inference from Dynatrace being in the room. That was a guess when this note was written from the invitation alone, and the listing confirms it.

"You'll hear real customer stories: what worked, what failed, and what the path from prototype to production actually looks like."

Vendor days do not usually advertise the failures. If that session delivers, it is the reason to attend; if it resolves into three case studies where everything went well, the day was a product demo with a lab attached. That is the thing to check, and it is checkable on the day.

The question to bring: what was the failure mode? The interesting answers are the ones about agents in production that this site's own notes keep circling — loops that do not terminate, context that rots over a long horizon, an eval gate nobody calibrated against a known-bad input, a governance story that is a label rather than a measurement. The uninteresting answer is "we needed a bigger model".

Adjacent to the standing question here: what separates an agent demo from an agent a team can operate. Related notes:

• CLI coding agents: 2026 Q2 comparison — the same production/prototype gap, from the tool side.
• Amazon Bedrock RAG workshop — the 2024 hands-on, for what has changed in the stack since.
• Boston AI Week 2026 and Tech Week Boston 2026 — the surrounding calendar.
• AICamp Boston, July 2026 — the same platform's local series.

The part to watch is Dynatrace's presence, and the listing settles it: two of the four speakers are Dynatrace, and operational visibility is a stated outcome. The unglamorous half of "production-ready" is observability of agent systems, and a co-presenting vendor whose business is telling you what is actually running is a better signal than the lab agenda.

Recorded because the local calendar is worth tracking as a set, not because attendance is planned. None of these has a note yet.

Note the Neo4j entry: the August workshop in this series was Neo4j and AWS. The October one is Neo4j and Google. Same graph vendor, different cloud, five weeks apart — which is its own small data point about whose stack the graph layer is courting.

Confirmed. The listing gates registration behind "Approval Required. Your registration is subject to approval" and a Request to Join, so the seat was the organizer's call; AICamp's confirmation arrived 2026-09-18. The email repeats the venue as 55 Pier 4 Blvd and asks attendees to re-verify date, time and location before travelling, which for an event this size is boilerplate rather than a signal.

Clojure/conj 2026 runs Sep 30 – Oct 2 in Charlotte. This is Sep 29 in Boston. No date overlaps, so the collision is travel, not calendar — and it is narrower than it first looks.

So the only thing genuinely at risk is the Wednesday workshop day. The talks on Thursday and Friday are livestreamed free to registrants, and whether the workshops are too is not recorded in our note — that is the thing to check before deciding, because it decides the whole question.

To be filled in after the event, or deleted if it is not attended.