Paul Colmer @ AWS ☁️🛰️🚀🇦🇺

😍 G'day Beautiful People...PostSmith Chapter 2: Kiro Spec Driven Development...is the subject of my 2nd edition newsletter: From Musician to Agents... 🚀

Subscribe here 👇
linkedin.com/pulse/postsmit…

Chapter 1 was musician to agentic orchestrator. Chapter 2 is spec-driven development with Kiro. Please subscribe and enjoy.... 😎

Key Highlights:

🔥Don’t start with a fully autonomous orchestrator. Paul fired the agent that chose the trending topics. Two human-in-the-loop checkpoints, beat “set and forget” when the output is your voice, your livelihood and your brand.

🔥Spec first, cathedral never. Kiro + steering docs + Scaled Agile Framework-style increments produces a working Trend Scout in small slices that each showed value. Specs (requirements → design → tasks) beat vague “build me an agent” prompts.

🔥Decouple early. A list in the terminal is not a product. An elegant UI plus a bridge so the front end and back-end agents aren’t welded together. This is the difference between a demo and something you can keep evolving in production.

🔥Give the scout rules, not vibes. Trend Scout scores on explicit criteria (Summit appearance, customer impact, GenAI/agentic focus), cap at 30, use a high-quality model for scoring, Pydantic for a common data model across agents. Consistency across agents starts in the specification.

🔥Memory stops you looking lost. DynamoDB is the memory, so the system doesn’t post the same trend twice in unglamorous fashion. Persistence is part of the architecture, not an afterthought.

Let me know your learnings from your AI adventures. Drop a comment below.... 🙏

At Digital Colmer....newslettering off... 📺

P

#TheDigitalCoach #AWS #AI

😍 G'day Beautiful People...your AI AgentCore agent can now start trusted AWS Step Function workflows via MCP... without rewriting the steps....🚀

Find the full multi-page AWS Agentic AI Handout here👇
lnkd.in/p/gAwFQcKa

The model chooses when. The workflow decides how. Existing Standard and Express workflows become tools through Amazon Bedrock AgentCore Gateway. The agent decides when to run and what input to pass. Step Functions runs each step exactly as defined... so the agent can't skip or reorder them. Available wherever AgentCore Gateway is available.

Letting an agent improvise a refund process... that's a finance meeting nobody wants...😂

I turned the What's New firehose into a Flight Simulator brief. Crash in the brief... not live in front of the customer. Let's unpack the hits...🤔

📄 Scannable handout for builders and executives is attached below.

This pack's page-one answer key...

🔁 AWS Step Functions and Amazon Bedrock AgentCore Gateway: agents run trusted workflows

📊 Amazon Quick Essentials: AI for every Quick Sight Reader... rolling out in phases at no additional per-user cost

🎧 Amazon Connect Customer and AgentCore Gateway: contact flows reuse agent tools... and it's in Sydney

🧩 Amazon SageMaker AI: several AI models share one GPU... up to seven isolated slices, not in Sydney at launch

🖥️ Amazon Bedrock AgentCore CLI 1.0: one tool to build and run AI agents

Tiny glossary...

🔁 Trusted workflow = approved steps, in the approved order

🧰 MCP = a standard plug that lets AI agents use your tools

Quick one...which process would you trust an agent to START for you first? 👇

💸 Refunds
👋 New starter onboarding
📞 Customer call routing
🧾 Monthly reports
😅 Be honest...none until I've read the audit trail

Drop your emoji below 👇

That's it. Enjoy. I'm off to enjoy the weekend with my family.

At Digital Colmer...weekly Gen AI briefing off...📺

P

#TheDigitalCoach #AWS #AgenticAI

Cowritten with GenAI...😎

Exactly what Dion said...🎉

😍 G'day Beautiful People...how did I miss this awesome new cyber security feature...you can now test service control policies (SCPs) in the AWS IAM policy simulator....and the simulator now appears as an option in the IAM console.... 🚀

lnkd.in/gxmk_5u5

This makes it easier to troubleshoot a principal (role or user) yourself, because you can now see what type of policy may be blocking you. Policies you can see in the policy simulator, that could potentially block include:

🔒Identity-based policies
🔒Permissions boundaries
🔒Resource-based policies
🔒Service Control policies

This update makes it really easy...to test your principals and understand what may be blocking you along the way.

That's it. I'm off to help more customers secure their environments.

At Digital Colmer....policying off.... 📺

P

#TheDigitalCoach #AWS #CyberSecurity

Written by a human on earth.... Video courtesy of Grok Imagine....the screenshots are not real...😎

😍 G'day Beautiful People...Edition 3 of my Agentic AI Newsletter is out...keep humans in the loop...loose coupling pays off...keep the front-end thin and spec-driven dev make changes cheap. These are just some of the many learnings from this latest edition...🎸

Newsletter is here 👇
linkedin.com/pulse/postsmit…

Learn about Vite, Trend Scout, the bridge and how using specifications in Kiro...I'm actually using Kiro Crew...can help speed up your software dev lifecycle and result in 100's of automated tests with zero defects.

For those not familiar with Kiro Crew.... it is a persistent, open source development workspace (IDE) that remembers your context, learns how you work, and coordinates across your unique tools and workflows, so you come back to progress instead of another workflow to restart.

The best part...is that I can ask it write and implement a spec with minimal interaction from me...which means it can run for 2-3 hours writing, testing and finalising code...leaving me to do the final UAT, without having to keep watching my screen. I can also run several jobs at once....🚀

What tools do you use for your agentic AI code development. Please comment below.... 👇

At Digital Colmer....coding off....📺

P

#TheDigitalCoach #AI #AWS