Hey there,

Week 1 of our Hacktoberfest Open-Source AI Challenge is officially live, and this week's theme is Touch Grass! Build something with open-weight models or open-source AI that gets people off the screen and out into the world, whether that's a bird call identifier that works on the trail with no signal, a garden planner tuned to your local frost dates, or a run club route builder that hunts down the best fall foliage. With $2,450 in prizes across 17 winners, including 16 partner categories, there are plenty of ways to win. Submissions are due October 11.

Looking for more ways to celebrate Hacktoberfest? Spend a week hacking with the MLH community at Global Hack Week: Hacktoberfest, packed with daily livestreamed workshops, mini-events in the MLH Discord, and social, technical, and design challenges that earn you points along the way. Kick things off at the Opening Ceremony livestream at 10:00 AM EDT, right on the DEV homepage

There's also just one week left in the Kaggle Benchmarking Challenge. Build a benchmark on Kaggle, run it against real models, and tell us what you found. Five winners will each take home $500, so get your submission in before October 11.

Have a great week ❤️

As always, here are our favorite posts from last week in no particular order:

Building an Offline Arduino UNO Q Cyberdeck That Identifies Birdsong and Draws Vintage Field Notes: a hands-on look at building a privacy-first field journal that listens for birds and sketches what you find.
I Built My First AI Agent With AWS AgentCore, and the Hardest Part Wasn't the AI: lessons learned from a first agent build where permissions and infrastructure turned out to be the real challenge.
How to Build a Real-Time Voice AI Agent with the Gemini Live API: building voice agents that respond in real time and can be interrupted mid-sentence.
TypeScript Compiler API: Preserving Child Node Narrowing in Reusable Type Guards 🔧: exploring how composable type guards can keep both parent and child narrowing intact.
The Physics of Socratic Prompting: Somatic Recoil, Chess Alpha-Beta, & The NLP Meta-Model: one dev's take on steering AI coding agents with questions instead of line-by-line fixes.
Views Measure Views: personal insights from nine years of writing data on what views and followers really measure.
Are Frontend Developers Cooked? Is Frontend design safe?: real experiences putting an AI agent to work on frontend apps, and rethinking where developer taste still matters.

Read Now

Here are some standout wins from last week's discussion:

@fm: finally found peace and closed 100+ browser tabs 🗂️
@bozhidar_valchev: joined DEV and migrated five blog posts, including a study pitting eight LLMs and DeepL against each other on translation 🌐
@earlgreyhot1701d: stepped away from the laptop to explore New Orleans with family 🎷
@yahhi: turned reader feedback into real fixes for their MCP server and published a step-by-step guide to building and shipping your own 🛠️
@anthonymax: jumped into Hacktoberfest 2026 with EmbedCatalog 🎃

Well done, everyone!

Share Your Win

Thanks for Being Awesome! 😎

Have we told you lately how awesome you are? Thanks a ton for all you bring to DEV. Your ideas, your content, your support—it's what makes this community great.

Happy coding!