What is a cozy fantasy? In this post, we explore the genre. We include the elements of cozy fantasy – with plenty of examples.

What Is A Cozy Fantasy?

A cozy fantasy is a sub-genre of the fantasy genre. A cozy fantasy has magic, dragons, and fairies. It can even have the undead. But because there is no saving the world, or the universe, required, cozy fantasy is considered ‘low fantasy’.

For starters, they are often described as having:

1. Low Stakes – This doesn’t mean there are no stakes, or that the stakes are choices between plain croissants or chocolate ones. The stakes are high as far as the characters are concerned. Sir Terry Pratchett’s Discworld series is a great example. No one is trying to save the world. None of the characters make your spine curl. They are endearing even in their criminality. Humans, dwarves, trolls, werewolves, magicians, purveyors of fine foods otherwise known as Rat-onna-stick, and vampires (who are on the wagon), all dwell in Ankh-Morpork. And while there’s literally nothing cozy about Ankh-Morpork itself, it quickly becomes familiar, a cozy place for readers to escape to for a few hours while curled up on the couch with a hot chocolate. The Harry Potter series, on the other hand, has very high stakes. Let the Dark Lord gain power and the world will be in danger. So why is it called a cozy fantasy? Because, like Discworld, The Hobbit, etc., it has the same other following elements. 
2. The Everyman Hero – Bilbo Bagins, Sam Vines, Harry Potter are all everyday, ‘common garden’ people you can meet anywhere. None of them want to be a hero. A comfortable, uneventful life, with lots of tea, books, kindness, and a comfortable armchair. And, as far as Sam Vines is concerned, old boots with very thin soles. That’s all they want.
3. Found Family – Friendship and community are core elements of cozy fantasy. Found family especially so. Harry Potter’s found family are Hermione Granger, Ron Weasley, Albus Dumbledore, and the whole Weasley family. For Sam Vines, it’s the motley gang of coppers of the Night Watch, human or otherwise. For Bilbo Baggins, it’s Frodo, Gandalf, and the Dwarves.
4. Created Homes – In the beginning of The Hobbit, the reader is introduced to the comforts of home. In fact, the second line of the book ends with, ‘it was a hobbit hole and that means comfort.’ Harry Potter finds his ‘home’ at Hogwarts. There are magical feasts, Quidditch, new friends, magic lessons, and safety from his aunt and uncle’s cruelty. Sam Vines went from the slum of the Shades, to living in the Night Watch headquarters, and then, after he marries Lady Sybil, he moves into her opulent Ramkin Residence. All of these are, or become, refuges for the characters and the readers. Everyone, no matter how old they are secretly wants to go to Hogwarts or live in a hole in the ground with a perfectly round door.

Could They Be Any Cozier?

Yes. In fact, Discworld, Hogwarts, and adventures beyond the Shire may be considered by some to be on the more active side of cozy fantasies. Other books in the genre exchange epic adventures for character journeys, action for a slow-paced, relationship-building, community-rich, positive, feel-good read.

Does that mean bad things don’t happen. On the contrary. The comfort that cozy fantasy offers is that they do, that life is not always easy for the characters, and there are, as in real life, problems that need working through and resolution. Just as readers of Romance know there will be a happy ever after, readers of cozy fantasy enjoy knowing the characters will be okay, recovery can be achieved, magic is gentle, and problems will, in the end, be solved.

The coziest cozy mysteries will have:

1. A fantastical setting
2. A gentle story where not much happens
3. A lack of explicit content, i.e. brutal violence, intense gore, sex or sexual assault
4. A magic system
5. A sense of fun
6. A sense of wonder
7. An enchanted home
8. An emphasis on comforting food and surroundings, and drink
9. A tonally feel-good atmosphere
10. Characters who achieve their dreams
11. Characters who resolve their past traumas and find peace internally and externally
12. Eccentric, quirky, or subtle humour
13. Everyday, yet important struggles – relationships, local mysteries, small business’ problems
14. Great worldbuilding
15. Main characters that don’t have to be human
16. Other worldly elements
17. Plots that are character-driven as opposed to action-driven
18. Self-discovery through unfolding life lessons
19. Settings that are whimsical and atmospheric – charming rural villages, cozy pubs, or bookshops
20. Talking creatures

Five Of The Best Cozy Fantasies

Here are the top five cozy fantasies on Goodreads:

1. A Fellowship Of Bakers And Magic by J. Penner – in which a human’s mouth-watering pastries are entered into a prestigious Elven Baking Battle.
2. A Witch’s Guide To Love And Poison by Aamna Qureshi – in which a garden witch specialising in poisons falls in love with the charmingly handsome son of the apothecary and who is known for creating cures.
3. The Between-Worlds B&B by Amy Mae Baxter – in which a human mistakenly books into a B&B for magical creatures with a surprisingly handsome receptionist.
4. The Shambling Guide To New York City by Mur Lafferty – in which a human writes a travel guide for the undead.
5. The Strangest Criminals by Blake Polden – in which Sally Li grows weird, glowing plants, is the StrangeEats delivery rider, and is The Orchard manager for the town’s crime syndicate.

