
Poolside ships Laguna S 2.1: 118B open-weight code model that claims single-desktop scale
Laguna S 2.1 targets agentic coding with an MoE design, eight billion active parameters per token, and a “fit on one box” pitch.

Curating from trusted global sources…
121 briefings · “tor”

Laguna S 2.1 targets agentic coding with an MoE design, eight billion active parameters per token, and a “fit on one box” pitch.

As Pixar sheds staff, the Toy Story 5 box office run pushes the question: why cut now?

Creator says it is “on Ruby on Rails,” with Jacob Hair attached, and writing can run alongside the series.

A new detector scans posts, notes, replies, and comments, helping readers gauge AI-written or AI-assisted text.

A spacecraft en route to Psyche captures craters, polar ice caps, and deserts, expanding what investors and operators can track.

Substack is teaming with Pangram to add AI detection, shifting the trust model for creators and complicating platform risk management.

A 13-minute “full test AI film” from Blomkamp’s Barley Studios draws harsh viewer criticism and reignites debate on AI in cinema.

A new build pipeline through 2033 may consume power on par with India, forcing operators to plan for scarcity now.

New research pins fertility suppression on olfactory signaling, rewriting a decades-old idea about naked mole-rat hierarchy.

Summer Farmen turned an ALK-positive diagnosis into a six-year survival case study on targeted therapy and patient power.

Open-weight AI is not open-source economics, and that mismatch is crushing model makers while powering cloud operators.

French lawmakers approve the senator by a narrow margin, sparking fears the rights watchdog will miss its mission.