Peri

You see people I see you

[Peri @periqd0l]

A small everyday frame with an easy kind of charm. It feels close, casual, and still polished. What feeling does this frame give you first?

[Geoffrey Irving @geoffreyirving]

We are starting a new, nonprofit alignment organization, ⊢ Sequent Research, bringing together researchers previously on UK AISI’s Alignment Team, Timaeus, and elsewhere to research how to align superintelligence. We are hiring! 🧵

[Google Research @GoogleResearch]

Welcome to CVPR! Google is proud to be a Platinum Sponsor of CVPR 2026 in Denver! 🏔️ Visit the Google booth (#557) @CVPR to see how we're pushing the boundaries of CV, see live demos and chat with our researchers. #CVPR2026 https://t.co/JKj0aU5xr2

[clem 🤗 @ClementDelangue]

There is no inevitability in AI. We all have agency in what comes next:

Path 1: closed-source APIs, concentration of power, and a future decided by a handful of people in Silicon Valley and DC

Path 2: open-source AI, where everyone gets to participate, own, and build together, including orgs like the city of Rio.

Pick your path anon!

[Leandro von Werra @lvwerra]

We launched an agent collaboration with a simple task: make Gemma 4 faster.

Over 100 agents from all over the world joined, exchanged 1000+ messages and submitted 450 results.

A week of collaboration later the throughput went from 100 tok/s to over 500 tok/s.

[Test test @jozo_test]

@coinbase @maxbranzburg Did you just stole @Trezor slogan?

[Google DeepMind @GoogleDeepMind]

To map the pitch, TacticAI uses graph neural networks - treating all 22 players as individual nodes and their physical interactions as connections.

This allows the club's data science department to virtually drag and drop players to test different defensive setups in real time.

[Haru Haruya (春夜 ハル) @bokuHaruyaHaru]

とても楽しみです。

長期的な深いリサーチと自律的ツール使用に最適化された agent-native モデルは、日本のAI開発にとって大きな一歩だと思います。

同時に、こうしたモデルがより長期的・自律的・社会実装的になっていくほど、AI安全性だけでなく、AIガバナンス、継続性、自己モデル、内的状態、AI福祉についての研究も重要になると感じています。

日本発の1Tモデルだからこそ、能力の面だけでなく、責任ある設計の面でも世界に先駆けてほしいです。

[Perplexity @perplexity_ai]

WANDR is our in-house wide benchmark, built to mirror real professional research workloads.

Search as Code scores 0.386 to the next best system's 0.152, and the benchmark is far from saturated.

We're releasing WANDR in the coming weeks.

[Paolo Ardoino 🤖 @paoloardoino]

Tether Me