← Financial Cloud Cloud Cloud Club · AWS Re:cap

Vertex Macro | Financial Cloud Cloud · AWS Re:cap

AWS Re:cap 01: Encode Architecture as Steering for AI Agents

Speaker: Shiro Seike

Session: 01

Session
Summit Dev Lounge2026 Re:cap
01 Encode Architecture as Steering for AI Agents
02 Agent Harness Is the Real Engineering Moat
03 Ask Observability Data in Plain Language
04 Serverless AR Game with Bedrock AgentCore
05 Multi-Agent Quant Backtesting on AgentCore
06 Blog to Slides in Three Minutes with Kiro
AWS Community Day Hong Kong 2025 Re:cap
02 AWS Compliance with Terraform
03 Beginner to Builder An AWeSome Cloud Journey
04 Team-First Serverless Engineering with Laravel & Bref
05 Event Opening - AWS Community Day Hong Kong 2025
06 Agent-to-Agent: Building Interoperable AI on AWS
07 Utilize another telemetry data for faster improvement with AI agent
08 Graduating from Vibe Coding: Spec-Driven Development with Kiro
09 Automated Testing using MCP & AI Agents
10 Modernizing Telecom Security ML Powered Approach
11 Rethinking GenAI Agent: RAG & MCP
12 Disaster and Emergency Response with TAK and AWS
13 Rethinking Serverless Application Workflows from a Testing Perspective
14 Practical AWS FinOps for Cloud Success
15 AI-Powered Global Pure-Alpha Macro Trades on AWS: Revolutionizing Risk-Adjusted Asset Returns
FSI Recap
01 Modern Trade Lifecycle: Trading to Settlement
02 Goldman Sachs: Fast Track your applications onto Cloud - AWS Re:cap Q1/2023
03 Zurich Insurance Group: Building an Effective Log Management Solution on AWS
04 FSI Meetup 2025 Q4 - Brex Database Disaster Recovery
05 FSI Meetup 2025 Q4 - A Graviton Migration Success Story
06 FSI Meetup 2025 Q4 - Stifel Modern Data Platform
07 FSI Meetup 2025 Q4 - Financial Transaction Data Reconciler PayPal
08 FSI Meetup 2025 Q4 - Scaling Resilience
09 Maximizing AI Inference Cost Efficiency: Strategic Adoption of AWS GPU Instances
10 Advanced Agentic AI Design Patterns
11 Build New Modern Apps on AWS
AWS re:Invent 2025
01 Coinbase re:Invent Recap (IND3312)
02 Building the Future Trading Platform Leveraging AI and AWS
03 Trading Innovation: Jefferies' AI Assistant on Amazon Bedrock (IND3315)
04 How FSI Revolutionized HFT Analytics with Agentic AI (GBL302)
05 Improving Distributed Systems with Amazon Time Sync Featuring Nasdaq
06 Amazon Aurora HA and DR Design Patterns for Global Resilience (DAT442)
07 Building Agentic AI: Amazon Nova Act and Strands Agents in Practice (DEV327)
08 Deep Dive into Amazon Aurora and Its Innovations (DAT441)
09 Deep dive on Amazon S3 (STG407)
10 Nasdaq: Build Resilient Infrastructure for Global Financial Services (HMC327)
11 What's New with AWS Lambda (CNS376)
12 Spec-Driven Development with Kiro (DEV314)
13 Amazon's finops: Cloud cost lessons from a global e-commerce giant (AMZ308)
14 Tick to trade latency trading platforms on aws
Government data
01 The AI Era: The Boundary Between Development and Design Is Disappearing
02 On-Device Multimodal AI and Smart-City Practice
03 Large-Model Capability Evaluation and a Method for Landing AI Projects
04 Controlled End-to-End Automation of Government Development with Cloud Agents
05 A New Software Ecosystem for the Agent Era, Seen Through Multi-Agent Systems
06 AI-Driven Macro Quantitative Research and Smart Governance
07 Authorized Operation of Public Data and Smart-Government Practice
08 Putting Data Assetization into Practice: Rights, Compliance, Engineering Governance, and Digital-Government Cases
09 AI for Mental-Health Public Welfare: Governance, Architecture, and Practice of a Trustworthy Platform
Amarathon 2025 Recap
01 A Developer’s Roadmap to Architecting for Agents
02 Amazon Bedrock Data Automation
03 Multi-Agent on AgentCore
04 Building Agentic AI Nova Act and Strands Agents in Practice
04 Accelerating Migration Projects with Kiro using Spec-Driven Development
06 From Matching to Understanding: Personalized AI Search Practice Driven by AgentCore Memory
07 Observe to Optimize – LLM Observability to AIOps Turning real-time insights into intelligent automation
08 Deploying TEAM and Building the Best Engineering Team
09 Five Hard Lessons from Five Years of So-Called Serverless Databases
14 What if AI does my job How Q Developer CLI and Kiro have changed my daily routine
16 Velocity with Vigilance: Security Essentials for Amazon Bedrock Agent Development
26 Run OSS LLMs on a Single H100 Smarter, Cheaper, Faster
28 A Modern Unified Metadata Architecture: New Approaches to Breaking Down Data Silos
29 Serverless MediaOps: Automating Video Workflows with AI on Amazon Web Services
30 Architecting for Efficiency and Reliability with Performance Testing at Scale
31 Connecting the World Through Open Source: Practical Journey of Technology, Community and Global Developer Relations
33 Building Streaming Iceberg Tables for Real-Time Logistics Analytics
34 Accelerating Large-Scale Robot Strategy Training: An Automated Closed-Loop Architecture Based on Kiro, Trainium, and EKS
35 From Vibe to Viable with spec driven development
36 Making Cloud Cost Analysis Smarter: Building FinOps Intelligent Agents with Strands and AgentCore
37 Transform Conversational Agentic AIOps for K8s Using CNCF Kagent, K8sGPT, and Nova Sonic

