Hey dev team!
Alexa+ has launched in India! That brings us up to 12 countries, with the US, Italy, Mexico, Canada, the UK, Spain, Austria, Germany, France, Brazil, and Australia. Where's next?
Our partners over on the AWS Builder team have launched an app, with articles, workshops with free sandbox environments, community spaces, events, courses, and a Wishlist where you can give feedback directly to AWS product teams.
A bunch of Dev Updates too, no matter what platform you're developing for: Amazon Appstore, Alexa+ Smart Home, Ring Appstore, Android, plus a new beta release from our friends at Expo.
If you couldn't make it out to Agent Conf last week, check out the recording of Anisha and Kourtney's talk, where they explain how a team of agents is like a football team (soccer for us yanks).
Our Build, Ship, Shape: Amazon Developer Hackathon is ongoing, so be sure to sign up, and we'll be in Berlin for Next App Dev Con in just a couple weeks!
:mobile_phone_with_arrow: Amazon Appstore ends support for Easy Port and the Billing Compatibility SDK — As of September 15, 2026, Amazon Appstore is ending support for Easy Port and the Billing Compatibility SDK; migrate to the current Amazon In-App Purchasing (IAP) SDK to keep purchases working.
🏠 Alexa+ smart home devices get an Automated Testing Service — the new Alexa Automated Testing Service (ATS) lets you create test plans that define voice commands and expected results (from templates or from scratch), then run them against your device and download reports you can save for a CI/CD pipeline.
🔘 Ring Developer Portal: FAQ and Certification Best Practices — Ring's Developer Portal now publishes a certification FAQ and best-practices guide, so review it before submitting your Ring integration for certification to avoid rejections.
🔘 Ring adds two video-encryption modes to the Partner API — Ring's Partner API now supports two additional video-encryption modes, so review the release notes and update your encryption configuration if you consume Ring video streams.
🔘 Ring updates its Appstore Marketing Guidelines — Ring's Marketing Guidelines now source approved product imagery from the Ring Press Center and put the Appstore badge download link on each badge entry, so pull fresh assets from there when you market your Ring integration.
🛠️ Expo SDK 58 Beta is now available — Expo SDK 58 Beta is available, letting you upgrade your Expo apps now to test the new release ahead of its stable launch.
🤖 Android Bench 2.0: Pushing the frontier with challenging long-horizon tasks — Google's Android Bench 2.0 adds harder long-horizon multi-step tasks for evaluating on-device AI agents, so benchmark your agent against the new suite to gauge how it handles extended, real-world Android workflows.
:alexa: Alexa+ launches in India
- Amazon's next-gen Alexa+ assistant just went live in India, widening the international footprint your skills and devices will eventually need to serve.
👷 Introducing the AWS Builder app
- The AWS Builder app puts the whole builder.aws.com experience (docs, community content, and re:Post answers) in your pocket.
🤖 Introducing Strands Harness
- Strands Harness, newly open-sourced under Apache 2.0, is a pre-assembled agent you spin up in one line of Python or TypeScript: it runs across Bedrock, Anthropic, OpenAI, Google, Ollama, or LiteLLM with built-in shell, file, and web tools, context management, and long-term memory, at a claimed 28% lower token cost than comparable setups.
🎬 Five-Service Streaming Bundle Launches on Prime Video in the U.S.
- Prime Video just launched the first-ever five-service streaming bundle in the U.S., a real peek at how people will subscribe to streaming going forward: one bundle, one bill, many services.
👓 New Ways to Work with SPECS for Enterprise
- Snap is opening its Specs glasses to enterprise deployments on Snap OS, including working with AWS to connect SPECS to Amazon Quick.
