@JeffONelson
Account based inUnited States!
About this account
- Account based in
- United States
- Connected via
- United States App Store
X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
Developer Advocate, Analytics + AI @GoogleCloud ☁️ Past life: economics. Philadelphia sports fan
Los Angeles, CA
Joined November 2011
- Tweets1.6K
- Following2K
- Followers2K
- Likes1.5K
Jeff Nelson retweeted
Lots of discussion out there about our next model(!), so I wanted to give an early look as soon as possible. Introducing Gemini 4 Argon!
It shows frontier performance in complex workflows, cyber defense and software engineering. Teams are using it extensively at Google, from coding to quantum computing, great feedback.
Here’s a look at the benchmarks:
I switch between agent tools depending on the task, and Data Agent Kit's MCP tools + skills for Google Cloud data travel with me.
Today it's generally available!
New: BigQuery Graph, Bigtable, serverless spark + Iceberg, dozens of devex upgrades.
Launch blog:
cloud.google.com/blog/topics…
Tested @typesafeai's Jev against BigQuery’s AI.IF and AI.CLASSIFY from 1 to 100k rows (and pushed BQ to 10,000,000). Findings:
- 1-50k rows: Jev via Cloud Run is fastest
- 50k-10M rows: BigQuery optimized flattens model cost
- Cascade (Jev + Gemini): top accuracy at 1/3 the cost
Great options for devs at any scale!
Blog: medium.com/@jeffonelson/jev-…
Jeff Nelson retweeted
Deploying DiffusionGemma-Jev (djev) just got a lot easier. You can now spin up a Jev API-compatible endpoint on Google Cloud Run using a single command.
Performance is solid: ~35-60 ms for single step latency and batch@32 is ~100-123 requests/sec.
It's a straightforward way to experiment without needing your own GPU. Runs at roughly $3/hr and drops to $0 when idle.
Get the code and instructions here: github.com/taeold/djev-run
Connected @typesafeai 's Jev to BigQuery using a Cloud Run remote function.
Every row gets evaluated with Choice and Noul, which turns those nuanced judgment calls about unstructured data into 0–1 probability columns you can filter, aggregate, and join in SQL.
Blog + GitHub repo follow below
Jeff Nelson retweeted
For 20+ years @ShaneLegg and I've discussed AGI’s potential impact on the economy, science & society. With the DeepMind Institute, we're expanding interdisciplinary research on key questions for the AI era. We hope it spurs the discussions needed to get the next steps right: deepmind.google/institute
Not only did Gemini 3.8 Live drop today. We’re also showing it ~live~ at the @theallinpod summit! Stop by - we even have robots from @GoogleDeepMind and @sippeyxp! Lets talk @GoogleStartups 🙌
Day two of @theallinpod Summit with @GoogleStartups and friends @JensenHuang and @satyanadella
Stop by to chat with us!
First day of @theallinpod summit with @GoogleStartups - stop by if you’re here!
Most data + AI demos query a single clean CSV. In reality, you're jumping across multiple cloud products.
Here's a new Codelab and explainer blog on cross-service investigations with the Data Agent Kit.
cloud.google.com/blog/produc…
codelabs.developers.google.c…
2026 Gemini Flash model releases:
3.5 - May 19
3.6 - July 21
3.7 - August 13
3.8 - September 2
Unfortunately I think 3.14 Flash would come sooner than Pi Day 2027
Rapid advancements in our suite of Flash models have been awesome to see (and code with!)
Machine learning is now easier in BigQuery with TabFM.
Think of it like prompt context for tabular data: you pass historical records as in-context examples, and TabFM generates predictions on your target table in a single SQL query. No training required.
Some example -
1. Classification: predict customer churn and renewal status
2. Regression: forecast revenue / ARR expansion values
3. Evaluation: precision, recall, and accuracy metrics
Blog: cloud.google.com/blog/produc… 🔮
Docs: cloud.google.com/bigquery/do…
TimesFM-3 was released today and brings multivariate support to zero-shot time series forecasting.
If you're new to time series foundation models, think of it like an LLM - you input historical time series data, and it outputs the forecast. No training required!
research.google/blog/timesfm…
v3 is exciting because you can also pass external signals (think marketing promotions, holidays, weather) in a single forward pass
My favorite part: it's coming to BigQuery soon! All of the good time series data you've got there can soon be forecasted with a single helper function.
Check it out on GitHub and Hugging Face:
github.com/google-research/t…
huggingface.co/google/timesf…
Got an app idea you want to test this weekend?
Check out how to build & host it using AI Studio and Google Cloud at goo.gle/builders !
