Another Crazy Day in AI 3/109

Another Crazy Day in AI: How Businesses Can Save Early Careers and Why They Should

Jan 19
4 min read

Hello, AI Enthusiasts.

Here's another crazy day in AI:

Stopping workforce erosion and why it's urgent
Why OpenAI is focused on self empowerment
What’s driving the buzz around Claude Code
Some AI tools to try out

🎧 Listen to a quick breakdown of today’s stories.

How Businesses Can Save Early Careers and Why They ShouldAnother Crazy Day In AI: The Podcast

0:00

TODAY'S FEATURED ITEM: Preventing a Lost Generation of Workers

A robotic scientist in a classic white coat with 'AI Scientist' on its back stands beside a human scientist with 'Human Scientist' on their coat, looking towards the AI Scientist.

Image Credit: Wowza (created with Ideogram)

What happens to early careers when the ground beneath them keeps changing?

Lisa Stevens, Chief Administrative Officer at Aon Corporation, recently wrote for the World Economic Forum about the pressures facing early-career professionals at a time when workplaces are being rapidly reconfigured. Many young workers are entering the job market after years marked by disrupted education, limited in-person learning, and fewer opportunities to build confidence through hands-on experience. At the same time, organizations are reassessing how work is structured, with entry-level roles often under the most scrutiny. Stevens outlines why this moment matters for employers as much as it does for individuals.

What the article brings into focus

Unemployment among workers aged 22–27 remains higher than the overall workforce, highlighting ongoing challenges at the start of many careers
Entry-level roles are frequently targeted for automation, even though they play a critical role in learning and professional growth
When early-career pathways weaken, long-term talent pipelines and leadership development are often affected
Limited guidance and support can contribute to lower trust in workplace technologies among younger workers
Separating routine tasks from developmental work allows organizations to preserve learning while improving efficiency
Human skills such as analytical thinking, ethical reasoning, collaboration, and accountability continue to matter in evolving workplaces

Entry-level positions have always done double duty. Yes, they fill immediate organizational needs, but they also serve as the place where people figure out how professional life actually works. How do you handle a mistake? What does it mean to collaborate under pressure? When should you ask for help? A lot of this gets learned through repetition and proximity—watching more experienced colleagues, trying things yourself, getting feedback in real time. When these roles get automated or eliminated, that learning has to happen somewhere else. Some organizations will figure out effective alternatives. Others might not realize what's missing until they need someone ready for bigger responsibilities and discover the pipeline isn't as deep as they expected.

The generation entering the workforce now is dealing with something previous generations didn't face in quite the same way. Many had their college years moved online, their internships cancelled, their first jobs started from home. The informal education that used to happen just by being in an office—picking up on how decisions get made, seeing how conflicts get resolved, understanding workplace dynamics—a lot of that simply didn't occur. Now these same workers are looking for positions in a market where the roles that might have provided some of those experiences are being fundamentally rethought. The decisions organizations make about automation, role design, and talent development will play out over years. These choices will affect not just individual careers but the depth and readiness of entire workforces. How deliberately those decisions get made, and whether they account for what comes after the immediate efficiency improvements, remains an open question worth watching.

Read the full article here.

OTHER INTERESTING AI HIGHLIGHTS:

Why OpenAI Is Focused on Self Empowerment

/OpenAI

OpenAI outlines how AI’s real promise lies not just in advancing capability, but in helping more people actually benefit from it. The piece introduces the idea of a “capability overhang,” the growing gap between what AI can do and how much value individuals, businesses, and countries are capturing from it. By focusing on access, accurate information, and self-empowerment, OpenAI argues that everyday users—not just power users—can unlock meaningful productivity and economic opportunity. The message is clear: AI’s impact depends on how widely and thoughtfully it’s explored.

Another Crazy Day in AI: How the World is Really Adopting AI (And Who's Being Left Behind)

Jan 14
4 min read

Hello, AI Enthusiasts.

Here's another crazy day in AI:

New data on worldwide AI adoption patterns
Gemini turns on Personal Intelligence
The simple framework behind great AI output
Some AI tools to try out

🎧 Listen to a quick breakdown of today’s stories.

How the World is Really Adopting AI (And Who's Being Left Behind)Another Crazy Day In AI: The Podcast

0:00

TODAY'S FEATURED ITEM: Microsoft's New Data on Worldwide AI Adoption

Image Credit: Wowza (created with Ideogram)

Which countries are actually using AI—and does yours know where it stands?

Microsoft's latest AI Diffusion Report tracks how artificial intelligence is spreading worldwide, and the findings reveal some unexpected patterns. In a recent Tools and Weapons podcast conversation, Microsoft Vice Chair and President Brad Smith sits down with Juan Lavista Ferres, Chief Data Scientist and Director of the AI for Good Lab, to discuss what their global data shows about AI adoption through 2026. They examine why certain nations have embraced AI faster than others, what’s driving these differences, and what the patterns might mean for economic development in the coming years.

What the numbers show:

Global North countries have reached 24.7% AI adoption while the Global South stands at 14.1%, with the distance between them growing rather than closing
UAE maintains the top spot globally with 60% adoption and continues growing at 4.6% even at high penetration levels
South Korea jumped from 25th to 18th place worldwide after implementing national AI training initiatives and seeing improved Korean language performance in newer models
United States ranks 24th in adoption while leading in AI innovation and infrastructure
DeepSeek, a free and open-weight Chinese AI model, has gained significant users in China, Russia, and multiple African nations
Countries investing in workforce AI training show higher adoption than those concentrating mainly on infrastructure
African countries display growing AI usage, particularly of Chinese models, as the continent's young population expands rapidly

The report presents something worth thinking about: being good at building technology doesn't always mean being good at using it. Countries like UAE and South Korea have gotten more of their populations using AI through focused training programs and making sure the tools work well in their languages. Meanwhile, nations with world-class AI labs sometimes find their own citizens lag behind in actual usage. The data doesn't say one approach is right or wrong, but it does show they produce different results.

