Listen to today’s podcast: https://www.youtube.com/channel/UC-nqwUyvLDEvs7bV985k-gQ
AI Daily Podcast — 12/08/2025
Today’s podcast episode was created from the following stories:
Google Antigravity’s rate limits are changing amid ‘incredible’ demand
By Brady Snyder — December 7, 2025
Google is tweaking Antigravity’s usage policies after “incredible” demand: paid AI Pro and Ultra subscribers now get priority access with the most generous limits refreshing every five hours, while free users move to a weekly quota. Importantly, limits are based on “work done” by the AI agent—not requests—so complex tasks consume more quota. Developers can still access Gemini 3 Pro and all features; if you’re running time‑sensitive projects, consider a paid tier to avoid bottlenecks.
Titans + Miras: Helping AI have long-term memory
By Ali Behrouz, Student Researcher, Meisam Razaviyayn, Staff Researcher, and Vahab Mirrokni, VP and Google Fellow, Google Research — December 4, 2025
Google Research unveiled Titans (an architecture) and MIRAS (a framework) that let models adapt at test time, updating a deep neural memory using a “surprise” signal to capture important new information. The approach outperforms leading transformer and linear RNN baselines on long‑context tasks like BABILong and scales to multi‑million‑token contexts while keeping linear-time efficiency. This could reshape how we build fast, long‑memory AI systems for language, genomics, and time‑series.
Refly claims to combine Notion, NotebookLM, Perplexity, and more into one all‑inclusive super productivity app
By Nolen Jonker — December 2, 2025
Refly is an open‑source, AI‑native workspace that merges a canvas of nodes, multithreaded chats, and automation to handle research, writing, and organization in one place. It supports mixed imports (PDFs, docs, links, images), attaches AI tasks to any node, and exports to Markdown/PDF—though access friction and self‑hosting setup can be a hurdle. Power users may find it replaces chunks of Notion, Perplexity, and NotebookLM, especially for complex, parallel workflows.
Amazon’s Trainium3 is the latest to conform to Nvidia’s mold
By Tobias Mann — December 7, 2025
AWS’s Trainium3 UltraServer racks mirror Nvidia’s GB200/GB300 NVL72 design philosophy, pairing modular compute blades and a switched scale‑up fabric (NeuronSwitch) to pool 144 accelerators as a rack‑sized engine. The industry is converging on shared chassis standards (like Nvidia’s MGX) and switch‑based fabrics, with AWS signaling future support for UALink and NVLink Fusion in Trainium4. Expect faster time‑to‑scale, cross‑vendor interoperability, and more uniform AI infrastructure across clouds.
Google Workspace Studio: la nueva forma de crear agentes de IA sin programar
By Natalia Polo — December 7, 2025
Google Workspace Studio lets Business and Enterprise users build no‑code AI agents that automate tasks across Gmail, Drive, and Chat—and connect to external apps like Salesforce and Asana—powered by Gemini 3. Admins can enable access now (rapid release) with broader rollout in early 2026; templates make it easy to set up daily summaries, auto‑filing, and team check‑ins. It’s a practical on‑ramp for organizations to turn repetitive workflows into reliable, context‑aware automations.
TDWI München 2026: Vorträge für die Konferenz zu Data, Analytics und KI gesucht
By Rainald Menge-Sonnentag — December 7, 2025
TDWI München returns June 23–25, 2026, with a call for contributions open through January 26, 2026, spanning 12 tracks from Data Architecture and Governance to a new Industrial Data & AI focus. The program favors real‑world case studies and also supports first‑time speakers with community mentors. If you have a data or AI story with measurable impact, now’s the time to submit.
Evidence that humans now speak in a chatbot‑influenced dialect is getting stronger
By Mike Pearl — December 7, 2025
Researchers and reporters are finding that AI‑style phrasing is bleeding into human speech and writing—from YouTube vocabulary shifts to Reddit moderators struggling to distinguish human posts from AI‑like prose. Even U.K. parliamentary speeches show patterns associated with chatbot output, suggesting cultural “smuggling” of AI idioms into public discourse. The upshot: media literacy and moderation are getting harder just as AI’s stylistic fingerprints become ubiquitous.
