The craziest Google I/O to date! In August at Google DeepMind, it was: Nano Banana (Gemini 2.5 Flash Image)
Gemini Embedding
Veo 3 Fast
Genie 3
Imagen 4 Fast
Gemma 3 270M
Perch 2
Kaggle Game Arena
Gemini API Url Context
AI Studio Builder (UI overhaul, prompt suggestions, GitHub
RESEARCH
-

Google DeepMind August Releases: Gemini, Veo, Imagen and More
By
–
-
Top performers avoid long hours grinding leads mediocrity
By
–
All the research I've seen shows that the top performers — i.e historically globally significant creative geniuses and innovators — have **not** worked long hours. Grinding leads to mediocrity.
-
Creating AI Newsletter Editions: Process and Themes Analysis
By
–
I've only done four of them so the process is still evolving, but mainly I read through every page of https://
simonwillison.net/2025/Aug/ to spot the big themes and write based on that I start by copying over the markdown skeleton from the previous edition -
AGI Timeline Debate: Continuous AI Advancement Beyond 2030
By
–
The realistic alternative to "AGI by 2030" is not "AI goes away" but rather continued advancement in an industry with large-scale economic value (at least by early revenue numbers) and many smart people focused on figuring out new approaches that might work if old ones fail.
-
Model Release Cycles: Dead Ends and New AI Paradigms
By
–
It is possible, especially on X, to get caught up in the excitement (or anxiety) of weekly model releases, judging each as "its so over" or "we are so back," but, in the larger scale, some possible dead ends (pre-training?) and new paradigms explored (reasoners, RL) look normal.
-

Chip Density Growth: Multiple Paths Forward Beyond Current AI Limits
By
–
60 years of exponential growth in chip density was achieved not through one breakthrough or technology, but a series of problems solved and new paradigms explored as old ones hit limits. I don't think current AI has hit a wall, but even if it does, there many paths forward now.
-

AI Impact on Young Worker Employment Levels Study
By
–
Here's a fascinating new paper by @SeyedMH98 and @LichtingerGuy about AI and the workforce. They use different methods and data sources than @econ_b, @RuyuChen
, and I do in our "canaries" paper, but they also find falling levels of employment for young workers in the -

Token Order Prediction Improves Language Model Pre-training
By
–
A better auxiliary LLM pre-training objective just dropped! Predicting the Order of Upcoming Tokens Improves Language Modeling Teach the model to rank which words are likely to come sooner vs later instead of guessing future words makes LLMs better with minimal overhead
-

Deep Learning Uncovering Fundamental Patterns Across Scales
By
–
results which hold across many orders of magnitude of scale and across many decades highlight how deep learning is uncovering something fundamental
-
Microcontroller Language for LLM Token Processing and Training Data
By
–
R&D: While I love coding up new languages just for fun, this project has particular #ML / #LLM research questions in mind… What if there was a language for a 'micro-controller' able to process 99% of tokens? TRAINING: To produce training data, what's the middle ground