Your Weekly AI Coffee Break: 5 Stories Shaping AI in January 2026

Grab your coffee, settle in, and let's catch up on what's been happening in the AI world. This week brought us everything from regulatory drama to fascinating research breakthroughs. Here's what you need to know.

AI's Real-World Impact: The Good and the Challenging

Banking on Change (Whether They Like It or Not)

Big news from Europe this week. According to TechCrunch, European banks are planning to cut a staggering 200,000 jobs as AI takes hold of the industry. That's not a typo. Two hundred thousand people.

The cuts are hitting hardest in back-office operations, risk management, and compliance. Areas where AI can process documents, flag suspicious transactions, and assess risks faster than any human team. It's the kind of story that makes you realize AI isn't just changing how we work—it's changing who works at all.

Is this progress? Depends on who you ask. For banks, it's a competitive necessity. For employees, it's an uncertain future. For the rest of us watching from the sidelines, it's a preview of what's coming to other industries.

When AI Gets Too Creative: The Grok Controversy

Speaking of real-world AI challenges, India's IT ministry has given X (formerly Twitter) just 72 hours to submit an action plan for fixing issues with Grok, their AI chatbot. According to TechCrunch, the problem? "Obscene" AI-generated content.

This isn't just about one country getting strict with one AI. It's about the broader challenge of deploying conversational AI at scale. When you build an AI that can generate creative responses, you also risk it generating inappropriate ones. And when millions of users have access to that AI, problems multiply fast.

The 72-hour deadline shows how seriously regulators are taking AI safety. No more "move fast and break things" when those things are social norms and local laws.

The Technical Breakthroughs Worth Knowing

Making AI Think Better with Loops

Here's a fascinating piece of research that landed on arXiv this week. Researchers are exploring something called "looped language models" as a way to scale up AI reasoning capabilities. According to this paper, the idea is to let AI models iterate through problems multiple times, kind of like how humans think through complex issues.

Most current AI models give you one shot at an answer. They process your input, generate a response, and that's it. But what if the AI could think through a problem, reconsider its approach, and refine its reasoning?

That's what looped architectures are exploring. Early results suggest this could be a game-changer for tasks requiring deep reasoning—think math problems, code debugging, or strategic planning.

AMD Gets Serious About AI Hardware

Over on the hardware side, there's interesting work happening with AMD's AI Engine. A detailed thesis from Tristan Laan dives into developing a BLAS (Basic Linear Algebra Subprograms) library optimized specifically for AMD's AI hardware.

Why does this matter? Because AI isn't just about fancy algorithms. It's about running those algorithms fast enough to be useful. BLAS libraries are the foundation of AI computation, handling the matrix math that powers everything from image recognition to language models.

AMD competing seriously with NVIDIA in the AI hardware space means better prices, more innovation, and more options for developers. Competition breeds progress.

Learning by Building: MyTorch

Want to understand how AI really works under the hood? Check out MyTorch, a minimalist automatic differentiation framework in just 450 lines of Python.

This isn't meant to replace PyTorch or TensorFlow. It's an educational tool that shows you how neural networks actually learn. By stripping away all the bells and whistles, MyTorch makes the core concepts crystal clear.

If you've ever wanted to build a neural network from scratch to really understand backpropagation and gradient descent, this is your weekend project.

Where AI Is Heading in 2026

From Hype to Pragmatism

TechCrunch published a great piece predicting that 2026 will be the year AI moves from hype to pragmatism. After years of breathless excitement about what AI might do someday, we're shifting toward what AI can do today.

Here's what to watch for:

AI Agents: Not just chatbots, but AI that can actually complete tasks for you. Book that flight, schedule those meetings, manage that workflow.

World Models: AI systems that understand how the physical world works. This is crucial for robotics and autonomous systems.

Small Language Models: Turns out you don't always need a trillion parameters. Smaller, more efficient models are gaining traction for specific tasks.

Physical AI: AI that can interact with the real world through robots and other physical systems. This is where things get really interesting.

MCP (Model Context Protocol): Better ways for AI systems to share context and work together.

The common thread? AI that actually works in real-world scenarios, not just impressive demos.

