Proud to have participated in the GPT4o vs Gemini 1.5 hackathon at SF last weekend with @notionsmith and @sachscode Ryota Yoda. Thanks @thomasschulzz , @AlexReibman and jacob at @solarissociety for the event and @AustinPeirson for the video (our project is at 1:13).
Last week OpenAI released GPT-4o and Google launched Gemini 1.5 Flash+ to the world
We wanted to see what model is better, so we brought together 100+ AI builders in SF
Who do you think won?
New AMD paper builds a small fully open language model Instella.
Instella is a 3B parameter model family where the authors share weights, training code, and data recipe so others can fully reproduce it.
They first train on huge open text, then on more reasoning heavy data, and finally on about 2.3M instruction examples with a preference step that shapes answers toward human choices.
They also make Instella Long, which can read inputs up to 128K tokens using extra long training data from books and papers, and Instella Math, which gets more math datasets plus reinforcement learning that rewards better solution steps.
Across many benchmarks these models beat earlier fully open ones and come close to the best open weight baselines while staying transparent.
----
Paper – arxiv. org/abs/2511.10628
Paper Title: "Instella: Fully Open Language Models with Stellar Performance"
Massive thanks to our early users! Your feedback has been crucial in shaping our roadmap since day one. Excited for what's ahead—2025 is going to be incredible!
#launch#hexus#jointhehive
🌎 We are changing the world of product storytelling, one collateral at a time.
Welcome to Hexus Hive—the central hub for product marketing for all your teams.
Check out - https://t.co/ACg6xakTki
So excited for our upcoming @ProductHunt launch - and what better way to go live other than in collaboration with @rajivayyangar himself! Can't wait to see everyone at the event. I would also love to meet you if you plan to join us - please come say hi!
In the lightning-fast world of AI and tech, staying current can feel like chasing a moving target. This week at @VentureBeat Transform, the "Women in AI" panel was a highlight. We're lucky to have @aparnabsinha as an early supporter. The key takeaway: community is everything.
3/ OpenRabbit - 🏆 Best use of Computer Vision @roboflow
Open sourced Rabbit R1 with local hardware. It can:
- Create music with Suno
- Send emails
- Curate playlists
And more
A better R1 but built in 24 hours
@notionsmith @sachscode @rama41296
Presented our team project "CityHub" in the
@Meta#Llama3Hackathon at @SHACK15sf.
It is an AI-powered conversational assistant that provides personalized information on San Francisco city services, aiding residents and visitors in navigating government websites.
8/ City Hub
Llama3 powered chat agent that lets residents ask questions about city laws and regulations. RAG’s over city documents to inform the answers
Chat with your city
@AravSrinivas I have 2 years of experience in Front-end application development, including optimizing performance of apps on devices.
Love to work around apps that work around AI. Let me know if that works for you.