AI moved fast this week. ChatGPT, Claude, Gemini, NotebookLM, and Google AI Studio all shipped upgrades that are actually worth paying attention to. Not tiny cosmetic changes. Real capability jumps.
The biggest headline is simple: Claude’s new Fable 5 and Sonnet 5 models are pushing the bar higher, Gemini now supports custom MCP connections and smarter automations, ChatGPT quietly added settings and app controls that make a big difference, and NotebookLM is becoming far more useful for content creation and study workflows.
If you use AI for research, coding, automation, design, writing, or content repurposing, these changes matter right now.
Claude just made a serious move with Fable 5 and Sonnet 5
If there is one platform that deserves immediate testing, it is Claude.
The standout release is Fable 5. For public access, this looks like one of the strongest models available right now, especially for tough tasks. It is positioned for:
- Research-heavy work
- Agentic tasks
- Coding
- Complex problem solving
The important catch is timing. Fable 5 was included only for a limited trial window through July 7, after which usage shifts to a credit-based cost. That makes this the kind of model you should test immediately if you have access, especially on a Pro plan.
What makes Fable 5 so interesting is not just raw quality. It is that it appears to handle difficult tasks with the kind of reliability people usually hope for but do not always get from general-purpose models.
Why Sonnet 5 may be the more practical upgrade
Fable 5 is the flashy release, but Sonnet 5 may be the more important everyday model.
Claude’s Sonnet line has typically been the more cost-conscious option. What changed here is that Sonnet 5 is no longer just a lighter general assistant. It is built to work more autonomously. It can:
- Make plans on its own
- Use tools like browsers and terminals
- Handle coding tasks far more effectively
- Perform much closer to higher-end models at a lower price
That matters because it narrows the gap between Sonnet and Opus. In practical terms, Claude is making lower-cost usage far more competitive without forcing people into the most expensive tier for serious work.
If you have been choosing models based on budget, Sonnet 5 may shift that entire calculation.
Claude Design is no longer a side feature
Claude also rolled out a major upgrade to its design experience, and this is one of those updates that can sneak up on people if they are only thinking about text generation.
The new design workflow supports a range of outputs, including:
- Slides
- Website prototypes
- Wireframes
- Documents
- Animations
- Blank creative projects
You can also feed it structured inputs like a design system, a screenshot, a codebase, a Figma file, or other assets. That makes Claude feel less like a chatbot and more like a fast design collaborator.
One especially impressive use case is animation. With a short prompt asking for an animated video around a topic, Claude can generate a polished animated asset in seconds. The kind of work that could have taken a designer hours to assemble can now be mocked up almost instantly.
That does not mean human design skill is obsolete. It means ideation, prototyping, and rough production just got dramatically faster.
If your work touches presentations, visual explainers, mockups, or campaign concepts, this is worth serious testing.
ChatGPT’s new UI and settings unlock more than most people realize
ChatGPT also shipped meaningful upgrades, though some of the most valuable ones are tucked away in the interface and settings.
A redesigned creation and tools menu
The new plus menu has been reorganized to make core actions easier to access. From one place, you can now jump into tasks like:
- Image creation
- Web search
- Deep research
- File and plugin access
- Skills and app-related actions
That may sound like a small UI refresh, but it points to a larger shift. ChatGPT is becoming more operational. It is no longer just a prompt box. It is a hub for tools, files, connectors, and actions.
Settings you should absolutely turn on
There are three settings in particular that deserve attention.
- Connector Search under advanced settings
- Higher Intelligence under general settings
- Enable Dictation under general settings
Connector Search is a big one. When enabled, ChatGPT can search connected apps and use that context to answer questions more effectively.
Higher Intelligence helps ChatGPT automatically step up its reasoning when a question is more complex. That means less manual fiddling and better default behaviour.
Enable Dictation is simple but useful if you think faster than you type or want to capture ideas more naturally.
Low-risk app permissions reduce friction
Inside the apps section, there is also a permission setting for allowing low-risk actions. This is one of those options that can make or break the experience.
If you leave it off, ChatGPT may keep stopping to ask permission every time it tries to perform a minor action. If you turn it on, routine workflows become much smoother.