If you need help creating a setting, buy The Setting Workbook from our shop.

The Last Word

So, if you enjoy fantasy, but would like something a little softer, gentler, where the slaying of dragons, or defeating evil empires isn’t on the agenda, and where raucous taverns give way to cake shops, and conversations, then cozy fantasy may be the genre for you.

[Cozy fantasy is a sub-genre of the broader cozy fiction genre, which includes cozy mystery and cozy horror.]

If you would like to learn how to write a book, sign up for one of the rich and in-depth workbooks and courses that Writers Write offers, and get your book off to a great start.

Source for image: Pixabay

by Elaine Dodge. Author of The Harcourts of Canada series and The Device Hunter, Elaine trained as a graphic designer, then worked in design, advertising, and broadcast television. She now creates content, mostly in written form, including ghost writing business books, for clients across the globe, but would much rather be drafting her books and short stories.

More Posts From Elaine

1. What is Romantasy & Why Is It So Popular?
2. How Much Personal Experience You Need To Write Fiction
3. A Quick Start Guide To Writing Crime Fiction
4. What Can Jane Austen Teach Writers Today?
5. A Quick Start Guide To Writing Dialogue
6. What Is Deus Ex Machina in Storytelling?
7. What Is True Crime & How Do I Write It?
8. How To Write A Paranormal Story
9. What Is Fan Fiction & How Do I Write It?
10. The 6 Pillars Of Young Adult Fiction

Top Tip: Find out more about our workbooks and online courses in our shop.

The post What Is A Cozy Fantasy? appeared first on Writers Write.

This week’s off to a hot start at work, and in the industry as a whole. It’s also earnings season across tech, so we can see who is making what type of real progress with AI.

[blog] Beyond Just Looking: Gemini 3 Now Has Agentic Vision. This is a bigger deal than we realize. Instead of “best guess” image processing, our model now does an agentic loop to truly understand an image.

[article] The Five Skills I Actually Use Every Day as an AI PM (and How You Can Too). Here’s a great challenge to PMs. If you’re not doing these types of activities, you’re going to quickly see people encroaching on your domain.

[blog] High-performance inference meets serverless compute with NVIDIA RTX PRO 6000 on Cloud Run. Truly impressive. Which other serverless stack is offering up to 44 vCPUs and 170+ GiB of RAM per instance and letting you run 70B parameter models on demand?

[blog] Summarizing Too Big for Context with MapReduce and LLMs. Smart approach from Wei here. If you’ve got a ton of input data, you can do a Map-Reduce-style exercise to distill the information.

[article] AWS’s inevitable destiny: becoming the next Lumen. It’s lucrative to be the backbone, but mindshare disappears.

[blog] The Rise of Coding Agent Orchestrators. Agent harnesses and orchestrators are going to have a big year. As will the management layers around them.

[blog] A Javelit Frontend for the Deep Research Agent. Ok, now I know what Javalit is, and why it’s great for building data apps without messing with the frontend.

[blog] Kubernetes Rolling Updates for Reliable Deployments. Solid post about how good Kubernetes has gotten at supporting rolling updates for your workloads.

[blog] LiteRT: The Universal Framework for On-Device AI. I don’t understand this space very well, but I read this to learn more. Getting cross-platform acceleration for AI workloads from a single framework is a good deal.

[article] Elon Musk’s SpaceX has acquired his AI company, xAI. The man goes big and plays to win, that’s for sure. The renewed investments in space are pretty exciting.

[article] Waymo raises $16B to scale robotaxi fleet internationally. Let’s go! Great to see this fantastic engineering get deployed more widely.

Want to get this update sent to you every day? Subscribe to my RSS feed or subscribe via email below:

This is a servicing release of Windows Package Manager v1.12. If you find any bugs or problems, please help us out by filing an issue.

New in v1.12

• MCP server available; run winget mcp for assistance on configuring your client.
• App Installer now uses WinUI 3. The package dependency on WinUI 2 has been replaced by a dependency on the Windows App Runtime 1.8.
• Manifest schema and validation updated to v1.12. This version update adds Font as an InstallerType and NestedInstallerType.
• Font Install, Uninstall, and a winget-fonts source have been added and are non-experimental.

