3-paragraph email at work today. Hit Outlook's summarize button for fun. Summary was the same length as the email. Said I saved 6 minutes. Somebody's counting wrong.
Danielle Heberling
@danielleheberling.xyz
A person trying to leave things better than I found them. 😀 Appalachian born, Cascadia based. 🌲 Software Engineer and AWS Hero. 💻
Open-weight models aren't just about cheaper tokens. @danielleheberling.xyz explains how NVIDIA's Nemotron 3.5 Lightning could fit into a routed AI stack alongside larger frontier models. Yells at Cloud Ep. 2: yellsatcloudpod.com/episodes/ep-2
I'm not super into Country music or Southern US culture...but the one exception to that was Dolly Parton. She was highly successful, made a lot of money, and used her seat at the table to lift others in the communities she came from up. The world needs more people like her. RIP.
Check out episode 2. We had more yelling to do. 😀
Yells at Cloud@yellsatcloudpod.com · last wk.Episode 2 is live. AI agents escape containment. Grok Bot wants to be your coworker. AWS runs out of room. Amazon builds a gas-powered AI data center. Twitch says nobody would opt in. The Slow Collapse of Civilization. yellsatcloudpod.com/episodes/ep-2
I use AI to build, and I build things that use AI. A job-search agent and CloudFormation troubleshooter are two public examples. I wrote about why I think those are different kinds of work: danielleheberling.xyz/blog/types-o...
Not All AI Builders Are Doing the Same Work | Danielle's Blog
An opinion on the current landscape of what being a builder looks like in the ever evolving age of AI.
danielleheberling.xyz
My EventBridge Scheduler invokes all "failed" into the DLQ while the agent succeeded every time. Cause was a ~30s timeout I couldn't find documented anywhere. Fix, code, and the async ack pattern on the blog 🛠️
EventBridge Scheduler DLQ'd My Successful Invokes | Danielle's Blog
EventBridge Scheduler kept DLQing successful invokeAgentRuntime calls. The cause was a ~30 second timeout I couldn't find in the docs. Here's the async AgentCore fix.
danielleheberling.xyz
At this rate...I wish I drove to Seattle this week. 🫠 First they switch my train to a much less comfortable bus to get there...now my train terminates in VAN with no shuttle bus to PDX (per the message I just received). 👎
Amtrak Cascades@amtrakcascades.wsdot.wa.gov · 4w agoOperations this week: Due to the warehouse fire north of the PDX station, Trains 506 & 517 will go to Vancouver, WA. Amtrak is trying to source buses for the other SEA-PDX trains except 500 & 511 which are canceled this week. All trains b/t SEA-VAC (BC) and PDX-EUG are all operating normally. 💚🤎🤍
All of these photos and videos don’t do justice how terrifying it is to be in the midst. My thoughts are with folks evacuating from the wildfires. Be safe. Get out. Wear a mask. 😷
Christopher Webb@cwebbonline.com · last mo.Terrifying. I hope everyone is okay.
There's always a nonzero chance something breaks after a deploy. That doesn't mean don't ship. Set up rollback plans and real monitoring first, then press the button. Ship, watch, be ready to roll back. That's the job.
Episode 1 of Yells at Cloud is live. Lambda MicroVMs, CloudFormation Express Mode's lies, Cloudflare temp accounts, and why multi-agent teams underperform their smartest member. Generally Available on Spotify, Apple Podcasts & YouTube. yellsatcloudpod.com/episodes/ep-1
This is happening July 29th...great opportunity for folks!
Annie Sexton@anniesexton.com · 2mo agoPORTLAND TECH LADIES, FEMS & THEYS: I'm running a workshop with @brittanyellich.com called "Putting Yourself Out There" for folx who want to give talks, write blog posts, make videos, but feel a little intimidated. Free ticket to Chain React included!!! June 29th, 9am-12pm ti.to/chainreact/c...
I'm bad at bragging about myself, so review season is rough. That's why I built brag-doc. Log what you shipped in Claude Code, a couple minutes a day. Just facts. When reviews hit, it drafts a promotion case or self review from your entries. github.com/deeheber/brag-doc
GitHub - deeheber/brag-doc: An engineering brag doc that writes itself — log daily wins in Claude Code, then generate promotion and review drafts.
An engineering brag doc that writes itself — log daily wins in Claude Code, then generate promotion and review drafts. - deeheber/brag-doc
github.com
Finally upgraded my Pop!_OS machine to COSMIC and thus far it's pretty great! 💛
HAPPY JUNETEENTH!!!! ✊🏾❤️🖤💚 🎥: @blairimani.bsky.social
This was a good read and something for all of us to consider. tombedor.dev/human-attent...
If You are Asking for Human Attention, Demonstrate Human Effort | Tom Bedor's Blog
An ever-increasing volume of debug investigations, document writing, and code is written by robots. This has created a new etiquette question when working with a team - when is it OK to forward the…
tombedor.dev
It's fun that Claude uses cute terms like "belt and suspenders" for what I would just call "overengineered". 🙃
Call me a purist but something about the Stanley Cup Finals teams current temperature being 78 and 96 degrees Fahrenheit is bizarre. Back in my day hockey was played in places that actually had ice! 🏒
Writing code got cheap. Everything around it didn't. Some recent thoughts written down in public.
Signal and Noise | Danielle's Blog
The AI hype is loud. Here's what I think actually matters underneath it.
danielleheberling.xyz
Scale to zero. 🙌 aws.amazon.com/blogs/aws/in...
Introducing the next generation of Amazon OpenSearch Serverless for building your agentic AI applications | Amazon Web Services
AWS rebuilt Amazon OpenSearch Serverless from the ground up for agentic AI and dynamic workloads. Get instant autoscaling and up to 60% cost savings.
aws.amazon.com
Running Claude Code on your laptop is easy. Getting it onto shared compute your whole team can use is the hard part. So I built a not-pretty-but-working tool for it that investigates broken deploys. 😅 Wrote up how:
Getting Claude Code off my laptop and onto shared compute | Danielle's Blog
How I got Claude Code onto shared compute so broken deploys don't wait on one person.
danielleheberling.xyz
"We don't control the whole system; it's not clear anybody does. But we are not powerless. We can still help hold things steady long enough for something wise and good to take shape." www.nosmallendeavor.com/the-human-co...
The Human Cost of AI: What Is It All For?
We’re building smarter, faster tools every day, but are they helping us live better lives or just accelerating us in the wrong direction?
nosmallendeavor.com
AI's already pretty good at writing code. In my opinion, the dev work that'll matter most isn't using AI, it's building the harnesses that make AI tools actually useful for a specific situation. That's what separates devs of the future from the rest.