We built an AI tutor because AI was making us worse at actually learning things
Hey Indie Hackers 👋 Prem here, founder of AllCognix.
Before I talk about what we built, I want to share something we noticed in ourselves first.
We were using AI tools every day to learn new things. New frameworks, new concepts, new tech stacks. It felt productive. But around month 3 we realized we couldn't actually explain or retain anything we had "learned." We were getting faster at fetching answers, not faster at understanding things.
That gap bothered us for a long time. So we stopped complaining about it and built something.
So we built AetherSage AI, a tutor at tutor.allcognix.com that guides you through learning instead of just handing you answers. It asks questions back, walks you through the reasoning, tracks your progress across sessions, and auto-generates clean notes so you actually remember what you covered.
We ran a private beta with 5 users over the past few weeks. The feedback pushed us to open it up today.
Who this is for
Students grinding through tough subjects, developers picking up a new stack, and professionals who want to genuinely understand something rather than skim a summary and forget it by tomorrow.
If you try it, three things I'm genuinely curious about:
Does it feel meaningfully different from a regular AI chat or does it still feel the same?
What would actually make you pay for and keep an AI tutor, not just try it once?
What learning problem do you have that nothing has solved well for you yet?
Brutal feedback is worth more to us right now than kind words. We are five users deep and every piece of real feedback shapes what we build next.
Try it here: tutor.allcognix.com
This is a good problem to build around because the pain is not “AI tutor gives bad answers.” It is that AI makes learning feel faster while quietly weakening retention. That is a sharper angle than another chatbot-for-learning pitch.
I’d lean harder into the “understanding over answer-fetching” frame. The product sounds more valuable when it is positioned as a learning retention system: questions back, reasoning steps, progress across sessions, and clean notes that compound over time. That separates it from regular AI chat much better than simply calling it an AI tutor.
One thing I’d watch is the naming. AetherSage AI feels very AI-tool-like and AllCognix sounds more institutional, but neither immediately carries the softer trust layer needed for students, professionals, and long-term learning support. If this becomes a serious guided-learning/wellness-adjacent education brand, Lyriso.com would feel cleaner, calmer, and more memorable.
I share your view that quick AI answers often stop us from truly understanding and retaining the things we learn. Moving toward a tutor that asks questions instead of just giving results is exactly how we fix that learning gap. It is great to see a tool that helps users master the reasoning process rather than just finding a shortcut.
What is the hardest concept you have tried to learn using a standard AI chat?
Thank you @LilyJeon for the thoughtful comment.
It may not have been the hardest problem, but even learning new topics through standard AI chat can be challenging. For example, concepts like agent memory or KV cache were difficult to grasp deeply because conversation history disappears and there are limited options for personalized, structured learning.
I noticed the same issue when my kids tried learning advanced mathematics and physics concepts — standard chat solutions often provided answers but didn’t fully support the learning process or long-term understanding.
I also use AI tools every single day for work, but I haven't really come across many tools designed to help us master the reasoning process. Honestly, I always thought the whole point of using AI came from our desire to just get problems solved faster! 😂
So it brings me to a fundamental question: why does a tool for learning how to reason have to be AI? There are already tons of great books out there on this topic. What's the special 'hook' that makes users absolutely need to use this service over traditional methods?