Bug Fixes

• Manifest validation no longer fails using UTF-8 BOM encoding when the schema header is on the first line
• Upgrading a portable package with dev mode disabled will no longer remove the package from the PATH variable.
• Fixed source open failure when there were multiple sources but less than two non-explicit sources.
• Corrected property of Font experimental feature to accurately reflect fonts as the required setting value
• Fixed an issue where App Installer would not update its progress.
• Fixed an issue with opening packages that require elevation in App Installer.
• Fixed an issue that prevented App Installer from launching on older OS builds when the Windows App Runtime is missing.
• Fixed an issue that blocked activation of MSIX apps with OnLaunch updates.
• Fixed an issue where App Installer would fail to open a file if the calling process exited too soon.

Font Support

Font Install and Uninstall via manifest and package source for user and machine scopes has been added.
A sample Font manifest can be found at:
https://github.com/microsoft/winget-pkgs/tree/master/fonts/m/Microsoft/FluentFonts/1.0.0.0

At this time install and removal of fonts is only supported for fonts installed via WinGet Package.

Fonts must either be the Installer or a .zip archive of NestedInstaller fonts.

A new explicit source for fonts has been added "winget-font".
winget search font -s winget-font

This source is not yet accepting public submissions at this time.

Experimental Features

• Experimental support still exists for the 'font' command.

Experimental support for Fonts

The following snippet enables experimental support for fonts via winget settings. The winget font list command will list installed font families and the number of installed font faces.

{
  "$schema" "https://aka.ms/winget-settings.schema.json",
  "experimentalFeatures": {
    "fonts": true
  }
}

The font 'list' command has been updated with a new '--details' feature for an alternate view of the installed fonts.

What's Changed

This release only contains bug fixes for App Installer, and no changes to winget.

Full Changelog: v1.12.460...v1.12.470

AI is ubiquitous in both the consumer and enterprise sectors. Yet few organizations are realizing AI’s full potential. Why? AI agents must make decisions and take actions based on a limited subset of overall data. Result: too much guesswork, the occasional hallucination, and failure to extract full value from AI.

The downfall of enterprise AI, then, is agents that falter without a comprehensive understanding of data, both customer- and business-derived. Companies need to be able to pivot from simple data ingestion to sophisticated content collection, integration, and curation that enable AI agents to respond accurately and take appropriate actions.

This can only be accomplished by advancing from traditional prompt engineering to context engineering, which combines a 360-degree view of the customer and a complete enterprise view of a dynamically changing business.

Why enterprise AI is data-rich but context-poor

Many companies implementing AI are data-rich. They use large language models (LLMs) that pull data from all over the internet. They have in-house models that access data from customer databases and product documentation libraries.

Their agents access these pools of information and attempt to guide their decisions. Sometimes they get it right. But too often, they take the wrong action or recommend an incorrect response. What is missing is end-to-end context.

Here is a common example: A person wants to buy a car, so before finalizing their purchase, they go on the manufacturer’s website to research the various options. This data is captured in the car maker’s systems, and over the following weeks, AI directs a series of marketing actions to generate interest in the car model. Without full context, the marketing agent doesn’t recognize that the person has already purchased the car.

This breakdown occurs when one system contains the details of a car purchase, another has records on the individual buyer, and a separate application tracks customer engagement details (such as website visits). Robbed of the rich context of data locked inside information silos, AI digital engagement agents only know that someone researched a car. They’ve missed the opportunity to promote extended warranties and maintenance plans.

Far from rare, such examples are all too common in agentic AI. Enterprises may be data-rich but are context-poor.

Key elements for achieving fluid, unified data

For AI to respond contextually, data needs to be fluid, harmonized, and unified. The walls between silos must be removed.

Achieving this requires several key elements:

Data catalog: The data catalog provides a single view of data across systems. This gives apps and AI agents a map of all assets residing in on-premises systems, the cloud, data lakes, and legacy infrastructure.

Data lineage: Consider this a data verification layer. It traces the full journey of data from origin to consumption, showing every change or transformation along the way. Data lineage enables AI agents to know where any piece of data came from, how it was produced, whether it aligns with organizational governance and regulatory compliance policies, whether it is secure and trustworthy, and whether it reflects the most current knowledge.

Connected signals and actions: Apps and AI agents rely on signals from every system to interpret what’s happening and trigger secure, meaningful actions.
Unified data context: There must be a central repository within an agentic AI architecture that collects, synthesizes, harmonizes, and unifies all information. This context interface for apps and AI agents must operate in real time without requiring file copying or data movement. Whether an AI agent is analyzing a trend or processing a product return, it must provide a single, shared, up-to-the-second view of the customer and the business, aligned with all relevant policies.

