Feeling sleepy every time you open a textbook? 😴
Try this:
→ Quiz yourself, don't reread
→ Move for 2 minutes
→ Bright light, cool room
→ Desk, not bed
→ Nap 10–20 min max
But if you're always tired, the real fix is more sleep.
#StudyTips
👀 Imagine typing:
"I want to understand how the universe works."
And getting a structured learning path built around your goal. 🪐
What would YOUR first learning goal be?
👇 Reply below!
🚀 Join the waitlist: https://t.co/ADaY84JfOG
What if your learning goal could become an entire course? 🎯
Tell https://t.co/8P8Dyh3paB what you want to learn.
AI helps turn your goal into a structured, personalized learning journey.
Your goal. Your course. Your pace. 🚀
#LearnRay#AILearning#EdTech
Big goal, no idea where to start?
LearnRay turns your goal into a learning plan: courses in stages, in order, next step marked.
Goal: space scientist. Plan: math + physics, then Python + scientific thinking, then Space Science for Everyone.
Waitlist: https://t.co/ADaY84JfOG
3/ Practice that helps you remember. Assessments with feedback, plus flashcards: try to recall first, then reveal and review. That's active recall.
Ready to learn your way? Join the waitlist → https://t.co/ADaY84JfOG
2/ Explanations your way. Pick your language, simple words or step-by-step, level of detail, analogies, everyday examples. New topics are explained with your choices in mind.
1/ A course from your goal. Tell LearnRay what you want, why, and what you already know. You get a syllabus with chapters and topics, so you always know your first step and the next.
Answers are easy to find. Understanding them is the hard part.
LearnRay builds a course around your goal, explains it your way, and helps you practice.
2 minutes, start to finish 👇
Join the waitlist: https://t.co/ADaY84JfOG
#AILearning#LearnAnything
A https://t.co/8P8Dyh3paB teaser about turning learning intentions into a clear next step.
https://t.co/8P8Dyh3paB is coming: goal-based courses, explanations your way, quizzes and flashcards.
Turn “someday” into a starting point.
Follow for the launch.
#LearnRay#Learning
Ever heard the same explanation again and again… but nothing clicks?
Then someone explains it differently.
💡 Suddenly, it makes sense.
Learning should adapt to how you understand.
That's the idea behind https://t.co/8P8Dyh3paB.
🚀 Coming Soon.
#LearnRay#EdTech#BuildInPublic
Same subject. Different minds.
Why the same way to learn?
What if every course understood your goals, adapted to your preferences, and connected everything you learned?
We're building something different.
🧠 https://t.co/8P8Dyh3paB
🚀 Coming Soon.
#LearnRay#EdTech#AI
I understand someone is paying for it.
But this will hurt the common man. Merchants always pass it on by raising costs indirectly.
I'm not sure about ATM maintenance costs. Shifting that cost here would have helped. ATM usage is at an all-time low.
On top of that, we already pay tax on everything. I still don't understand what are we getting in return. Something like UPI that helps the population should have stayed free, in my opinion.
It doesn't affect me much personally, but not everyone is privileged and can take this hit. The government needs to think about what it actually gives back for all the taxes it collects.
Yandex published interesting research, and it exploits the KV cache to do something super cool. Here's the gist. It is long, but I promise it's super interesting. So, buckle up.
Start with what the KV cache normally does.
When a model generates text, it does not want to recompute everything it already read every single time it produces a new word. So it stores the internal representation of everything so far, and just reuses it. That storage is the KV cache.
Most people, myself included, think of it purely as a speed optimization. Nothing more.
But here's what they did with it...
The KV cache is not just stored data; it is the model's active memory of the conversation. It decides what the model can see and in what order.
And if that is true, then changing how the cache is organized changes what the model is capable of, even if you never touch the model's weights.
So how do you change it.
The trick (as mentioned in the paper) is that a chunk of that stored memory can be kept in one place (shared across), but shown to different readers in a different order.
Imagine two people looking at the same shared notebook, except each person can be handed the pages in a different sequence. Nothing in the notebook is copied or rewritten, only the order each reader sees it in changes. That is the whole mechanism.
Once you have that, a few things become possible that were not possible before.
1. You can run several model instances against the same shared memory, so each one sees what the others are doing as they do it, instead of waiting for a full answer before reacting.
2. One model can play two roles off the same memory: one part keeps thinking, another part talks. The talking part does not wait for the thinking part to finish; it just uses what is ready so far. And the thinking part can see what has already been said, so it can adjust as it goes.
Yandex showed a demo of an agent that played Doom. The model gets a constant stream of game frames while it is reasoning and while it is choosing actions, all happening over the same shared memory setup.
Nobody trained this model on Doom. The only thing that was built was the way its memory is organized and exposed to it, and that alone was enough to make it act in real time.
The way I would summarize it is this.
Interactivity did not come from retraining a model to be interactive. Some of it can come from changing how a model's own memory is arranged and shared, which is a much cheaper lever to pull than most people assume.
I am still digging deeper, but yes, I spent a good 90 minutes reading this and trying to make sense of what was written, letting my brain get cooked in the process.
Hope this helps.
I have a genuine question.
Anthropic has probably been working on Fable 5.1 for a very long time. OpenAI has probably been working on GPT-6 Astra for just as long. These labs are almost certainly building in parallel, not reacting to each other in real time of course.
So why does Fable 5.1 launch just a couple of days before Astra, and then Astra beats it on almost every benchmark by a good margin?
And this isn't a one-off.
Every single launch follows the same pattern: when Anthropic ships a new model, it beats all of OpenAI's previous models.
When OpenAI ships a new model, it beats all of Anthropic's previous models. Now they are actually beating what the other just launched days earlier.
Benchmarks are completely vanity metrics.