Off-by-none: Issue #347

December 9, 2025

Werner's final keynote 👋

In our previous issue, we dropped the latest announcements from re:Invent, got an introduction to O11ywashing, and saw Akamai attempting to buy their way into the serverless space. This week, Werner gives his final re:Invent keynote, Strands Agents adds TypeScript support, and Anthropic asks Claude, "How do we IPO?". Plus, we have plenty of serverless and cloud content from the community!

News & Announcements

The beginning of last week gave us a flurry of announcements from re:Invent, but the rest of the week seemed quieter than usual.

We did learn that Strands Agents added a preview of TypeScript support as well as Strands evaluations to test agent correctness. Amazon Bedrock now supports Responses API from OpenAI, which is probably not what you think it is, and Amazon Bedrock added reinforcement fine-tuning that simplifies how developers build smarter, more accurate AI models. A similar feature was also added to Amazon SageMaker AI (read the official blog for a detailed look).

AWS recently announced the general availability of GPU-accelerated vector (k-NN) indexing on Amazon OpenSearch Service, and this post from AWS explains how you can build billion-scale vector databases in under an hour. 🤯

If you're looking for additional re:Invent recaps and reactions, Steef-Jan Wiggers collected some developer vibes regarding the new Durable Functions and Lambda Managed Instances. Plus, Marco Troisi shared his thoughts on why Durable Functions might be the most important addition to the serverless toolbox in years. Ran Isenberg gives his serverless and agentic AI takeaways and Yan Cui breaks down the biggest re:Invent 2025 serverless announcements.

If you want a more official list of announcements, check out AWS's official blog post: Top announcements of AWS re:Invent 2025

The other day Cloudflare went down again, but as they explained, it had to do with patching a vulnerability in React. Despite the recent fire fights, they still managed to make their Python Workers more performant with lower cold starts.

And finally, Anthropic is prepping for an IPO. I want to assume that's a good thing, I'm just not sure we need more wild speculation in the stock market. The Magnificent 7 provides plenty of that.

The next generation of observability won’t be a static dashboard or a chat window, but a non-deterministic UI that reshapes itself around your system’s state. Honeycomb’s Canvas is an early example of this idea, letting AI use what it learns from each query to build a visual trail of evidence you can actually trust. We might be stuck with conversational interfaces for now, but Honeycomb is clearly designing for what comes next. Sponsored

Tutorials

• AWS Lambda Durable Functions: Build Workflows That Last by Eric Johnson
• The Replay Model: How AWS Lambda Durable Functions Actually Work by Eric Johnson
• Testing AWS Lambda Durable Functions in TypeScript by Eric Johnson
• Developing AWS Lambda Durable Functions with AWS SAM by Eric Johnson
• Webhooks on AWS Lambda - Tips & Tricks by James Gilmore
• SQS Fair Queues with AWS EventBridge by Robert Slootjes
• What you need to know about Lambda Managed Instances by Yan Cui
• Designing an Exactly‑Once DynamoDB Stream Lambda DynamoDB Pipeline by Code Green
• Building fullstack Amplify applications with the new AWS MCP by Praneeta Prakash
• Quarkus 3 application on AWS Lambda- Part 9 Measuring Lambda cold and warm starts with GraalVM Native Image and REST API by Vadym Kazulkin
• Durable AWS-Lambda functions by Vinodbokare
• Streaming Agents on AWS - Part 1 and Part 2 by Mike Chambers
• Streaming AI Agent responses with AWS REST Gateway and Lambda Function

Reads

AWS AI Agents: Amazon’s Desperate Bid to Dominate Enterprise AI at re:Invent 2025
This was an interesting article. I think that AWS is definitely playing catch up here, but as the article notes, they already have a massive customer base and the infrastructure to scale these AI workloads.

re:Invent Proves It: DevOps Isn't Dead, Cloud Native Isn't Fading, and Platform Engineering
Nice summary from Alan Shimel here on the balance that was struck at re:Invent this year. It wasn't just Werner's keynote that put developers at ease with the rise of AI, but the chalk talks, breakout sessions, and hallway track that highlighted AI as a developer enabler, not a replacement.