Enterprise understanding: Apps and AI agents should not have to relearn the business from scratch. They must act in accordance with the definitions, rules, and principles that underlie each portion of the business. If they don’t, they may appear “AI smart” but “corporate stupid.” Why? Deep metadata intelligence in the enterprise is unavailable to customer-facing systems.

Building enterprise understanding for smarter AI

Enterprise context is vital in defining core business entities and their interrelationships. This context encompasses historical records, master data management (of products, suppliers, assets, and more), business rules, regulatory compliance, and organizational workflows. Comprehensive customer and enterprise records must be unified to supply AI agents with a shared data vocabulary that helps them infer the right context for the right situation at the right time.

Case in point: Large enterprises typically include numerous accounts and corporate entities. The names of various entities may be similar, but there are hierarchies, as well as specific rules and tax schemes that apply by geography and industry. In such a complex organizational structure, if names are entered incorrectly or data is assigned to the wrong corporate entity, AI-based errors are practically inevitable.

Why complete context is key to preventing AI errors

Only the complete unification of customer and enterprise metadata and systems can prevent costly errors and keep AI agents and apps supplied with the applicable context. This way, organizations can consolidate all enterprise and customer data and connect related data from multiple sources to transform trusted context into a meaningful story.

Learn more about Data 360 from Salesforce and how it transforms scattered, fragmented enterprise data into one complete view of your business to fuel real-time workflows, better decision making, and more intelligent agents.

The post Unlocking AI’s full potential: Why context is everything appeared first on The New Stack.

OpenAI is betting that coding agents are outgrowing the terminal.

The company launched a macOS desktop app on Monday for its Codex software development agent, which focuses less on the code itself and more on managing multiple software development agents in parallel.

Until now, the main way to interact with Codex was through the terminal, similar to Anthropic’s Claude Code or Google’s Gemini CLI. While the new desktop app still allows developers to drop into an IDE like VS Code or the terminal for hands-on work, OpenAI’s mission with this app is to make agentic coding more accessible to a wider range of users — even those who may never want to touch the code itself.

As OpenAI CEO Sam Altman pointed out during a press briefing ahead of Monday’s launch, over the last few months, software agents have “crossed a threshold of real utility.”

More than a million people already use Codex regularly, he said, but because these agents are now so capable, the team also wanted to look into ways to make them more accessible to a wider audience. 

Credit: OpenAI.

“5.2 [Codex] in particular is a model that many of us have found can do extremely complex things. We realized we started to feel limited by the interface. And so as we have gotten to these incredible reasoning models with this incredible capability, and we’ve been thinking about what we need to do to make it easy for people to get use of that capability. We have made this Codex app, which is a way to work with the model and manage agents as they do these like long-horizon tasks running on your laptop,” Altman explains.

The Codex team also argues that as developers begin trusting their agents with more complex work, an app like Codex is necessary to help them orchestrate and supervise them.

“Existing IDEs and terminal-based tools are not built to support this way of working,” OpenAI argues in its announcement.

Usking skills in the Codex desktop app (credit: OpenAI).

Developers can point the Codex app at existing repositories, and it integrates directly with GitHub for creating pull requests. Multiple agents can run in parallel, and developers can switch back and forth between different tasks just like they would switch between files in an IDE, because the agents all run in separate threads. 

Since Codex supports Git worktrees, multiple agents can work on the same repo in parallel, each using an isolated copy of that code.

Credit: OpenAI.

From vibe coding to software engineering

As Altman noted, one question the company has been asking itself is whether these agents can transition from vibe coding to serious software engineering. Altman believes so.

“I think we’re over the bar for that,” he says. “I think this will be the way that most serious coders do their job in very rapidly from now.”

Codex beyond code

One interesting aspect of Codex is that it isn’t just about writing code. As Anthropic quickly realized with Claude Code, the core agentic loop that powers these tools is also useful for automating other kinds of workflows. Anthropic launched Cowork as a separate app for this use case, but OpenAI notes that Codex users can use Agent Skills to extend Codex to do this. 

With skills, users can “extend Codex beyond code generation to tasks that require gathering and synthesizing information, problem-solving, writing, and more,” the team writes.

That’s not something OpenAI is emphasizing in this release, but given that the company has set its sights on enterprise use cases, we’ll likely hear much more about this in the future.

Availability

Codex is now available to anyone with a ChatGPT account, including, for a limited time, ChatGPT free and Go users. For those on paid ChatGPT Plus, Pro, Business, Enterprise, and Edu plans, OpenAI is doubling the rate limits during this time.

The post OpenAI’s Codex desktop app is all about managing agents appeared first on The New Stack.