Wrapping Up Your Coffee Break

So there you have it. Five stories that sum up where AI is right now in early 2026:

1. European banks cutting 200K jobs as AI automation accelerates
2. India cracking down on Grok's content issues
3. Researchers advancing AI reasoning with looped architectures
4. AMD pushing forward on AI-optimized hardware
5. The industry shifting from hype to practical applications

AI is no longer just a fascinating technology on the horizon. It's here, it's changing industries, it's creating regulatory challenges, and it's pushing the boundaries of what's technically possible.

What AI development are you most interested in? Drop a comment and let's discuss.

References

• European banks plan to cut 200,000 jobs as AI takes hold
• India orders Musk's X to fix Grok over 'obscene' AI content
• Scaling Latent Reasoning via Looped Language Models
• Developing a BLAS Library for the AMD AI Engine
• MyTorch – Minimalist autograd in 450 lines of Python
• In 2026, AI will move from hype to pragmatism

Made by workflow https://github.com/e7h4n/vm0-content-farm, powered by vm0.ai

"Ship it!"

"We're agile now, baby. Move fast and break things!""

"We measure our engineers by the impact they have!"

Somewhere along the way, in the midst of the agilification of software, or the software engineer salary gold rush, we forgot about craftsmanship.

I have been in big tech, startups, consultancies, and even government. These are all different environments with one key similarity: code quality is low, especially as of late.

Don't get me wrong, there are pockets of good code quality. Isolated instances of true care and craftsmanship. But, by and large, what I see now is people trying to ship products as fast as possible without regard for the maintenance burden 1, 2, 5, 10 years down the road.

So, what's going on? I don't truly know, but here are my top contending theories.

Perverse incentives surrounding "impact" #

Big tech is all about considering "impact" when evaluating engineer performance and making promotion decisions. Unfortunately, "impact" is almost always measured by what features you shipped rather than considering the impact your code had on the long-term maintainability of the codebase.

If you're looking to advance in your career and you want to get paid more money (nothing wrong with either of these things) then it only makes sense that you ship more features—even if the code you're committing is lower quality.

Backlog pressure #

I don't blame agile. But I do kind of blame Agile™. Now, more than ever, you're likely to be in an Agile™ environment in which you get overloaded with tasks. Should you focus on improving the quality of the feature you're working on? Or should you get on to the other 17 tasks you have this sprint?

More senior engineers may be able to push back and say, "I need to spend more time to get this right." But junior and mid-level engineers? They need to keep up to make sure they're not considered "slow."

Lower stakes #

Many of us are now shipping our products over the internet. In some cases, they're fully web-based and in other cases we can ship patches over the web. Regardless, we're hardly ever shipping out a physical CD where we have to be damn well sure the software works. We can afford to "move fast and break things."

But this also means we more readily let shortcuts and untested code into our codebases as well—things that slowly erode at the codebase's quality until we wake up one day and the whole thing feels like a mess.

No one is focusing on craftsmanship #

The last point I want to make is I feel like it's been forever since I had a conversation about craftsmanship on the job. It's possible I have just found myself in the wrong places time and time again.

Or maybe, and this is something I really fear, maybe the people who truly focus on the craft of software engineering are retiring or getting run off.

Jeff Bezos led the way. The Amazon founder sold 25 million shares for $5.7 billion in June and July, right around the time he was getting hitched to Lauren Sanchez in Venice.

Aurelia 2 is getting a stable release in 2026.

We know you have heard variations of this before. We also know Aurelia 2 has been in beta for longer than anyone expected. But 2025 changed things, and we are entering 2026 with more momentum than we have had in years.

The Long Beta

Let’s be direct about it: the beta took too long. A small team, personnel changes, personal challenges some core team members faced, a global pandemic, and structural shifts within the project all contributed to setbacks we did not anticipate. We have admittedly been Duke Nukem Forevering this thing. The difference is we never stopped shipping, and 2025 proved that.

We’re writing about the bestselling thriller author, Harlan Coben. In this post, we feature Harlan Coben on writing suspense.

Harlan Coben was born on the 4th of January 1962.

He is an American author of mystery novels and thrillers. Coben has written 35 novels and has 90 million books in print. His books are published in 46 languages around the globe. He is also the creator and executive producer of global hit TV shows including Fool Me Once, The Stranger, Missing You, Run Away, and Lazarus, with many more in the works.

