Cut through AI noise with weekly insights from Japan to future-proof your career or business. We aim to bring you information that is: digestible, actionable, and unique (breaking through the firewall of language).

In this week’s newsletter:

First, what did they actually launch?

Rather than a single product, Google released a coordinated set of education features across its ecosystem that together redefine how learning happens:

It forms a sort of dynamic learning interface, one that adapts content, pacing, and format around the learners needs in real time.

The old model was teacher → content → student.
The new model is teacher + AI system → guided learning paths → student.

AI increasingly handles explanation, repetition, and personalization, while educators focus on discussion, judgment, and application. Teaching shifts from content delivery to learning design and coaching.

With AI handling explanations and remediation, classroom time becomes less about synchronized lectures and more about discussion, problem-solving, collaborative work and applied learning. Or at least it should!!

Rather than expecting long, uninterrupted focus, these systems:

This signals an acceptance that attention fragmentation is structural, and learning systems must adapt accordingly.

Flashcard apps, quiz generators, note tools, and lesson builders are no longer defensible on their own. Platforms like Google Classroom and Gemini are eating up all these functions.

For learning tool makers, differentiation shifts from features to:

As learning happens through guided paths, systems generate ongoing data about understanding, gaps, and progress. Static exams weaken as the primary signal of learning, replaced by continuous assessment embedded in the learning flow.

Its aiming to be a sort of redefinition of how learning interfaces shape behavior, attention, and understanding.

Google is betting that the future of education is:

AI is going to be the infrastructure layer. And the competitive edge will belong to those who design systems that genuinely improve how people think and learn.
It will be interesting to see how the problem of scale is solved with these AI systems, in huge populations where standardized testing and scoring is baked into the consciousness.

A story that stuck with me this week came from Katie Parrott, a writer from the excellent excellent AI newsletter Every.

Despite positive feedback from her editors, Katie carried persistent professional self-doubt. When she was asked to set goals for 2026 based on her past performance, she had access to all the data but didn’t quite know how to interpret what it meant.

So she tried something simple: she uploaded her Q4 performance data into Claude and ChatGPT and asked a question she’d never feel comfortable asking her boss:

What came back surprised her. The AI surfaced patterns she hadn’t been able to see clearly herself:

Eventually, she asked the real question directly: “Does this mean I’m good at my job?”

Both models independently said yes.

This isn’t about asking AI for praise. You can use it like a neutral mirror.

1. Gather your data you already have
2. Ask for analysis, not validation
Start with something simple, like “Here’s X months of performance data. Please analyze my contributions and identify patterns in where I add the most value.”
3. Go deeper
Follow up with questions like:
“What patterns show up in my highest-performing work?”
“Where does the data suggest I should focus more energy?”
“What am I underestimating about my impact?”
4. Ask the uncomfortable question
“Based on this data alone, does this indicate I’m effective at my job?”
5. Challenge your inner critic — with evidence
Run the same analysis through multiple models. Separate facts from the stories you’ve been telling yourself.

I’ve not been a content creator but, I have tried this with a subset of emails and slack messages in the past to analyze with AI how people feel about working with me. It was surprisingly helpful. I think the key is to gather very relevant data for your job, and then ask AI both sides of the question.. get through the default positivity to the light criticism and challenge yourself. Its quite illuminating!

Bears have been in the news in Japan recently - there’s been a surge in bear attacks throughout 2025, making it the deadliest year on record.​

Over 230 people were injured or killed, including 13 fatalities, with nearly 47,000 black bear sightings reported from April to November which is double the previous record.​ Tohoku region prefectures like Akita and Iwate saw the highest incidents, driven by food shortages and booming bear populations pushing into urban areas. Authorities responded with record culls, military aid, and new hunting rules.

So it would be wise to use technology to help! I found this cool AI project, Kumap.
Kumap delivers real-time maps of bear sightings, encounters, and removals across Japan, helping hikers and outdoor enthusiasts stay informed.​

You can:

Do share feedback with them as its in beta! (No affiliation with the maker)

​​A rising Silicon Valley startup, Command Center, is hosting a New Year–themed hackathon! 🌅
​🪩 What is the De-Vibed Hackathon? 🪩 This hackathon goes against the “vibe coding” boom and focuses on competing in code quality and software design skills.​About halfway through the development time, a technical challenge will be announced. In short-term, AI-assisted development, the key differentiator will be how well you can write code that is resilient to change.

​🎍 Co-hosts 🎍
​Tokyo AI (​​​TAI)
​The biggest international AI community in Japan, with 3,000+ members mainly based in Tokyo (engineers, researchers, and technical product managers).

McDonald’s Netherlands foolishly released a fully AI-generated commercial titled "The Most Terrible Time of the Year," which depicted surreal, creepy scenes of Christmas chaos (Santa in traffic jams, cyclists falling in snow). They finally had to pull the ad following a massive social media backlash where users called it "soulless slop." Let’s hope public pressure continues on this type of slop!