No slop · no ads · daily

The stack

4 picks · curated
01
Hacker News txt 421 c · 834 pts

Muse Spark Open

Meta released the weights for Muse Spark Open, a natively multimodal mixture-of-experts model with 816B total parameters (42B active) and a 1M-token context window, built for coding and long-horizon agentic work. Meta calls it the most capable US open-weight model to date and says it runs entirely on-device across two 512GB M5 Ultra Mac Studios via MLX.

02
AP Wire 8h

Carney says US is trying to break Canada as Alberta votes on separation path

Alberta votes Monday on whether to remain a province or instruct its government to begin the legal process toward a future independence referendum. Campaigning in Calgary, Prime Minister Mark Carney warned that outside powers are trying to break Canada apart so they can own it, a pointed nod to Trump's trade war and 51st-state talk, and told Albertans a protest vote is a dangerous bluff.

03
Reuters 3h

Beijing seeks tighter grip on AI exports amid US rivalry

China's commerce ministry proposed export controls on 23 AI technologies, including models, agents and computing clusters, within a broader list of 40 AI, biotech and solar technologies. The rules would restrict overseas deployment of models containing controlled technology but exempt access through APIs, apps and SaaS, so DeepSeek and Alibaba can keep selling abroad while Beijing gains a veto over the underlying transfers.

04
Lobsters txt 6h

Why are we even doing all this?

In a guest post on Terence Tao's blog, theoretical physicist Rikab Gambhir skewers OpenAI's release of hundreds of AI-generated math write-ups, likening the output to Borges's Library of Babel: an endless supply of possibly-true results that nobody understands or checks. Under the satire is a serious claim that the point of math and science is human understanding, not a growing pile of factoids.