This autumn we're across Europe: at Next App Dev Con Berlin, Developer Advocate Anisha Malde joins a panel, Giovanni Laquidara leads a workshop, and I give a lightning talk, and I'm also giving a talk at CityJS Athens. Our Build, Ship, Shape: Amazon Developer Hackathon is underway online too, so build your project before the deadline, and we hope to see the rest of you in Berlin and Athens!
| Date | Event | Location | Description |
|---|---|---|---|
| Aug 31 – Oct 23 | Build, Ship, Shape: Amazon Developer Hackathon | Online | Our first global developer hackathon: build on Fire TV, Alexa+, Ring, or Bee across four tracks for a share of a $190K pool of cash and AWS credits, with submissions closing October 23. |
| Sept 24 | React Native Connection | Paris, FR | French React Native conference covering the framework, ecosystem, and platforms built on RN. |
| Oct 7–9 | Next App Dev Con Berlin | Berlin, DE | Dev Advocate Anisha Malde joins a panel, Giovanni Laquidara leads a workshop, and Moses Roth gives a lightning talk. Mobile and app-platform conference at Berlin's CityCube, with workshops, talks, and platform announcements. |
| Oct 21–23 | CityJS Athens | Athens, GR | Dev Advocate Moses Roth is giving a talk. Athens edition of the CityJS conference series, covering modern JavaScript and frontend frameworks. |
| Oct 21–22 | NAB Show New York | New York City, US | East-coast edition of NAB Show — broadcast, media, and entertainment-tech industry conference, at the Javits Center. |
| Oct 22–23 | AGNTCon + MCPCon North America | San Jose, US | Linux Foundation conference for the open agentic AI ecosystem, with MCP and agent-systems tracks for builders, contributors, and enterprises. |
| Oct 23–26 | React Advanced London | London, GB | Annual advanced-React conference in London for senior practitioners working with the framework at scale. |
| Oct 28–29 | GitHub Universe | San Francisco, US | GitHub's flagship two-day event for developers, maintainers, and technical leaders, focused on collaboration and workflows in the agentic era. |
| Oct 29–31 | React India | Goa, IN | India's largest React conference, covering web, mobile, and emerging React platforms. |
| Nov 9–12 | Web Summit | Lisbon, PT | One of Europe's largest tech conferences, with broad coverage across AI, devices, startups, and platforms. |
| Nov 10–12 | AI Engineer Code Summit | San Francisco, US | The code-focused summit in the AI Engineer conference series, on agent and LLM development. |
Anisha and Kourtney's talk from Agent Conf last week is up, check it out. What AWS's Matt Wood told employees about the future of AI is a candid look at where the CTO thinks all of this is headed, straight from the internal all-hands. If you'd rather build than read, Build Your First AI Agent on AWS (in about the time it takes to drink a coffee) does exactly what the title promises, and Quit Tokenmaxxing: 4 Context Engineering Techniques for AI Agents is the video I'd send to anyone whose agents are burning through context.
📽️ Agents FC: Harness Agents Into a Winning Team | Anisha Malde & Kourtney Meiss at Agent Conf 2026
Amazon Developer Advocates Anisha Malde and Kourtney Meiss use a football-team analogy at Agent Conf 2026 to break down the engineering behind reliable multi-agent systems: clear roles, shared rules, coordination, and observability. They evolve a demo called Agents FC from an opaque black box into an observable, governed system, covering the agent loop, orchestration, context engineering, and the three layers of a strong harness. The ideas then map onto Amazon Bedrock AgentCore in a live multi-agent football simulation, so if you are building agent teams on AgentCore the talk gives you a concrete pattern for structuring and monitoring them.
📰 Connect your AI coding agent to AWS
By Ana Cunha
The Agent Toolkit for AWS closes the gap between your model's training cutoff and AWS's actual current state: it wires your coding agent to live AWS docs and API references, task-specific Skills written by the service teams, and scoped calls to your own AWS account within your existing permissions. It runs on the agent you already use (Kiro, Claude Code, Codex, Cursor) via the AWS MCP Server, so setup is either a copy-paste prompt from the Console banner or the aws configure agent-toolkit wizard. The Skills are open source, so you can read exactly what workflow each one steers the agent through, or contribute your own. If you have ever watched an agent confidently recommend a deprecated API, this is the fix worth wiring in.
📽️ What is an API? (The Only Explanation You'll Need)
Axel Espinosa builds the case for splitting logic out of a monolithic website so a web frontend and a mobile app can share one backend, then codes a working API endpoint live. A restaurant analogy grounds the concepts (the menu is the interface, the waiter is access control, the kitchen is the backend you never have to think about) before it lays out where REST, WebSocket, and GraphQL each fit. A solid explainer to hand a junior teammate or to sharpen how you frame API design in your own reviews.
