🧠 Gemini Files, Stripe Agents, and ChatGPT’s Goblin Phase

 🔥🔥🔥 Three Highly Curated AI Updates

1. Gemini can now generate files (finally)

Google just rolled out the ability to generate files directly inside Gemini. That means you can now create things like documents and structured outputs without jumping between tools.

On paper, this is useful. In practice, it feels like catch-up. This is the kind of functionality users have expected for a while, especially as AI tools move from “chat” to actual work environments.

This signals where Gemini is heading. Less assistant, more workspace. Google clearly wants Gemini to sit closer to your daily workflow, not just answer questions.

But the timing stands out. Competitors have already trained users to expect this level of utility. So this isn’t a leap forward. It’s Google closing a gap.

2. Stripe launches a wallet for AI agents

Stripe introduced a new “Link wallet for agents,” which lets AI agents spend money on your behalf, with approvals and security built in.

The immediate reaction online was loud for a reason. This is more than a feature. It’s infrastructure for a new kind of marketplace where agents can transact, buy services, and interact economically.

If agents can spend, they become economic participants. That opens the door for entirely new business models. Think tools, APIs, or services built specifically for agents to use and pay for. Right now, that marketplace is basically empty. Which is exactly why people are excited. Early builders could shape what “agent-native businesses” actually look like.

This isn’t guaranteed gold. But it is a very real shift in how software might get bought and used going forward.

3. ChatGPT 5.5 has… a goblin problem

OpenAI shared a quirky deep dive into how ChatGPT sometimes produces strange, goblin-like outputs and artifacts. Yes, goblins.

This stems from unexpected behaviors that show up when models generate content, especially in more creative or edge-case scenarios. Think odd phrasing, strange imagery, or outputs that feel slightly… off.

Like this interesting interaction OpenAI’s Chief Scientist had with GPT‑5.5:

As models get more powerful, these edge cases become more visible. And understanding them is important if you’re using AI for anything serious. Even advanced models still have quirks. You just notice them more now because expectations are higher.

Also, let’s be honest. “Goblins in the model” is a much better way to explain weird outputs than a long technical breakdown.

https://openai.com/index/where-the-goblins-came-from

🔥🔥 Two Cool Tools From Google

1. 👗 Google Photos AI Wardrobe

Google is testing an AI try-on feature inside Google Photos that lets you visualize outfits on yourself using your own images.

It’s simple but powerful. Instead of guessing how something might look, you can actually see it. This could change how people shop online, especially for apparel where fit and styling matter.

If this rolls out widely, it brings AI directly into everyday decisions, not just work tasks.

2. 🎨 Google Flow (better way to use Nano Banana)

If you’ve been using Gemini’s Nano Banana image model, there’s a better way to control outputs. It’s called Google Flow.

The big difference is control. You can upload multiple reference images, tag exactly what you’re referring to, and control things like aspect ratio more precisely.

This makes it much better for consistent outputs like infographics, brand visuals, or repeatable creative assets. Same model, but way more usable for real work.

🔥 One prompt you should try today

Optimise your desk setup for maximum productivity

Here’s how to do it:

1. Take a clear photo of your current desk setup.

2. Upload the image into ChatGPT.

3. Paste the prompt below exactly as is.

4. Review the visual audit with suggested fixes.

5. Implement the top 2–3 changes for immediate improvement.

Here’s the creator who inspired this:

Here’s the prompt used:

Create a visual-first, editorial-style infographic auditing the desk setup in the attached photo. Show a side-by-side of current vs. optimized setup with annotations on monitor height, chair position, lighting, cable management, and clutter. Rate each issue with symbols like Top Fix, (working), (acceptable), (costing energy), and (actively hurting), tying each to a consequence like back pain, eye strain, or focus loss. Rank fixes by impact and group them into free fixes, under $50, and worth the investment. Include a Focus Forecast gauge predicting daily deep work hours possible with the current setup vs. after the top 3 fixes. Keep it clean, minimal text, no paragraphs.