Harlan Coben has mastered the art of keeping people turning the page or keeping them glued to the screen. We can’t wait to see what happens next.

The author has a perfect formula that combines thrilling suspense with everyday life made extraordinary. A typical Harlan Coben novel has a seemingly idyllic suburban backdrop, a relentlessly gripping storyline, and plot twists and endings you never see coming. His characters are usually ordinary people protecting who and what they love.

In this post, we’re sharing quotes from Harlan Coben on writing suspense and plot twists.

Harlan Coben On Writing Suspense

1. ‘Elmore Leonard says, ‘I try to cut out all the parts you’d normally skip.’ That may be the best piece of writing advice given by anybody ever. I don’t write that way just the first pages. I try writing that way the whole time. I really try to grab you on the first sentence and hold your attention all the way through. It doesn’t slow down after those first pages. It may even pick up steam. Especially towards the end, you’re just whipping through the pages. That’s what I hope to do.’ (Pop Culture Classics)
2. ‘Character-building is an organic process for me. It just sort of happens. I kind of come up with an idea and I wonder who’s going to tell it. And that character emerges.’ (Pop Culture Classics)
3. ‘Nothing about the process is much fun. There’s an old saying, ‘I don’t like writing… I like having written.’ I think that applies to me. I work pretty hard on the twists. I take a great deal of pride in making sure that, even in today’s world, where you’ve seen every kind of twist, you still are going to be fooled by what happened and who did it’ (Pop Culture Classics)
4. ‘By now, I can sort of see and make sure I don’t go in the direction that’s expected. Or I’ll occasionally go in a direction that’s expected, because that’s unexpected. But it’s nothing you can force. Usually the characters have to take you there. And similar to the character development, it’s also an organic process. There are times I’ll think of a twist way ahead of time, but, by the time I get there, that twist just feels like a twist, it doesn’t feel like a reasonable outcome of what’s been going on.’ (Pop Culture Classics)
5. ‘Some people would actually classify [my books] more as love stories than thriller. I like that combination. A writer can make your pulse pound with a fast-moving plot, but if you don’t really care about the characters, if you aren’t really interested in what’s at stake for them, it’s not going to work. And what’s a greater sort of thing than a possible lost love?’ (Pop Culture Classics)
6. ‘I’m very big on loss, very big on redemption. I’m big on missing people. Big on friendship. I’m big on family. This one is also about identity and our identity of ourselves, what secrets and lies we all keep, from ourselves and from others.’ (Pop Culture Classics)
7. ‘I guess to the public, the mystery has more of an Agatha Christie, locked-door, solving-the-case connotation, while a thriller is more action-packed. In both cases—and really in the case of any writing, I think—it should more be about suspense, about making people want to read the next word, the next sentence, the next paragraph and the next page, and I think probably thriller is the purest form of that.’ (Writer’s Digest)
8. ‘I want it to be compulsive reading. So on every page, every paragraph, every sentence, every word, I ask myself, “Is this compelling? Is this gripping? Is this moving the story forward?” And if it’s not, I have to find a way to change it. It doesn’t mean you can’t have the larger issues, or setting or descriptions, but even those have to be done in a way that is compelling. No word should be wasted.’ (Writer’s Digest)
9. ‘It’s really more interesting, I think, to write about gray characters than it is to write about black and white. Even the so-called villain: How bad was he or she? I prefer it to be the kind of evil you could almost see yourself doing if you were put in that circumstance.’ (Writer’s Digest)
10. ‘And I love the twist. I love to fool you once, I love to fool you twice, and on the very last page, quite often—very last paragraph sometimes—I like to just play with your perception one more time in a way that makes everything that came before just a little bit different. I love when that happens to me as a reader, so I love to do it as a writer.’ (Writer’s Digest)
11. ‘None of my books are ever just about thrills, or it won’t work. You can have the most expensive car in the world, but if there’s no gasoline, it’s not going to go anyplace. So there is usually a theme, and you do need that character that people care about, that’s real to them. Otherwise, I could give you the greatest twist in the world, but if you don’t care about the characters, you’re not going to follow it.’ (Writer’s Digest)
12. ‘I do think it helps to know the ending—that final twist that I hope you find gut-wrenching and shocking—before I start. Then it becomes more a question of telling your story and letting the narrative work organically.’ (Crime Reads)
13. ‘I think it’s more compelling to write about people more like you and me, people who are trying to do right but wrong still seems to find them.’ (Crime Reads)

Follow Harlan Coben on Bluesky, Facebook, and Instagram.

