Major League Hacking

It's that time of year again: **YCP HACKS 2026**! Mark your calendars for **November 6-8**! More workshops, more prizes, more activities, more coins, more whoopee pies! This year is RETRO themed, so bring your old gear, clothes, and whatever your mom has lying around! Registration is now open at YCPHACKS.IO so get signed up and ready to attend! Tell your friends, tell your roommates, tell your grandma (as long as she's enrolled in school)!

Jev from TypeSafe AI is the most hyped model right now and I love it. At Major League Hacking we put it on a real job: routing the plain-English assertions our eval runner grades to cheap, deterministic code or an LLM judge. With some tuning, Jev made zero mistakes, but we're still using Gemini Flash-Lite in production. Here are the three most surprising things I learned about Jev: 1. "Make no mistakes" effectively worked 🤣 2. Few-shot prompts examples decreased accuracy 🤯 3. Jev + Flash-Lite was an incredibly cost effective combo 💵 Full write-up with the data: https://lnkd.in/eqZV9h3U Content from my lightning talk at AI Tinkerers in NYC yesterday. #MLH #AI #LLM #Gemini #Evals #LearnByDoing

Want better answers from AI? Start by telling it who to be. 🎭 On this episode of the Prompt Pad, Rosendo Pili shows how framing your model's persona instantly improves the quality of its responses.

What if your next national park trip planned itself AND taught you something along the way? 🏞️ Meet ExploreEase, a hackathon project built by Sibi Tiruchirapalli, Kal Pandit, Jay Prasad, and Dev Rijhwani to make the outdoors more engaging, educational, and accessible. The app uses Gemini to generate personalized itinerary recommendations based on what you love, then pairs them with an "expert" algorithm to fine-tune the plan while keeping it fully customizable. Along the way, an AI chatbot helps you prepare for your trip and shares the history and stories behind the landmarks you visit. Need a meal after a long hike? Yelp and Google Maps API integrations highlight the best food spots near the park for every meal of the day. At its core, ExploreEase is a social good project: getting people off their screens, into nature, and caring more about conservation. A great example of builders using tech to bring us closer to the real world.

Join Week 1 of the Hacktoberfest Open-Source AI Challenge! 🌲 This week's theme: Build something with open-weight models or open-source AI at its core that gets people off the screen and into the world. Hiking, gardening, birding, run clubs, fall foliage: if it gets someone outside, it counts. Missed the Weekend Challenge? Every challenge is a fresh start, so everyone competes on equal footing. 🟢 $2,450 in prizes: 17 winners total (1 overall + 16 partner category winners) Not sure where to start? The partner categories are a great jumping-off point. Because winners are drawn from a smaller pool of submissions per category, your odds of taking home a prize are meaningfully higher if you build with one of the featured tools. 🟩 Submissions due October 11 at 11:59 PM PDT, winners announced the week of October 12 Full details here: https://lnkd.in/ew6kN8RN

The AI landscape moves fast. New models, new tools, and new techniques show up almost every week. That's why we're launching the Prompt Pad, a new video learning series hosted by Rosendo Pili. Rosendo brings years of hands-on experience to the screen, sharing the practical tips and tricks that help builders get more out of AI tools. Each episode is focused around something you can put to work right away. Whether you're brand-new to AI or a seasoned developer looking to sharpen your workflow, we think there's something here for everyone. Watch the intro and subscribe to stay in the loop: https://lnkd.in/gDUt_VNy Know someone diving into AI? Share this with them!

New York friends! I'll be in town Sunday Oct. 11 for AI Engineer NYC. Join me at the hackathon we're organizing? It's free and we'll feed you 😊 https://lnkd.in/gZmapK5r

Hacktoberfest is here, and we're kicking it off with the Open-Source AI Weekend Challenge! 🎉 This is the first of five challenges running all month long. Every challenge uses the same prompt, with a new theme revealed each Monday in October. The Prompt: build something with open-source AI at its core. Run an open-weight model, build on an open-source agent harness or framework, run inference locally, or all three. The Theme: Build for a Friend. Pick one person in your life and build something that solves a real problem for them. A meal planner that knows your roommate's allergies, a patient practice partner for a friend learning a new language, a family recipe book made from grandpa's voice memos. It doesn't have to be big. It just has to matter to them. 17 cash prize winners: one overall winner and 16 prize categories from our partners. Submissions due October 5 at 6:59 AM UTC https://lnkd.in/gqT6PRpe