The Idea
Online STEM acquisition is, successful practice, a room of speech videos. Think Khan Academy, Udemy, Coursera, MIT OpenCourseWare, YouTube, etc.
But why videos?
Picture a professor school a people astatine a blackboard. The professor is performing a shape of presentation: they constitute an equation while explaining it, circle a term, tie a plot, past constituent to what matters and continue. Record that presentation, and you person a speech video. Put it connected the internet, and you person the paradigm down today’s online STEM education.
Producing a bully speech video is really hard. But let's opportunity you managed to do it. Two weeks later, you announcement a mistake. Now you person a problem: the video is already finished, and you cannot spell backmost and hole it. In practice, you either time off the video arsenic it is aliases spell backmost into accumulation and make it again.
While reasoning astir this problem, a thought occurred to us.
We constitute codification and vessel it, but later, we find mistakes, hole them, and vessel a caller version.
What if we could do the aforesaid pinch speech videos? What if we wrote speech videos arsenic root codification truthful that they are maintainable?
Lecture videos arsenic code
Start pinch what the coach really does. The coach says something, writes an equation, circles a term, past keeps talking. Each of those actions tin beryllium described arsenic code:
Now ideate location is simply a compiler that takes this root codification and makes a video from it, pinch matter to reside and machine graphics.
Now the speech becomes thing you tin edit and amended complete time, conscionable for illustration software.
The thought of speech videos arsenic codification is already powerful connected its own. But successful the era of AI, you get thing overmuch much powerful.
In the era of AI
Think astir wherever LLMs person had the biggest effect truthful far: package engineering.
Why software? Because package is written arsenic text. LLMs publication and constitute text.
The aforesaid logic applies elsewhere: move thing into text, and an LLM tin commencement moving connected it directly.
Once speech videos go code, they go text.
Every method taxable connected earth
Nobody is going to walk weeks producing a speech video for obscure mathematical theorems, niche engineering methods, investigation papers, etc. There are excessively galore subjects and excessively fewer group willing successful each one.
But if lectures are code, you tin inquire an LLM for a speech connected fundamentally each method taxable connected earth.
Every language
If lectures are code, you tin inquire an LLM to construe the speech into immoderate language. And the consequence is simply a first-class type of the aforesaid lecture, not a dub aliases a subtitle. One speech tin beryllium successful 80+ languages.
AI-native student experience
If lectures are code, the speech itself tin go interactive. You tin region it, inquire a question, and get the reply arsenic a video, generated for you successful existent time.
The lectures themselves tin besides beryllium personalized to you. A people tin travel your syllabus, lucifer your level, walk much clip connected what you struggle with, and skip what you already know.
Where we are
We decided to build a startup astir this idea. Its sanction is Academa.
We built the underlying lectures-as-code technology.
All the videos present are generated successful a azygous changeable by an LLM, but we will support and amended them complete time. They are conscionable code.
You tin besides inquire our AI to make 1 for you.
We are building galore caller things that we are very excited about, and this is only the beginning. We will stock overmuch much complete the coming weeks and months.
If you would for illustration to travel along, subscribe to our newsletter aliases subordinate our Discord server below.
English (US) · 
Indonesian (ID) · 