She architects: Bringing unique perspectives to innovative solutions at AWS
Great post from the underrepresented voices in tech. I love how AWS is continuously innovating on ways to elevate and promote women. They're not perfect, but these stories show that a path to equality is possible.

Vector Databases Aren't Enough: Why AI Needs Multi-Modal Memory Architectures
I would hope that the message that Sowjanya Pandruju is delivering is common knowledge, at least amongst practitioners. If you don't know why it matters, you should give it a look.

AWS shows Rust love at re:Invent: 10 times faster than Kotlin, one tenth the latency of Go
I've spent most of my career working with Perl, JavaScript, PHP, Python, Ruby, and ColdFusion (don't judge, it was very promising at the time), but when it comes to low-latency performance with Lambda Functions, it's hard to deny that Rust is king. You don't necessarily need to build everything in Rust, but if you need high performance and low latency, this is hard to beat.

Podcasts, Videos, and more

AWS Weekly Roundup: AWS re:Invent keynote recap, on-demand videos, and more (December 8, 2025)
This post gives you links to all the keynotes and a few sessions along with links to YouTube playlists broken down by topic and segment. I know what I'm doing for the rest of the year. 😉

AWS re:Invent 2025 - Keynote with Dr. Werner Vogels
If you don't have time to watch all those session videos, at least watch Werner's final re:Invent keynote. There were no major product announcements, no marketing hype, just a clear vision of what the future of development looks like and the integral part that humans will continue to play.

AWS re:Invent 2025 - Building Scalable, Self-Orchestrating AI Workflows with A2A and MCP (DEV415)
re:Invent this year heavily leaned into AI agents. This 400-level talk by Allen Helton and Gunnar Grosch gives a detailed overview of how the process works, and the AWS services used to deliver them.

New from AWS

• AWS Elastic Beanstalk now supports Python 3.14 on Amazon Linux 2023
• Announcing Spatial Data Management on AWS to accelerate spatial-data insights
• Amazon Bedrock now supports reinforcement fine-tuning delivering 66% accuracy gains on average over base models
• Amazon Q now can analyze SES email sending
• AWS launches simplified enablement of AWS CloudTrail events in Amazon CloudWatch
• AWS Elastic Beanstalk now supports Node.js 24 on Amazon Linux 2023
• SES Mail Manager is now available in 10 additional AWS Regions, 27 total
• Amazon SES adds VPC support for API endpoints
• Amazon OpenSearch Service now supports automatic semantic enrichment
• Amazon SageMaker HyperPod now supports checkpointless training

Final Thoughts 🤔

I have plenty more thoughts on re:Invent that I hope to find time to organize and share. High-level feedback from me: it was underwhelming. In technology, at least for the enterprise, boring is good. For the innovators, it feels stale and makes them look elsewhere. There was some innovation, some hardening of services, and some long awaited features that are sure to pique a developer's interest. But I think most of us would agree that the "next big thing" will likely not be announced on a re:Invent stage.

My biggest takeaway was from Werner's keynote. He posed a slightly different question than the panicked, "Will AI take my job?" Instead he asked, "Will AI make me obsolete?" His answer was a confident "NO", with the caveat that we must evolve as developers for this to hold true. Human creativity, problem solving, and connection give us a unique perspective that Werner thinks machines won't replace. I agree with him. And in his words, "Keep Calm and Carry On {Coding}".

Thanks for 14 great keynotes, Werner. You'll be hard to replace (though my money is on Claire Liguori). 😉

Until next week,
Jeremy

I hope you enjoyed this newsletter. We're always looking for ideas and feedback to make it better and more inclusive, so please feel free to reach out to me via Bluesky, LinkedIn, X, or email.