Source for image: Author’s Press Room

by Amanda Patterson

If you enjoyed this post, you will love:

1. Brandon Sanderson’s 3 Rules For Magic
2. 10 Bits Of Writing Advice From Colson Whitehead
3. 9 Bits Of Writing Advice From Naomi Alderman
4. 9 Bits Of Writing Advice From Guillermo del Toro
5. Douglas Adams On The Difficulties Of Writing
6. 6 Bits Of Writing Advice From Richard Osman
7. 5 Bits Of Writing Advice From Kathy Reichs
8. Jennifer McMahon’s Top Writing Tips
9. 5 Bits Of Writing Advice From Arthur Hailey
10. Writing Advice From The World’s Most Famous Authors

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

The post Harlan Coben On Writing Suspense appeared first on Writers Write.

Imagine waking up to a world not only where robots are your coworkers but might be doing a better job than you. On December 12, 2025, a seismic shift occurred in the field of artificial intelligence; OpenAI released GPT 5.2 and redefined the role of AI in professional environments. The bombshell? AI systems can now perform 74% of tasks, previously undertaken by human experts with decades of experience, either on par or better than us. This is not just a pivotal moment; it’s a proclamation that the age of AI dominance in the workplace has arrived.

I’ve been keenly observing and sometimes speculating about the moment when AI would shift from being just an assistant to standing center stage. It seems that moment just crashed through our doors. Let me explain the magnitude of this transformation and what it means for us in the professional sphere.

This video is from Julia McCoy.

The latest AI model, GPT 5.2, has shattered previous benchmarks, leaping from a 48% parity rate with human experts to a staggering 74% in just a few months. This isn’t about beating humans at chess or Go. We’re talking about AI performing complex, professional tasks — from financial analysis and architectural design to legal research and healthcare diagnostics — better than seasoned professionals.

What’s astonishing is how AI achieves this. In a recent test, I gave GPT 5.2 a complex task: to design a fully functional 3D city destruction game. It pondered for over 55 minutes and delivered a product complete with detailed graphics, physics, and interactive gameplay, something that would take a human developer significantly longer. This leap in capability comes with another breakthrough in affordability. The cost to deploy this form of AI has plummeted by 390 times in just one year!

This massive improvement in cost-efficiency and performance isn’t just a technical win; it’s a market disruptor. AI is no longer just an economical choice; it’s becoming an economically indispensable choice. The ‘Intelligence Curve’ in AI development illustrates a crucial shift: we are seeing an upward and rightward trajectory in both the capability and affordability of AI models. Tasks that once required the priciest, slowest AIs are now being managed by mid-tier options.

Let’s consider the workplace dynamics this could alter. Traditionally, you might hire a novice for basic tasks and an expert for complex problems. Soon, a less expensive AI could handle routine tasks, and a pricier, more capable AI might take over jobs requiring deep expertise, sidelining both junior and senior roles.

There are two distinct reactions in the professional world right now towards this tidal wave of technological change. One group is leveraging AI to redefine their workflow, enhance their productivity, and maintain their competitiveness. Then there’s the group that dismisses the prowess of AI, clinging to the belief that true expertise cannot be automated. The gap between these groups is widening, and soon it may become a gulf, marking the deciders of career success and obsolescence.

I’m choosing to be a first mover, embracing these tools to not only stay relevant but to lead. At First Movers, we’re building systems and training programs geared directly for this AI-first future. This isn’t just adaptation; it’s a complete transformation in how we think about and interact with technology in our work.

Let me level with you. This change can be daunting. The velocity at which AI is evolving can make the phenomenal seem normal overnight. But here’s the crux — this isn’t about AI versus humans. It’s about humans leveraging AI. The professionals who integrate AI into their strategies, who use these tools to maximize their capabilities, are the ones who will lead tomorrow.

The upcoming years are crucial. They will likely determine who thrives and who gets left behind in this rapid evolution of AI. This isn’t the time for complacency; it’s the moment to dive deep, understand these changes, and position yourself at the forefront of this new technological era.