Tim Messerschmidt

Agents & Developer Experience Ecosystems Lead EMEA at Google. O'Reilly author. Local AI, fine-tuning. Building tool-eval-bench & awesome-local-ai. Views my own.

[Tim Messerschmidt @SeraAndroid]

I couldn't keep track of all the local AI tools flying around anymore. New engines, harnesses, tooling, benchmarks... every week.

So I started awesome-local-ai: a list sorted by what you want to do, not by hype.

The twist: it's built for agents too. Point your coding agent at it and ask it to recommend a setup for your hardware.

It's early and DGX Spark-heavy. Help me fill the gaps:
https://t.co/aO5UiUnmjh

[Tim Messerschmidt @SeraAndroid]

Massive quantum computing breakthrough! Congratulations, @GoogleQuantumAI 👏

[Sundar Pichai @sundarpichai]

Some magical progress towards "magic state" in quantum!

Our @GoogleQuantumAI team has demonstrated "magic state cultivation" on physical hardware for the first time. The result is a 40x error reduction on our Willow processor vs. previous methods, achieving a state-of-the-art 99.99% logical fidelity.

It's another step towards useful, error-corrected quantum computers to help solve the hardest problems. The result is published in Science today, watch the video below to see how it works

[Google Antigravity @antigravity]

We’re rolling out Gemini 4 Argon in Antigravity to paid Google AI Pro subscribers today. Demand has been incredible and we've scaled capacity deliberately to get here. We’ve rolled out a separate weekly-only Argon quota for Pro on top of the existing Gemini bucket for 3.8 Flash so you have the best of Gemini for all your tasks.

[Tim Messerschmidt @SeraAndroid]

Loving Gemma 4? Point your agent at its new documentation and let it help you deploy it!

[Google Gemma @googlegemma]

Welcome to the new Gemma documentation!

We've built a brand new home for developers to explore, experiment, and build.

What’s new?
🌐 Global model selector for customized code snippets
🤖 Better agent support (turn any page into Markdown!)
📊 Data-driven backend for a consistent experience
⚡️ On-device demos running directly in your browser!

Check it out: https://t.co/yx9K4xGH8M

[Tim Messerschmidt @SeraAndroid]

@JustinoMolinaro So many people underappreciate the impact of a good harness!

[Tim Messerschmidt @SeraAndroid]

Part of my weekly paper series.

What if you didn't train the agent, but trained a small model to rewrite its harness?

A 4B "editor" beat its 35B teacher on average across 21 unseen task families. No weight updates at test time.

[Tim Messerschmidt @SeraAndroid]

https://t.co/pmSLhTYTCU

[Tim Messerschmidt @SeraAndroid]

https://t.co/pmSLhTYTCU

[Tim Messerschmidt @SeraAndroid]

@ashxhart Impressive! @MiaAI_lab has been cooking!

[Tim Messerschmidt @SeraAndroid]

@johnny_everson @ashxhart that's what I am looking for. I wondered if there's a good kvm given i have two macbooks and a mac studio. so I'd love to be able to swap between at least two devices if possible.

re mirrorless camera: same! have a sony dslr above my screen sicne covid and love it!

[Tim Messerschmidt @SeraAndroid]

So.. @rseroter - can I have a TPUv7 rack, please? 😁

[SemiAnalysis @SemiAnalysis_]

BREAKING NEWS🚨: Google has submitted its FIRST TPUv7 FP8 vLLM results to our Agentic Inference Benchmark and has demonstrated 1.67x better perf per dollar than even GB300 NVL72 FP4. 🚀

It is amazing to see Google go full steam ahead with externalizing the power of TPUs outside of Google.👇️ (1/4)🧵

[Tim Messerschmidt @SeraAndroid]

@EspenJD It’s not unusual that a finetune also pushes out some knowledge. Challenging to increase across the board. Still fun release!

[Tim Messerschmidt @SeraAndroid]

Try it in your browser, no Coral needed: https://t.co/Rer1P5qAhC

Code (Apache-2.0): https://t.co/ARvygGogH9

[Tim Messerschmidt @SeraAndroid]

Remember Google Coral? Maybe it's in a drawer, or still accelerating Frigate in your homelab. I've got a new job for it.

Reef: an 11.2M-parameter decision model that runs entirely on the Coral. 4.73 on Decision Index, within 7% of a model 24x its size.

Send an alert or a Frigate event plus typed questions, get a calibrated probability for every option. About 30 ms at around 2 W on a Coral, fully on your LAN. The demo below runs in your browser.

Links in the reply.

[Tim Messerschmidt @SeraAndroid]

https://t.co/AEnJuN6jcO

[Tim Messerschmidt @SeraAndroid]

https://t.co/AEnJuN6jcO

[Mia @MiaAI_lab]

That Argon gummy bear video turned out to be a FAKE.

I ran the prompt myself, and the results from Gemini 4 Argon is MUCH better. Opus is still better, but Argon is not bad at all.

Full prompt below.

[Tim Messerschmidt @SeraAndroid]

@BabushkaRobot Thanks for testing using tool-eval-bench and also: 94%!

[Tim Messerschmidt @SeraAndroid]

A great way for people to get their hands on Gemini 4 Argon. Take it for a spin and let us know what you think!

[OpenCode @opencode]

Gemini 4 Argon is now available in OpenCode Go

$60 usage for $10 til Nov 30

[Tim Messerschmidt @SeraAndroid]

@LiminalPanda Haha thanks. Normal as in Shimano Dura Ace? Could do that. I have the SRAM one at home.

[Tim Messerschmidt @SeraAndroid]

I make no secret that I love cycling. So when I saw the name Gemini 4 Argon, I knew exactly what I had to build.

An Argon 18 Nitrogen Pro demo, made with a single prompt: https://t.co/w2UY180MLQ

Can't wait to see what you build with Gemini!

[Tim Messerschmidt @SeraAndroid]

@txqueue The numbers were derived on the same date with the same eval.

[Tim Messerschmidt @SeraAndroid]

Update on yesterday's Gemma 4 E2B decision model.

v2 vs v1 on the same eval:

Accuracy: 0.513 to 0.804
KL from gold: 0.367 to 0.182
Brier: 0.197 to 0.082

For reference, the teacher model scores 0.735, so the student now beats its teacher. A dedicated fine-tuned decider (OpenDecider-small-td) lands at 0.792.

Not free: p50 latency went from 0.17 s to 0.61 s, and calibration (ECE) slipped a little, 0.041 to 0.061.

Small model, one consumer GPU, an afternoon of iterating. If you've been waiting to try fine-tuning, this is your sign.

[Tim Messerschmidt @SeraAndroid]

Fine-tuning my own decision model on my old gaming rig: @googlegemma Gemma 4 E2B + LoRA in @UnslothAI Studio.

16 GB of VRAM, about 12 minutes for the run in total.

If you've never fine-tuned a model, this is your sign. The barrier is a lot lower than you think.

[Tim Messerschmidt @SeraAndroid]

@blue_checkmarc @MiaAI_lab Haha, true. Probably some triathletes, too. They’re great bikes