What makes this particularly relevant now is the speed at which these patterns are developing. The gap between regions with high and low AI adoption is widening in real time, measurable from one six-month period to the next. Whether you see that as a problem, an opportunity, or just an interesting data point probably depends on where you sit and what you care about. Either way, knowing where countries actually stand seems more useful than assuming we already know.

Watch on YouTube here.

Listen on Apple Podcasts here.

Listen on Spotify here.

OTHER INTERESTING AI HIGHLIGHTS:

Gemini Turns on Personal Intelligence

/Josh Woodward, (VP, Google Labs, Gemini & AI Studio), on Google Blogs — The Keyword

Google is rolling out a new feature that lets Gemini pull from your own Gmail, Photos, YouTube and Search history to give more tailored answers. Instead of acting like a generic chatbot, it can now recall things you’ve done, places you’ve been, and preferences you’ve shown, as long as you choose to connect those apps. In a personal example, Gemini helped identify a minivan’s tire size, license plate, and even suggest road-trip-friendly options by referencing stored emails and photos. Google says the feature is designed to make AI feel more like a helpful assistant that understands your real life, not just the web.

Another Crazy Day in AI: Anthropic Releases Cowork as Claude Code Alternative

Jan 13
4 min read

Hello, AI Enthusiasts.

Here's another crazy day in AI:

Anthropic launches Cowork for non-coding work
How to govern fast-moving AI
Microsoft and LinkedIn release a book on AI and work
Some AI tools to try out

🎧 Listen to a quick breakdown of today’s stories.

Anthropic Releases Cowork as Claude Code AlternativeAnother Crazy Day In AI: The Podcast

0:00

TODAY'S FEATURED ITEM: Claude Cowork Handles Files Without the Coding

Image Credit: Wowza (created with Ideogram)

What would your workday look like if a digital assistant could quietly handle files, drafts, and messy folders while you focused on the bigger picture?

Claude recently introduced a new feature called Cowork, designed to let its assistant move beyond conversation and into hands-on work. Instead of only responding to prompts, Cowork allows Claude to interact directly with selected folders on a user's computer, making it possible to organize files, build spreadsheets, and turn scattered notes into more structured documents. What started as a system aimed at developers is now being made available to a much broader group of people who deal with everyday digital tasks.

Rather than relying on long back-and-forth chats, Cowork is meant to feel more like assigning work to someone. You describe what needs to be done, Claude puts together a plan, and it works through the steps while keeping you updated and asking before making any significant changes. With access to approved folders, connected tools, and even the browser when paired with Chrome, the system is designed to handle much of the behind-the-scenes coordination that usually slows down everyday work.

What the announcement lays out

Cowork can read, edit, create, and organize files inside folders you choose
Tasks are handled with more independence than a typical chat, guided by a plan and regular updates
The same technology behind Claude’s coding tools is being used here for everyday, non-technical work
It can create documents, spreadsheets, and presentations, and pull in web information when paired with Chrome
Users decide what Claude can access and must approve major actions before they happen
There are still risks, including errors and misleading content, which makes clear instructions and oversight important
The feature is being released as a research preview, with more updates and broader platform support planned

Anthropic labels this a research preview, and that framing matters. The company is transparent about current limitations—Claude can misread instructions, particularly when they're ambiguous or layered with context. There's also the documented risk of prompt injection, where Claude might encounter malicious instructions hidden in files or online content that could alter its behavior. These aren't edge cases being mentioned for legal cover. They're known problems in AI agent systems that researchers across the industry are still figuring out. For lower-stakes tasks like organizing personal files, the risks might seem acceptable. For anything involving sensitive work documents or irreplaceable data, those same risks deserve serious consideration.

What Cowork really brings to the table is a different way of interacting with software. Most tools we use require explicit commands for every action—click here, type this, save that. Cowork asks users to describe an outcome and trust the system to figure out the steps. That requires a comfort level with uncertainty that not everyone shares, and reasonably so. Some people will find it freeing to offload tedious file management. Others will feel uneasy about software making judgment calls on their behalf. As this type of feature becomes more common, users will likely develop their own boundaries around what they're willing to delegate and what needs hands-on control. The learning curve goes both ways—users learning what the tool can handle, and companies learning where the technology still falls short.

Check it out here here.

OTHER INTERESTING AI HIGHLIGHTS:

How to Govern Fast-Moving AI

/Amir Banifatemi, (Chief Responsible AI Officer, Cognizant), and Karla Yee Amezaga, (Initiatives Lead, AI and Data Governance, WEF), on World Economic Forum

As AI systems become more autonomous and adaptive, the authors argue that traditional, slow-moving regulation can no longer keep up. Instead of relying on one-time audits, they propose continuous, real-time oversight that can detect risks as systems evolve in the wild. Governments, companies, and international bodies are already experimenting with monitoring tools, policy sandboxes, and shared safety infrastructure. The piece frames agile governance not as red tape, but as a way to make innovation safer, more trusted, and more scalable.