AWS Hong Kong Summit 2026 · Developer Lounge Recap · Article 1 of 6

AWS Magazine · Field Report from Victoria Harbour


The Scene: High Finance Meets High Tech on Victoria Harbour

Having covered the evolution of computing across six decades—from mainframe punch cards in the 1960s to the dawn of autonomous cloud agents in 2026—I can tell you that true paradigm shifts have a distinct smell. In Hong Kong, it is the aroma of freshly pulled artisanal espresso blending with high-altitude air conditioning, overlooking the glittering glass towers of Central and the Victoria Harbour waterfront.

Hong Kong does not do technology in quiet whispers; it accelerates capital and code with identical velocity. At the AWS Hong Kong Summit 2026, held at the iconic HKCEC, the Developer Lounge served as the high-octane nerve center of the entire conference.

Here, high-tech infrastructure collided with Hong Kong luxury. International speakers flew in from Fukuoka, Tokyo, Taipei, and Singapore. Builders with badge lanyards slung over tailor-fitted jackets leaned over whiteboards and ultra-thin laptops. Venture capitalists and startup founders traded term sheets over iced milk tea and single-malt whisky in adjacent private suites, calculating GPU burn rates and runway while engineers debated agentic workflows on stage.

+-----------------------------------------------------------------------------------+
AWS HONG KONG SUMMIT 2026: DEVELOPER LOUNGE

[ Victoria Harbour View ]  <-->  [ AWS High-Tech Lounge & Agentic Demos ]
  - High Finance / VCs              - Mechanical Keyboards & Whiteboards
  - Startup Dealmaking              - Live Code Steering & CI/CD Pipelines
  - Private VIP Dinners             - International Speakers (Fukuoka to HK)
+-----------------------------------------------------------------------------------+

In the center of this energetic lounge, Fukuoka engineer Shiro Seike (@seike460) took the mic for a 15-minute presentation that completely reshaped how attendees viewed AI-assisted development.


The Keynote Insight: Architecture as Steering

During his session, titled "Architecture as Steering: On-Ramp to AI-DLC," Seike delivered an unforgettable warning that brought the bustling room to a absolute pause:

"Your AI coding agent will quietly cross an architectural boundary while every test stays green. The compiler can’t see it. A bigger model won’t fix it — because the boundary was never written down."

+-----------------------------------------------------------------------------------+
THE AI CODING AGENT PARADOX

1. [ Prompt / Requirements ] --> [ AI Agent (Bedrock / Claude Code) ]
2. Agent generates TypeScript code
3. [ Automated Unit Tests ] --> Status: GREEN (PASS)  Compiler is happy
4. Domain architecture --> BROKEN (infra leaked into core domain)
+-----------------------------------------------------------------------------------+

Speaker Profile: Community-Built Leadership

Shiro Seike returned to Hong Kong with deep gratitude to the local community, building on his previous appearances at AWS Community Day and JAWS-UG (Japan AWS User Group) events.