📰 Build Your First AI Agent on AWS (in about the time it takes to drink a coffee)
By Vishal Alhat
A hands-on walkthrough with the full code inline: a "why did my bill go up?" agent that pulls a cost breakdown, reasons over the numbers, and names the top drivers in plain English. The stack is the current AWS answer to the duct-tape era, one line each: Strands Agents is the open-source loop, Bedrock AgentCore is the managed runtime with sessions, memory, and guardrails, and Kiro handles the spec-first build. The whole agent is two files, and it ships with mock cost data by default so you can run it without touching your billing account, flip one flag to hit the real Cost Explorer API. A useful demo-craft note too: the mock keeps the data deterministic on stage while the reasoning and tool-calling stay entirely real.
📽️ CI/CD for Beginners
Jonathan Vogel builds a deployment pipeline from scratch on a real app, starting with a manually-deployed Lambda function and adding one layer at a time. The host replaces ad-hoc CLI commands and half-remembered deploy steps with a single GitHub Actions workflow file triggered by a git push, then layers in tests for continuous integration. If you have ever gone back to a project weeks later and had to reconstruct how you shipped it, this is a clean template for turning that manual process into something a file handles for you.
📰 You Earned AWS Credits. Now What?
By Rufeda Ali
A practical guide for turning Student Rewards AWS credits ($30 across 7 and 14 badges) into a first hands-on project. It covers how credit redemption works in the Billing and Cost Management console, which services credits do and don't cover (S3, EC2, Lambda, DynamoDB, Bedrock, Polly, and more, but not Marketplace or Training and Certification fees), and the Free Tier interplay for new accounts. If you mentor students or run a workshop, it's a clean reference for getting newcomers past the "I have credits, now what" hurdle with console-only starter projects like Polly speech synthesis or Comprehend sentiment analysis. Super helpful if you get some credits through the Build, Ship, Shape hackathon.
📽️ Quit Tokenmaxxing: 4 Context Engineering Techniques for AI Agents
Morgan Willis breaks context engineering down into four patterns for keeping an agent's context window lean: compress what's old, externalize what's large, select what's relevant, and isolate what doesn't belong. Each technique comes with working Strands Agents code, so you can see how to shape what enters, stays in, and gets transformed out of the window rather than just stuffing everything in. If you're building agents and watching token counts (and latency) climb, this is a practical reference for spending that budget deliberately.
📰 What AWS's Matt Wood told employees about the future of AI
At the AWS Global Meeting on September 17, AWS Chief AI and Technology Officer Matt Wood framed the company's AI strategy as a rerun of its cloud playbook, naming Trainium for compute, Bedrock for inference, and AgentCore for agents as the building blocks. To make the point he brought startup Odyssey on stage, whose Odyssey-3 foundation world model trains and runs on Trainium3 and learns physics from raw video rather than hand-coded rules. The developer-facing surfaces here are the ones Wood named, Trainium, Bedrock, and AgentCore; the world models themselves are a reference for where frontier workloads are heading, not yet an API you can call.
📰 CW@60: Nobody notices when it works
For Computer Weekly's 60th anniversary, Amazon CTO Werner Vogels reflects on two decades spent building the kind of infrastructure people only notice when it fails. It's a thoughtful read on the unglamorous discipline behind reliable systems, and why the work that runs in the background is often the work that matters most.
📰 How Hyundai Turned Amazon Autos Into Dealership Sales
Hyundai was the exclusive launch partner for Amazon Autos, and this case study is about the measurement plumbing behind it: a privacy-safe attribution framework on AWS Clean Rooms and Amazon Marketing Cloud that matches pseudonymized customer tokens against Amazon shopping and media signals without either side exposing raw data, with Amazon DSP driving the display, video, and streaming campaigns. The reported results: a 10x+ offline sales multiplier and 78% of offline Amazon sales new to the Hyundai brand, on a 90%+ match rate that cut duplicated reach 22% for a 2.2x efficiency gain. If you work on clean-room attribution or AMC/DSP integrations, it's a concrete end-to-end example of tying online media to offline conversions.
📬 Feedback
Questions or comments? Have an item for next week's post? Let me know!
Top comments (0)