For anyone using AI as an active assistant instead of a passive chatbot, reducing that friction matters.
ChatGPT’s app ecosystem is getting much bigger
The enabled apps section has expanded heavily. Instead of a tiny list of integrations, there are now many more apps available, including tools tied to communication, business operations, and ad workflows.
Once connected, ChatGPT can draw context from those systems. Gmail is the clearest example. With access to Gmail capabilities, ChatGPT can use email context to answer questions and support tasks more intelligently.
This is where ChatGPT starts acting less like a single app and more like a control layer across your digital tools.
A prompt optimizer that fills a gap the platforms still have
One tool called My Prompt Buddy stands out because it addresses a problem the major AI platforms still have not solved cleanly.
It works as a Chrome extension across Claude, ChatGPT, and Google Gemini. The main value comes from two things:
- Prompt optimization for different task types like standard prompts, reasoning prompts, deep research, AI video, and AI image generation
- Prompt shortcuts so frequently used prompts are always available without retyping
If you use repeating instructions every day, this saves time immediately. More importantly, it improves consistency. Instead of rebuilding a good prompt from memory, you can store and reuse what already works.
The extension is available with a free trial at mypromptbuddy.ai.
For related workflow ideas, it would also make sense to link internally to pages like best AI prompts or how to write better ChatGPT prompts if your site has them.
Gemini’s custom MCP support changes what it can connect to
Gemini may have shipped one of the most strategically important updates of the bunch: custom MCP access.
Inside Gemini’s connected apps area, you can now add a custom MCP endpoint by supplying a URL and credentials. Once that connection is established, Gemini is no longer limited to a narrow built-in integration set. It can potentially connect to far more systems.
That opens the door to much broader workflows, especially for people who want Gemini to interact with custom tools, business systems, or unique data environments.
If MCP-based interoperability continues expanding, this becomes one of the most important upgrades in the AI assistant space.
Recurring tasks and event-based automation in Gemini
Gemini also upgraded its scheduling and task automation capabilities.
You can now create recurring tasks or have Gemini react to updates across:
- Gmail
- Google Calendar
- Google Drive
- The web
- Any MCP-connected source
A practical example is email drafting. If a specific person sends a message, Gemini can prepare a reply draft in Gmail based on your instructions. That turns Gemini into something more than a conversational assistant. It becomes a workflow engine.
For anyone managing repetitive knowledge work, this is a very real productivity upgrade.
Gemini Skills make reusable instructions far easier
Another strong addition is Skills.
Skills let you create reusable instruction sets so Gemini behaves consistently across repeated tasks. Preloaded examples include things like:
- Matching your writing style
- Focusing your energy
- Generating fresh ideas
- Getting more perspectives
- Preparing for meetings
The real value is customisation. If you use Gemini for script writing, research formatting, content planning, or recurring analysis, you can define a repeatable process once and reuse it across topics.
That consistency is huge. It saves time, reduces prompt drift, and helps outputs stay aligned with your voice and standards.
NotebookLM is becoming much more useful for creators and learners
NotebookLM added two upgrades that stand out immediately: video overview shorts and customisable flashcards.
Generate shorts from long-form source material
The video overview feature can now create short-form videos from a source. One particularly smart workflow is to upload a single long-form video as your source, choose the shorts option, and generate multiple short clips around the same topic.
NotebookLM can:
- Pull focus topics from the source material
- Generate animated visual sequences
- Add changing text overlays
- Create graphics and motion elements automatically
And if the first result is not exactly right, you can inspect the prompt used, iterate on it, and refine the outcome. That is especially useful because it makes the output editable at the instruction level rather than forcing you to start from zero.
The key here is specificity. The clearer you are about topic and source, the better the generated short-form content will be.
If you publish content regularly, this is an efficient repurposing workflow.
Suggested image: a screenshot of NotebookLM’s shorts generation interface.
Suggested alt text: NotebookLM video overview shorts tool generating animated short-form content from a long-form source.
Custom flashcards for study and revision
NotebookLM’s flashcard feature also got better. You can now customise:
- Number of cards
- Difficulty level
- Specific topic focus
This makes NotebookLM more practical for studying, training, certification prep, or even internal team onboarding. Instead of generic review cards, you can create targeted study material that fits the exact scope and depth you need.