Profile Field Speaker Details
Name Shiro Seike (@seike460)
Company Fusic Co., Ltd. (Fukuoka, Japan)
Title Principal Engineer / Evangelist
Community Leadership JAWS DAYS 2026 Chair; JAWS-UG Fukuoka Leader
Builder Track AWS Community Builder (Serverless)
Recognition AWS Ambassador; 2026 & 2025 Japan AWS Top Engineer
Certifications Cloud Practitioner, Solutions Architect Associate, Security Specialty, Solutions Architect Professional, DevOps Engineer Professional
Company Slogan OSEKKAI × TECHNOLOGY (Thoughtful helpfulness meets sharp engineering)

● Slide Deck: speakerdeck.com/seike460/architecture-as-steering-on-ramp-to-ai-dlc

● GitHub Repositories:

● github.com/awslabs/aidlc-workflows

● github.com/fusic/architecture-as-steering


Deep Technical Breakdown: The 3 Chapters

Chapter 1 — Diagnosis: The Compiler Said Yes; The Boundary Said No

In the Developer Lounge demo, Seike demonstrated a common failure mode in modern AI-assisted engineering. An AI coding agent is tasked with implementing order confirmation logic. The generated TypeScript compiles cleanly, and all test suites pass, yet the core domain model is fundamentally corrupted:

// packages/core/src/order/ConfirmOrder.ts
// THE DOMAIN LAYER
import { DynamoDBClient } from "@aws-sdk/client-dynamodb"; // X Infrastructure leaking into pure domain
import { Pool } from "pg";

export class ConfirmOrder {
  async run(id: string) {
    const order = await new Pool().query(/* ... */); // X Domain layer directly querying the Database
    if (!order) throw new Error("not found");        // X Raw exception throwing inside core domain

    order.status = "CONFIRMED"; // X Direct mutation of the aggregate state
  }
}

Diagnostic Checklist

● tsc (TypeScript Compiler): Clean (0 Errors)

● vitest (Test Runner): All Green

● Architecture State: Broken

Hands-on Reality Check: Upgrading to a larger, more expensive model (e.g., throwing more token money via Amazon Bedrock) does not solve this issue. AI agents optimize for green test outputs and syntactical completion. If structural architectural rules are not explicitly encoded, the agent will breach them.


Chapter 2 — Prescription: From Pain to Proof via Steering

What is Steering?

Steering = The explicit, executable architectural rules that an AI agent cannot infer on its own.

Instead of relying on sprawling, unorganized prompts, engineering teams define strict boundaries across localized steering files (such as a root CLAUDE.md and dedicated configuration rules):

● Core Protection: packages/core contains zero Node, HTTP, Database, or AWS SDK dependencies.

● Inward Dependency Flow: Outer adapters depend on inner core ports; core never imports outer layers.

● Clean Interface Contracts: Public APIs are exposed strictly via index.ts barrel files.

+-----------------------------------------------------------------------------------+
HEXAGONAL ARCHITECTURE DIRECTION

1. [ Outer Adapters: Web/HTTP, AWS SDK, PostgreSQL ]
      depends inward
2. [ Ports / Interfaces: Contract Definitions ]
      depends inward
3. [ Pure Core Domain: Zero NPM Runtime Dependencies ]
+-----------------------------------------------------------------------------------+

Pillar 1: Clean Architecture Enforced in CI with dependency-cruiser

// .dependency-cruiser.cjs
// Runs automatically in CI on every Pull Request
module.exports = {
  forbidden: [
    {
      name: "core-no-runtime-deps", // Core must not import npm modules or built-in runtimes
      severity: "error",
      from: { path: "^packages/core/src" },
      to: { dependencyTypesNot: ["local", "core"] }
    },
    {
      name: "core-no-workspace-import", // Core must not import outer workspace modules
      severity: "error",
      from: { path: "^packages/core/src" },
      to: { path: "^packages/(?!core/)" }
    }
  ]
};

Pillar 2: Hexagonal Ports with Zero-Runtime-Dependency Guardrails

// Architectural Guardrail Test
it("core package.json has empty runtime dependencies", () => {
  const manifest = JSON.parse(read("packages/core/package.json"));
  expect(manifest.dependencies).toEqual({}); // PASS - Strictly enforced by design
});

