Bureaucracy is a big motivation killer! What's your top tip for dealing with it?

After 57 hours of Silksong I finally finished it and reached act 3. Yes, in this order! This game is amazing!

Today I'll buy grown up clothes, wish me luck! (to be fair it is only a jacket/blazer)

I have 50 hours of Silksong and I'm enjoying every minute of it! Are you playing it?

As someone that does live demos frequently, I'm sending a virtual hug to the person doing the demo! that is hard!

"you are loved and wanted in London!"

Twitch from now until next week:

The stars aligned today for two legends: 🎮 Silksong. The wait is over. 🦋 🏆 EmbeddingGemma. A tiny new #1 embedding model. 🧠 A small hero in a vast kingdom, and a small model in a vast universe of data. Both are fast and punch way above their weight class 🥊. check 👇: huggingface.co/spaces/bebec...

Google released open-weights of 380M parameter EmbeddingGemma that is small enough to run on less than 200MB of RAM with quantization - Trained on 100+ languages - Flexible embeddings (768 to 128 dims) with Matryoshka - Works with your favorite open tools

Runs on <200M RAM On a device, locally chef's kiss

The stars aligned today for two legends: 🎮 Silksong. The wait is over. 🦋 🏆 EmbeddingGemma. A tiny new #1 embedding model. 🧠 A small hero in a vast kingdom, and a small model in a vast universe of data. Both are fast and punch way above their weight class 🥊. check 👇: huggingface.co/spaces/bebec...

We've just released an amazing Embedding model: EmbeddingGemma, the new best-in-class open embedding model! 🚀 🏆 Top multilingual model on MTEB (<500M) 💾 Runs on <200MB RAM ⚙️ Customizable output for on-device use 🧩 Integrated with your favorite tools developers.googleblog.com/en/introduci...

Google just launched EmbeddingGemma: an efficient, multilingual 308M embedding model that's ready for semantic search & more on just about any hardware, CPU included. Details in 🧵:

LFG!

almost there!!!

How many copies of Silksong will be sold by Sunday? My guess: 4 million

I can’t think of any other time in modern video-game history that a game as highly anticipated as Silksong has arrived with no early reviews, impressions, leaks, guides, videos, or even forum teases. Soon, everyone will be playing Silksong for the first time. It’s unprecedented and kinda beautiful

vLLM breakdown blog post this is an excellent breakdown of how vLLM works (think ollama but for legit production workloads) great reading if you want a deeper understanding of inference www.aleksagordic.com/blog/vllm

Are you also just waiting for Thursday to play Silksong?

I need to follow more UK people and news reporters, any suggestions?

Old Soviet joke for today: A man walks into a newsstand every day, looks around, and leaves. After a long time of this, the owner says “Can I help you find something?” “I’m looking for the obituaries.” “The obituaries are in the back of the newspaper, comrade.” “Not the one I’m looking for.”

the new Gemini 2.5 Flash Image Generation is amazing! I used it to create a album of pictures of my dog in her secret life! (first image is the base for all the others)

Get Google's lightweight Gemma 3 AI model running in minutes! ⏱️ The post walks through deploying on Google Cloud Run—a perfect solution when you need a fast and incredibly easy AI endpoint for your applications. ⚡ Check out the guide: geshan.com.np/blog/2025/06...

I've played around 15 hours of Shinobi and it's fantastic! Highly recommend! share.google/ijzF1Y2R0a5X...

This is amazing! Denise Herzing, the brilliant mind whose research into dolphin communication powers the DolphinGemma project, is on the TIME100 AI list! 🏆 So well-deserved. Great to see DolphinGemma get a shout-out in the article too! 🐬📣 Link: time.com/collections/...

LLMs are not going to replace Developers in the next 5 years at least

Google just shipped Gemini 2.5 Flash Image, aka nano-banana, a SOTA image generation and editing model. aistudio.google.com/prompts/new_...

The new Shinobi game is BEAUTIFUL! If you like metroidvania, you should definitely check it out! (bad day to comeback from holiday, maybe I should have extended a couple of days...)

AI Singapore launched SEA-LION v4, a powerful open-source multimodal model for Southeast Asian languages. It's built on Gemma, a fantastic base model. SEA-LION v4 even supports low-resource languages and dialects, and it's topping the charts on the SEA-HELM benchmark huggingface.co/aisingapore/...

I like NotebookLM very much and now the video overview is available in 80 languages (the same languages available for audio overviews) share.google/KXxPQYUvbU1p...

Hey, this weekend you can generate some videos for free using the Veo 3 model on the Gemini app! 🤯 Those TPUs are doing overtime! 😂 Gemini letting free users generate Veo 3 videos this weekend share.google/k0cxp1CiokU0...

Nice post by the community about the architecture evolution of the Gemma models 🤩 medium.com/@krupagaliya...

I use Google flights for every flight I take The presentation is just on another level and it makes finding the perfect flight (cheapest 😉) super easy Now it has an AI feature to find deals based on more open ended queries, what do you think? share.google/5QRmz7IviP24...

Next step: they will start using Jax + TPU www.theinformation.com/articles/met...

I can't believe I missed a 10 million offer... I guess I won't be a millionaire in this lifetime www.wsj.com/tech/ai/meta...

I'm quite impressed with some new features of the new Pixel 10 phones, specifically magic cue which highlights important information for you while you're trying to do a task on your phone And this is using on device AI with Gemini Nano (which of basically Gemma 3n 🤩) share.google/chyHdl9OJ6zQ...

As a complete amateur of taking pictures, the Pixel 10 camera has two new features that I'm really looking forward to: Camera coach and best takes. Pixel 10 camera features: Auto Best Take, Pro Res Zoom and more. share.google/OpO2VR6WwOA6...

This experience is so cool!

I lived in São Paulo for 18 years—7 of them right next to the MASP museum—and never visited! 🤦‍♀️ After 10 years living in London, I finally paid my due. It was absolutely worth the wait! So beautiful. ✨ What's a landmark in your hometown you've always meant to visit? 👇

Google released the Gemma 3 270M. An instruction following compact model. developers.googleblog.com/en/introduci...

⚡🤯

Here's a fun demo that runs directly in a WebGPU-enabled browser (mainly Chrome variants) using transformers.js: huggingface.co/spaces/webml... Source code for that demo here: huggingface.co/spaces/webml...

Introducing Gemma 3 270M! 🚀 It sets a new standard for instruction-following in compact models, while being extremely efficient for specialized tasks. developers.googleblog.com/en/introduci...

This kind of posts always makes my day: 5 self-hosted LLMs I use for specific tasks share.google/tbdrk2KDSeYq...

Directly accessing your model from a storage bucket can drastically shrink your container image size and load times. Cloud Storage FUSE is the bridge that can link your container to object storage. 🎬 Video: www.youtube.com/shorts/J3KP0... #aidevelopment #cloudcomputing #cloudrun #cloudstorage

Ok, this was a good day!

I love these comics but this one specific is extra funny 🧠❤️!

Genie 3, with a text prompt, can generate dynamic worlds that you can navigate in real time at 24FPS!! The blog post below has many cool examples, take a look. I hope they publish a demo for us to play with soon! 🎮 deepmind.google/discover/blo...

This looks very cool!! Aside from the great Kaggle Benchmarks new feature that you can use to benchmark your models, this event with the chess matches will even feature live commentator and daily recaps! 🤯 I guess I know what I'll watch for the rest of the week!

The Naked Gun is very funny!