If you strip away the noise, this week tells a simple story: AI is moving from “impressive” to useful. We’re seeing models designed for agents, evidence of real-world usage, and a growing pushback against low-effort AI content.

This edition highlights those shifts and where they are starting to show up in practice.

Let’s dive in.

NVIDIA introduced Nemotron 3, a family of open models built specifically for multi-agent systems.

It’s a bet on coordination, planning, and reliability, the things agentic systems actually struggle with in production.

Why this matters:

As agent-based systems mature, the organisations shaping how models reason and interact will play an increasingly important role.

A new study found six leading AI models passed every level of the Chartered Financial Analyst (CFA) exams, with Gemini 3.0 Pro scoring 97.6% on Level 1.
This reflects progress in structured reasoning, long-form decision-making, and domain understanding under exam conditions.

Passing exams doesn’t replace professionals.
But it absolutely reshapes:

Perplexity AI and Harvard University analysed how users actually use AI agents inside the Comet browser.
Agents are mostly used for cognitive work:

Usage tends to focus on assisted thinking rather than full automation.

Reality check: Agents are commonly used as thinking companions within existing workflows.

Merriam-Webster named “Slop” its 2025 Word of the Year. This is the backlash phase. AI made content easier to create. It also increased the importance of taste, intent, and quality.

If you publish anything online, this should make you pause.

OpenAI released GPT Image 1.5:

~4× faster generation, Much better text rendering Consistency across edits and much better than its previous version. Performance varies by use case. In my experience, Gemini 3 still performs better in some creative scenarios.

This update is about art and practical visuals: slides, marketing assets, product mockups.

NotebookLM lets you upload documents and turn it into a podcast.

Additionally it helps you understand, organise, and generate content from your personal notes and uploaded documents.

It works as a personalised thinking partner, allowing you to explore your own documents through questions, summaries, and structured outputs grounded in your sources.

If PDFs are part of your job, this is a quiet productivity upgrade.

How to use:

Upload a photo of a child or pet, then use this prompt with Nano Banana.

Prompt:

Create a warm, festive Christmas portrait featuring Santa Claus with the uploaded subject. Santa should appear gentle, friendly, and comforting, sitting beside the subject in a natural, non-overpowering way. Santa should be smiling and relaxed. Style: cinematic holiday photography, soft warm lighting, shallow depth of field, cozy Christmas environment.

Preserve the subject’s real facial features, proportions texture exactly no cartooning, no over-smoothing. Santa should look realistic, warm, and timeless (classic red suit, soft textures, natural beard). Mood: joyful, cozy, magical but grounded. Photorealistic, high-resolution, professional holiday portrait.