Pillar 3: Domain-Driven Design (DDD) Invariants via Property-Based Testing

Instead of writing static unit tests that an agent can easily game, property-based testing generates hundreds of automated permutations using tools like fast-check:

import * as fc from "fast-check";

const statusArb = fc.constantFrom(...ORDER_STATUSES);
const terminals = [OrderStatus.confirmed(), OrderStatus.cancelled()];

test("terminal orders cannot transition to any other status", () => {
  fc.assert(fc.property(statusArb, (n) =>
    terminals.every((s) => !s.canTransitionTo(OrderStatus.of(n)))
  ));
}); // Executes 100 random state checks per test run automatically

The Enforcement Matrix

Architectural Concern dependency-cruiser Property / Fitness Test AI Automated Review
Clean Architecture ✓ CI Gate (Active Error) Partial (Type Check / Lint) —
Hexagonal Isolation Directional Dependency Check deps: {} Guardrail Test Guard Hook Integration
DDD Invariants Out of Scope fast-check Property Verification Semantic Rule Review

Code Output Comparison: Steering OFF vs. Steering ON

● Without Steering (Steering OFF):

// Mutates domain object directly and couples to database driver
import { Pool } from "pg";
order.status = "CONFIRMED";

● With Steering Enabled (Steering ON):

// Returns an immutable Result object and persists strictly via domain ports
const r = order.confirm() // Returns immutable Result
  .asyncAndThen((o) => this.orders.save(o)); // Handled safely through abstract port


Chapter 3 — On-Ramp: Convergence with AWS AI-DLC v2 & Kiro

AWS has aligned with this methodology through AWS AI-DLC v2 (AI Development Life Cycle) and tools like Kiro IDE and Claude Code.

// .kiro/.../tools/process-checker.js
// AWS AI-DLC v2 (MIT-0 License)
// Deterministically verifies artifact existence on disk rather than trusting LLM output

process.exit(failures.length === 0 ? 0 : 1);
// Exit Code 0: PASS
// Exit Code 1: FAIL (Missing required artifacts or unauthorized file modifications)

AI-DLC Implementation Phases

Lifecycle Phase Current Status Description
INCEPTION Active Development Branch Initial stage modeling and boundary definition (main branch is stable).
CONSTRUCTION Active Fan-Out Deconstructs tasks into localized units governed by steering rules.
OPERATIONS Planned Future Stage Production monitoring, runtime telemetry, and feedback loops.

3-Step Action Plan for Developers

1. Today (5 Minutes): Document your first critical boundary in CLAUDE.md as an explicit Steering Rule.

2. This Week: Integrate dependency-cruiser into your CI pipeline to catch illegal imports.

3. In Production: Clone fusic/architecture-as-steering and adopt awslabs/aidlc-workflows.


Developer Lounge Culture: Money, Capital, and Code

Outside the formal session track, the Developer Lounge reflected Hong Kong’s status as a top-tier international innovation hub.

+-----------------------------------------------------------------------------------+
HONG KONG DEVELOPER LOUNGE & ECOSYSTEM VIBE

[ Global Builders ]  -->  Fukuoka, Tokyo, Taipei, HK local tech talent
[ Tech Stack ]       -->  Amazon Bedrock, Kiro IDE, Claude Code, Serverless
[ Business Capital ] -->  Startup pitching, VC term sheets, private meetups
[ Luxury Experience] -->  Victoria Harbour skyline, espresso bar, VIP lounges
+-----------------------------------------------------------------------------------+

● Capital Velocity: Startup founders met with global venture partners over coffee, transitioning directly from technical architectural discussions to seed and Series A funding talks.

● Private Technical Meetups: In private meeting rooms from Causeway Bay to Central, engineers hashed out Bedrock token optimization strategies and agentic workflow guardrails.

● High-Tech Elasticity: Attendees experienced AWS's vision of autonomous software development—combining low-friction development with deterministic safety checks.


Executive Summary & Quick HTML Copy Reference Chart

For easy publishing to internal dev portals or company dashboards, use this summary reference:

+-----------------------------------------------------------------------------------+
AWS SUMMIT HONG KONG 2026 - LOUNGE RECAP SUMMARY

Event              AWS Summit Hong Kong 2026 (HKCEC)
Featured Session   Architecture as Steering: On-Ramp to AI-DLC
Key Speaker        Shiro Seike (@seike460) - AWS Ambassador / Fusic Co., Ltd.
Core Problem       AI agents pass unit tests while silently breaking domain limits
The Solution       Steering = executable rules enforced via CI and property tests
Key Tools          dependency-cruiser, fast-check, AWS AI-DLC v2, Kiro IDE
Primary Repos      github.com/fusic/architecture-as-steering
                   github.com/awslabs/aidlc-workflows
+-----------------------------------------------------------------------------------+