Jaana Dogan ヤナ ドガン (@rakyll) 's Twitter Profile
Jaana Dogan ヤナ ドガン

@rakyll

Principal Engineer at Google. Simplicity and optimism. Personal opinions.

ID: 10809412

linkhttps://github.com/rakyll calendar_today03-12-2007 13:08:02

14,14K Tweet

122,122K Followers

1,1K Following

Jaclyn Konzelmann (@jacalulu) 's Twitter Profile Photo

Excited to finally share what my team has been busy working on - an updated Project Mariner that allows you to multitask, supporting up to 10 simultaneous tasks, and a new Teach and Repeat feature. #GoogleIO

Excited to finally share what my team has been busy working on - an updated Project Mariner that allows you to multitask, supporting up to 10 simultaneous tasks, and a new Teach and Repeat feature. #GoogleIO
Edouard Leurent (@eleurent) 's Twitter Profile Photo

Excited to share what I've been up to: Gemini Diffusion is FAST! I'm convinced this will revolutionise iterative workflows: refine, get instant feedback, repeat! So proud of what our small team achieved here🪐

Jaana Dogan ヤナ ドガン (@rakyll) 's Twitter Profile Photo

It remains hard to build a sticky product in the AI landscape. It still makes so much sense to invest in relatively thin frontends. Models are nowhere close to slowing down in the next few years.

Jack Rae (@jack_w_rae) 's Twitter Profile Photo

The Gemini Diffusion release feels like a landmark moment. For text generation, autoregressive models have always outperformed diffusion models from a quality perspective. It wasn't clear that the gap could ever be closed. The team behind this have kept laser focused, broken

MBZ (@babaeizadeh) 's Twitter Profile Photo

#Veo3 further blurs the lines between reality and imagination with audio, stronger text adherence, and richer visual details.

Colaboratory (@googlecolab) 's Twitter Profile Photo

Introducing our reimagined AI-First Colab 📓 The fully agentic experience features a Gemini 2.5 Flash-powered collaborator and enhanced key capabilities: 🔸 Effortless code transformation 🔸 Next-gen data science agent 🔸 Flexible interaction 🔸 Iterative querying 🔸 And more

Jaana Dogan ヤナ ドガン (@rakyll) 's Twitter Profile Photo

I didn’t understand how serious celiac is until lately. Two days ago, I got trace levels of exposure from a shared kitchen. Today, I spent the entire evening with chills, extreme gastrointestinal distress, and breakouts all over my skin. I’m exhausted.

Quinn Slack (@sqs) 's Twitter Profile Photo

You have MPF (Model-Product Fit) when your product: • gets better when the models get better • uses (SOTA) models to the best of their abilities • isn't overfit to today's model capabilities and can evolve Like Product-Market Fit (PMF), you know it when you have it.

Steren (@steren) 's Twitter Profile Photo

The decision of making Cloud Run accept multiple concurrent invocation per instance by default wasn't an easy one. Back in the day, serverless meant Lambda, so deviating from that model was a risk. It also had obvious big consequences on revenues and margins.

Jaana Dogan ヤナ ドガン (@rakyll) 's Twitter Profile Photo

If I have to stack rank all of my employment experiences, the current one would top it all. Once you have the privilege of working with really good engineers who can move mountains, you feel on the top of the world. Otherwise, this industry isn't really worth the hustle anymore.