A haunted checkout gives AWS Developers an AI troubleshooting project
Morgan Willis and Rafael Lopes use a mock candy store to build an agent that investigates alarms and reports evidence to Slack.
Why this story? The channel's weekly summary: what went up this week, what is drawing views now, and where the channel stands among similar channels.
Trick or Trace: Build an AWS incident triage agent with Strands harness
Things we can learn from the video
Press to jump
- 3:07A staged checkout error at Midnight Candy Co. supplies the incident the agent will investigate.
- 6:38CloudWatch, EventBridge, Lambda and AgentCore Runtime connect the alarm to an investigation and Slack report.
- 12:12The agent loop builds context, queries the model, executes tools and feeds results back.
- 14:28Willis distinguishes the lower-level Strands Harness SDK from the packaged createHarness() tool.
- 19:44The TypeScript live-coding demonstration begins.
Midnight Candy Co. can take an order, until its checkout develops a deliberately staged problem. Morgan Willis and Rafael Lopes turn that broken purchase into the starting point for AWS Developers' most-watched new upload of the week, "Trick or Trace: Build an AWS incident triage agent with Strands harness".
The AWS product tutorial follows an incident-response architecture from CloudWatch alarms through an AI investigation to a Slack report for the on-call engineer. Willis explains the agent as a model plus its harness: the surrounding machinery that supplies context, executes tools and feeds results back. Lopes adds a practical caution about giving the agent the right permissions. The TypeScript coding demonstration begins roughly 20 minutes in, after the architecture gets its walkthrough. A viewer's question about the GitHub repository provides a familiar developer footnote: where's the code?
That stream collected about 3,200 views, the largest total among the week's three uploads. "Optimize Your Prompts with Amazon Bedrock" followed, contrasting quick prompt rewrites with more elaborate testing across models. These are company-made demonstrations of AWS tools, rather than independent product reviews.
The freshest upload, "How Streaming AI Responses Work", is already drawing the most daily views of any video on the channel. Its useful distinction: sending tokens as they are generated makes an answer appear sooner, but doesn't make the model itself faster.
Subscribers were unchanged over the week. Daily channel viewing is below its usual level, while Android Developers is experiencing a substantial lift among nearby technology channels. For AWS Developers, the new streaming lesson supplies the week's clearest change of subject: from investigating a failed checkout to shortening the wait at a blank screen.
Published on Trari
Where next?
What viewers say
Where's the github repo url?
@dahowlinSee it on YouTube
Public comments from the video's page on YouTube
The videos drawing views on AWS Developers right now
- Channel
- AWS Developers
- Subscribers
- 1,710,000
- Channel views, last 7 days
- 55,600
- Videos published this week
- 3
- Video published
- October 28, 2026
- Video views so far
- 3,206
- Views on its first day
- 1,958
The numbers
Here is the week in views, subscribers and uploads for AWS Developers.
💡 Test your knowledge
What is the mock candy store called in the incident-triage tutorial?
Midnight Candy Co. is the demonstration storefront whose checkout error starts the investigation.





