Jeremiah Liu

Principal Scientist@MSL on frontier capability. Past Life: Coding@GDM, Uncertainty Quantification, Probabilistic ML & Biostatistics.

[Jeremiah Liu @jereliu]

@grx_xce @Intelligence_ai @IndexVentures @conviction @A_StarVC Woohoo! Congrats Grace and the team! What a milestone!!!

[Addy Osmani @addyosmani]

@rezaul_arif Great work by the team including @jereliu! Antigravity capacity and quotas are being worked on rn.

[Ian Goodfellow @goodfellow_ian]

Amazing test of Gemini 3’s multimodal reasoning capabilities: try generating a threejs voxel art scene using only an image as input

Prompt: I have provided an image. Code a beautiful voxel art scene inspired by this image. Write threejs code as a single-page

[Jeff Dean @JeffDean]

The webdev capabilities of Gemini 3 had some of the biggest leaps from the prior Gemini 2.5: the small but mighty team we had working on these capabilities really did awesome work!

[Jeremiah Liu @jereliu]

(3/3) Massive shoutout to the team!👏 @clarahuiyihu Quan Yuan @qzscgy, @sunjiao123sun_, @suzzzylin, Pulkit Mehta! Building this best full-stack capabilities wasn't a cake walk and took so many first-principles thinking and relentless iteration. Y'all crushed it. :)

[Jeremiah Liu @jereliu]

(1/3) Incredibly proud to see the Gemini 3.0 is landed!
Our team has worked insanely hard to push the frontier LLM's fullstack capability, from tasteful design aesthetics to deep, deslightful functionality. Seeing it top the charts across the board is absolutely surreal:

[Jeremiah Liu @jereliu]

(2/3)🏆 #1 on WebDev Arena (Vibe Coding Agents) https://t.co/Om1Tcb7L9x
🎨 #1 on Design Arena (Website, 3D Design, UI Component, Game Design) https://t.co/c8ytMh3wlA
👀 Powering @Replit and @figma's "non-slop" AI design experience
https://t.co/PNbufRPuGT, https://t.co/qofoMLeKLw

[Amjad Masad @amasad]

Announcing Design, a new Replit experience focused on beautiful UIs.

The first non-slop AI design experience, powered by Gemini 3.0

[Arena.ai @arena]

🚨BREAKING: @GoogleDeepMind’s Gemini-3-Pro is now #1 across all major Arena leaderboards

🥇#1 in Text, Vision, and WebDev - surpassing Grok-4.1, Claude-4.5, and GPT-5
🥇#1 in Coding, Math, Creative Writing, Long Queries, and nearly all occupational leaderboards.

Massive gains over Gemini-2.5:
🔸WebDev in Code Arena: 1487 (+280 pts vs 2.5)
🔸Text: 1501 (+50 pts)
🔸Vision: 1328 (+70 pts)
🔸Arena Expert: Top-3 (just 3 pts behind #1)

Huge congrats to the @GoogleDeepMind team on this breakthrough! 👏

[Abishek⚡ @eter_inquirer]

yo they COOKED with gemini 3

i literally one-shotted this

[Arena.ai @arena]

Gemini 3 Pro shows impressive coding capabilities on the new WebDev leaderboard, powered by Code Arena - also ranking #1, with a massive score jump from Gemini-2.5 (+280 pts)

[Grace Li @grx_xce]

This is the biggest performance delta we’ve seen since launching Design Arena

Gemini 3.0 Pro has taken #1 overall and #1 in 4 of our 5 code arenas - Website, Game Dev, 3D Design, and UI Components

Well-earned congratulations to the @GoogleDeepMind team on a remarkable advancement (and humble version naming)

[Amjad Masad @amasad]

Announcing Design, a new Replit experience focused on beautiful UIs.

The first non-slop AI design experience, powered by Gemini 3.0

[Balaji Lakshminarayanan @balajiln]

Check out our work (w/ @jereliu, @suzzzylin, @shreyaspadhy, @dustinvtran, Tania Bedrax-Weiss) on improving single model uncertainty today at #NeurIPS2020 poster session 6, 09:00 – 11:00 PST

Spectral-normalized Neural Gaussian process (SNGP)
https://t.co/JotOwJZuCh

Thread 1/n:

[Dustin Tran @dustinvtran]

I find it genuinely troubling that every awarded paper is a deep learning paper, and there’s even a need for a “best non-DL paper” award https://t.co/JXXOTBcNes

[Joe Blitzstein @stat110]

P(Harvard beats Yale) = 0.92, based on games played this millennium. #HarvardYale #MaximumLikelihoodEstimate