Listen to today’s podcast: https://www.youtube.com/channel/UC-nqwUyvLDEvs7bV985k-gQ
AI Daily Podcast 12/22/2025
Welcome back to the show. Today’s podcast episode was created from the following stories: a mix of capability breakthroughs, product updates that put users in control, and policy moves that could shape how AI rolls out in 2026 and beyond.
Measuring AI ability to complete long tasks
Read the full story | By Thomas Kwa, Ben West, Joel Becker, Amy Deng, Katharyn Garcia, Max Hasin, Sami Jawhar, Megan Kinniment, Nate Rush, Sydney Von Arx, Ryan Bloom, Thomas Broadley, Haoxing Du, Brian Goodrich, Nikola Jurkovic, Luke Harold Miles, Seraphina Nix, Tao Lin, Neev Parikh, David Rein, Lucas Jun Koba Sato, Hjalmar Wijk, Daniel M. Ziegler, Elizabeth Barnes, Lawrence Chan | 2025-03-19
METR proposes measuring AI agents by the longest human-time tasks they can complete with a given reliability, finding a consistent exponential trend with a roughly seven‑month doubling time. Today’s frontier models reliably handle tasks measured in minutes, but extrapolation suggests week‑ to month‑long autonomous projects could be feasible within a few years. The work offers a forecasting framework with real‑world relevance for benchmarks, deployment planning, and risk management.
Untitled (Yahoo Germany)
Visit the source | By N/A | Date: N/A
The dataset includes a link to Yahoo Germany, but no title, author, date, or article text was provided. Follow the source to see the latest updates directly from Yahoo’s German portal.
Firefox is becoming an AI‑powered browser, but Mozilla says every new feature will come with a full off switch for privacy‑minded users
Read the full story | By Adam Hales | 2025-12-21
Mozilla’s new leadership is steering Firefox toward an AI‑powered future while promising a comprehensive “kill switch” that lets users disable all AI features. Unlike some rivals, AI will be opt‑in, preserving Firefox’s identity around privacy and user control. The strategy tests whether a browser can embrace AI without compromising trust.
New York state just put itself on a legal collision course with Trump’s AI policy
Read the full story | By Mike Pearl | 2025-12-21
New York’s RAISE Act requires large AI companies to formalize safety procedures and report issues within 72 hours—stricter than California’s approach. The law directly challenges President Trump’s executive order asserting federal primacy over AI policy and creating a DOJ task force to contest state rules. Legal experts question the EO’s constitutional footing, setting up a high‑stakes preemption fight that could determine how AI is governed in the U.S.
Instagram’s new AI tool lets you control your algorithm
Read the full story | By Kurt Knutsson, CyberGuy Report | 2025-12-21T07:14:00-05:00
Instagram’s “Your Algorithm” reveals and lets you tune the topics shaping your Reels in real time, rolling out first in the U.S. Users can dial interests up or down and even share their settings, with similar controls headed to Explore. The result is a feed that’s more intentional, less random, and better aligned with what you care about now.
Elon Musk promet l’intelligence artificielle générale en 2026 (il l’avait promis en 2025)
Read the full story | By Olivier | 2025-12-21T15:02:00
Elon Musk now targets 2026 for xAI to achieve AGI, after previously saying it would arrive in 2025. The piece underscores the slippery definitions around AGI and how ambitious timelines help sustain attention and investment. Funding momentum appears strong for xAI, but skepticism persists until concrete capabilities materialize.
Collibra CEO describes what he looks for in AI‑first employees
Read the full story | By Brent D. Griffiths | 2025-12-21T10:51:01.219Z
CEO Felix Van de Maele says it’s a red flag when candidates haven’t experimented with AI, expecting an AI‑first mindset tailored to each role (engineers, for example, using agents like Cursor). Collibra positions itself as an independent data layer that provides the context agents need, while urging enterprises to avoid single‑model lock‑in. The goal is pragmatic AI adoption that changes how work gets done while staying flexible as models evolve.