It is a simple feature, but a good one.
Google AI Studio may be the sleeper upgrade of the week
One of the biggest developments did not come from the mainstream chatbot interfaces at all. It came from Google AI Studio, where you can now build with agents directly inside the platform.
That includes templates or pathways for things like:
- Data analyst agents
- Customer support agents
- Document processors
- Repository maintainers
- AI talk radio style systems
The phrase that matters most is this: you can build agents, tools, and apps for free.
That is a big deal.
For a lot of people, the biggest barrier to experimenting with agents is cost and complexity. AI Studio lowers both. It gives access to models, real-time capabilities, speech and text tools, and a built-in design experience without the heavy overhead that often comes with other platforms.
If you have been curious about building an AI app or workflow but did not want to pay premium tool fees just to test ideas, AI Studio is now much more compelling.
Google Labs hints at what is coming next
Another useful place to keep an eye on is Google Labs. Many AI experiments launched there tend to show up later across Google products.
Examples include projects related to:
- Personalised story collections
- AI design tools
- Marketing agents
- Music generation
- Game world creation
Not every experiment becomes a core product, but Labs is often the early signal for where Google is heading next.
What these AI upgrades mean in practical terms
Stepping back, a clear pattern is emerging across Claude, ChatGPT, Gemini, NotebookLM, and AI Studio.
AI platforms are moving in the same direction:
- Better models
- More tools
- Deeper app integrations
- More autonomous workflows
- More reusable instructions and skills
- Faster creation of media, prototypes, and agents
This is no longer just about asking smarter questions. It is about building systems that can actually do more on your behalf.
If I had to prioritise what to test first, the shortlist would be:
- Claude Fable 5 for high-difficulty work
- Claude Sonnet 5 for coding and autonomous tasks
- Gemini custom MCP access for broader integrations
- Gemini schedules and skills for recurring workflows
- NotebookLM shorts for content repurposing
- Google AI Studio agents for low-cost experimentation
- ChatGPT connector settings and app permissions for smoother daily use
Meta description
New AI upgrades across Gemini, ChatGPT, Claude, NotebookLM, and AI Studio, including Fable 5, Sonnet 5, custom MCPs, agents, and automations.
Suggested categories and tags
Category: AI Tools
Tags: Google Gemini, ChatGPT, Claude, Fable 5, Sonnet 5, NotebookLM, Google AI Studio, MCP, AI agents, AI automation, prompt engineering
FAQ
What is the most important AI model update right now?
Claude Fable 5 stands out as one of the strongest public models right now for research, coding, agentic work, and difficult tasks. Sonnet 5 is also a major upgrade because it delivers near-premium performance at a lower cost.
What does custom MCP access in Gemini do?
Custom MCP access lets Gemini connect to external systems using an MCP URL and credentials. That expands Gemini beyond standard built-in integrations and opens up more advanced workflow and automation options.
Which ChatGPT settings should be enabled?
Three especially useful settings are Connector Search, Higher Intelligence, and Enable Dictation. It also helps to allow low-risk actions in app permissions so ChatGPT can complete routine actions with less friction.
How can NotebookLM help with content repurposing?
NotebookLM can generate shorts from a long-form source, automatically adding text, graphics, and animations. It also allows prompt iteration, which makes it easier to refine short-form outputs without rebuilding from scratch.
Why is Google AI Studio such a big upgrade?
Google AI Studio now allows you to build agents, tools, and apps directly inside the platform, with access to multiple model capabilities and built-in design support. The fact that much of this can be done for free makes it especially significant.
Final thoughts
This wave of upgrades is a reminder that the AI landscape is not just improving incrementally. It is getting more connected, more autonomous, and more practical for real work.
If you want the biggest upside, test the features that reduce repeat effort. Reusable skills, connectors, MCP access, automation rules, design generation, and agent builders are where a lot of the leverage now lives.
If you found this useful, share it with someone who is trying to keep up with the latest AI tools, and explore related articles on prompt engineering, Gemini workflows, and AI agent building.
