John Ritsema

Excited to Speak at Atlanta Developers' Conference I’ll be speaking at the Atlanta Developers’ Conference on October 11th, 2025

If you're in Atlanta and want to talk Agentic AI and MCP, I'd love to connect with you @ ATLDevCon johnritsema.com/posts/speaki...

Starting today, the way you will build infrastructure automation is AI Native. You'll prompt the agent, and then System Initiative will get to work building your automation for you in a safe, simulated change set. Come try it - it's so fun. www.systeminit.com/blog/ai-nati...

Exploring Bedrock AgentCore: Building an AI Agent Accelerator Reference Implementation This post describes the new Bedrock AgentCore services and tools announced by AWS, and how the author used them to build an AI agent reference implementation. The key features highlighted are session ...

Built an AI agent chatbot with Bedrock #AgentCore preview services - Agentic RAG, runtime session isolation, managed memory, S3 Vectors & observability. Terraform deployment in 5min. Evolution of my year-old chat accelerator.

johnritsema.com/posts/agentc...

This is crazy

This is an awesome feature!

I’ve been using this internally for a few weeks now! Extremely productive 🙌

@simonwillison.net MCP makes sense for dev tools. what are your thoughts on using it for apps? could be kinda cool if llm cli had support for it

Will likely need some sort of hybrid search with embeddings and keywords possibly with reranking. Also thinking about graphrag

My challenge has been finding a good local technique for semantic search on my markdown files which are quite hierarchical

Ooh, this could be big

Haha. Always 😝

Cool. So you’re running in a VM on Mac vs bare metal?

Awesome setup @marcoceppi.bsky.social!

How well does it run on Mac mini m1?

Also does it run on a first gen arm rpi 4?

p.s. - which usb hub is that? Looks cool

Citations - Anthropic

Anthropic’s new citations API is a welcome addition to folks building RAG apps docs.anthropic.com/en/docs/buil...

New release of one my favorite tools

Excited! 😆

This makes you a legend!

Interesting approach for runtime prompt augmentation. Could be useful for inference debugging community.aws/content/2rS5...

Building effective agents A post for developers with advice and workflows for building effective AI agents

Really nice article from @anthropic.com on building agentic systems with some practical advice from the field www.anthropic.com/research/bui...

repo.new and gist.new are two URLs I use frequently that I still don't think people know about

When people ask me why I'm excited about LLMs for programming when the productivity increase might reduce demand for programmers I remind them I've been contributing to open source for 20+ years with the exact goal of making other programmers massively more productive

Beautiful

lol 😂

I will miss Go Time 😥

A huge thanks to Mat, Johnny, Adam, Jarod, et al. For all their effort in doing this for the community. It was an excellent resource

Also bsky.app/profile/aws-...

bsky.app/profile/aws-...

Agreed. For Simon’s LLM CLI example I think most existing customers already have some CLI they use for obtaining temp creds. But I hear you, it’s not good for individuals or novice users. What do you think would be a good solution here?

Hi Corey. Not sure what you mean by “non-obvious that IAM Identity Center is the correct modern path”. You can invoke bedrock with any IAM principal. Looks like Simon’s example creates a user, but you could also assume a role (most companies use some type of SSO). Not sure where IDC comes in

Excited for this one 🎉

@jcfarris.bsky.social hey good to see you in the blue skies 😀 what do you think of reinvent so far?