@dynamicwebpaige.bsky.social
β¨ Keep it simple, make it scale. AI should be about empowering users and building understanding. π©βπ» AI Developer Experience @ Google DeepMind, ex-Github, ex-Google
Reminder: DeepMind's Gemma 3n model is performing about as well as Gemini 1.5 Pro β better, in many cases! β and is only 4B parameters in size.
The best model 6 months ago is now small enough to be run on a laptop. Now play that forward 6 months from now, thinking about the best models of today. π€―
Get ready to enter the simulation...
Genie 3 is a new frontier for world models: its environments remain largely consistent for several minutes, with visual memory extending as far back as 1min. These limitations will only decrease with time.
Welcome to the future.π
deepmind.google/discover/blo...
π Tired of having too many tabs open, but no time to read through all of them?
Just built a Chrome extension using AI Studio's URL Context feature - all you have to do is add an API key, and the extension uses Gemini 2.5 Flash Lite to summarize all of your open tabs.
π github.com/dynamicwebpa...
π The Gemini APIs in Google AI Studio can be used to understand β and also to generate! β many different output modalities:
Video, images, audio, text, code, or all of the above all at once.
If you've only been using the models to generate text and code, you're missing out. π
ai.google.dev
ποΈ More problem sets like this one from @mitocw.bsky.social, please:
"The intent of this first civil engineering design project is to let you get a feeling for [material properties and constraints] -- by going from playing to building and assessing *what* you do and *why*."
Standing room only for the @majorleaguehacking.bsky.social + Google DeepMind + Google Cloud AI Roadshow in San Francisco!
"This is the single greatest moment in history to become a software engineer, or a technologist β to be anyone who is creating digital artifacts. And we're only just beginning."
πΊοΈ New favorite use case for Google AI Studio's Build:
1οΈβ£ Use Gemini w/Search Grounding to identify specific places (ex: dog-friendly parks in SF; pickleball courts that are bookable on weekends; aestheticians that use Dermalogica products; etc.)
2οΈβ£ Then use @openstreetmap.bsky.social to visualize.
Lee Pace is great in *all things*, my kingdom for this man π₯Ή
01.08.2025 17:17 β π 1 π 0 π¬ 0 π 0
somehow I missed that there is a Foundation TV series (!!) and that Lee Pace is in it (!!!), bonus points with the shirtlessness
this is going to be a long night
[...] 15 year archive totaling more than 116M tokens in a single batch. It took just one command and just 39 minutes to process the complete archive, yielding 154,739 distinct stories: an average of 10.5 stories per broadcast.
Most incredibly of all: it cost just $153.80 to do all of this!"
We used the standard off-the-shelf public Gemini 2.5 Flash Thinking model without any modifications. No data was used to train, tune or otherwise contribute to any model: we used Gemini only to create an index of each broadcast.
We make use of Gemini's batch processing to analyze the entire [...]
"In collaboration with the @archive.org TV News Archive we examined 14,749 evening news broadcasts spanning the big three networks from July 2010 to present, asking Gemini 2.5 Flash Thinking to make an index of the stories covered in each broadcast, geography, sentiment, frame, narrative structure.
31.07.2025 23:43 β π 11 π 3 π¬ 1 π 0
π₯³ The @majorleaguehacking.bsky.social + Google DeepMind AI Roadshow is happening tomorrow in SF!
Six hours of learning and experimenting with the latest Gemini and Veo models in Google AI Studio, seeing what others have built, great food and drinks, and even better company.
π lu.ma/SFroadshowML...
Just finished presenting on the Gemini APIs, Google AI Studio, Gemma 3 and 3n, @googlecolab.bsky.social's Data Science Agent, and @googlelabs.bsky.social's Jules to a group of educators and engineers at Stanford.
Thanks for having @developers.google.com and Google Cloud come visit today! π² βοΈ
π Series A founders:
Want to work directly with DeepMind and Google Cloud engineers, have a fast track to support, and understand how others are using AI at their companies?
Apply for the Google for Startups Gemini Founders Forum, a two-day, in-person event in the Bay Area. Apply by Sept. 10th!
obsessed with the prufrock bunny πβ
30.07.2025 08:38 β π 8 π 0 π¬ 0 π 0
β¨ π Grateful to spend this morning and afternoon with ~100 founders from London, as part of the Google for Startups AI sprints!
Projects included everything from healthcare, to finance, to marketing and environmental science - all shipped in minutes using aistudio.google.com + the Gemini APIs.
comparison of the present-state of agentic capabilities in
@github.com Copilot, @sourcegraph.com Cody, @cursor.com.web.brid.gy, Cline, and several others - I just wish it was a constantly refreshing website instead of a PDF π
www.preprints.org/frontend/man...
Interesting comparison between Gemini CLI and Claude Code, across 50+ software development tasks:
"These measurements come from real development teams working on production applications and show that tool choice should depend on your specific use case rather than following blanket recommendations."
ποΈ Grateful for Joyce and @nathanbenaich.bsky.social, who shared the surreal, extremely Kubrickesque EDP headquarters in Lisbon last night!
For anyone else whose physics brain is breaking, the cantilevered bit over the stairs is an ultra-long balcony -- you can actually walk down it:
β€οΈβπ₯Gemini 2.5 to build the plan for migrating a codebase, and Claude Code (or Gemini CLI) to execute.
Pretty soon the concept of "painful and time-consuming migration from one version of an SDK to another" will feel like as ancient of an idea as punchcards:
π So sad to see paperswithcode is discontinued β but grateful that the @hf.co team is, as always, stepping up to support the community!
It was an incredible resource for use cases, common themes in papers, and visualizing how models have improved on evals over time:
huggingface.co/papers/trend...
Google is processing 980 trillion+ monthly tokens across our products and APIs (up from 480T in May) π€―
wen googolplex? π€©β€οΈβπ₯
"...in which an anonymous poet laments that women of his era know much about artifice and fashion, but little about practical matters such as stewing chiles with cheese." π§ π€
www.foodrepublic.com/2017/10/03/h...
From Prompt to Deployed App in Less Than 2 Minutes, by @dynamicwebpaige.bsky.social:
https://dev.to/googleai/from-prompt-to-deployed-app-in-less-than-2-minutes-dh3
#ai #prompting #deploying #google #guides
Very proud to be part of the team at Google DeepMind, give it a listen:
lnkd.in/dEk6qKA8
π Loved this Lex Fridman conversation with Demis!
His excitement about how our latest video generation model, Veo 3, is learning intuitive physics (ex: fluid dynamics) just from watching videos is contagious... like reverse-engineering the physical world, or reality itself.
π Calling all early-stage founders:
DeepMind and Google Cloud are organizing a startup founders forum in November! It's going to be two full days of technical presentations, feedback sessions, prototyping, demos, and insights into product roadmaps.
π Apply below:
blog.google/outreach-ini...
More on both at the links below:
π½οΈ labs.google/fx/tools/flow
π ai.google.dev/gemini-api/d...
