Your AI/ML study partner · 🎓 Explain mode · /research reports · /review papers · 📎 CSV data for ▶ Run
/research
/review
Enter your password to continue.
"Everywhere" logs out every phone, laptop and browser where you used NEXUS. You'll need your password again on each.
Learned automatically from your chats. When something changes, the old fact moves to History.
Save answers as notes (🗒️ under any answer). NEXUS also uses your notes when answering. Tick notes to export them together as a study guide.
Topics from alirezadir/AIMLInterviews (MIT). Click ⚪ to mark 🟡 learning → ✅ done. 🎓 = learn it with NEXUS, 🃏 = flashcards.
Real ML interview questions from AIMLInterviews. Answer in chat — NEXUS grades it /10, shows a model answer and asks a follow-up.
Upload notes, PDFs or Word files once. In every chat, NEXUS searches them and uses the parts that match your question.
A skill changes how NEXUS answers. Pick one (it stays on until you turn it off).
Tip: just ask in chat — "remind me tomorrow at 9 to call Rohan".
How many answers each free AI gave today, and roughly how close it is to its free daily limit.
Save the open chat as a file:
🔗 Share this chat with a read-only link (only this chat — nothing else):
📥 Import your ChatGPT history (ChatGPT → Settings → Data controls → Export → conversations.json):
conversations.json
Build an expression from Value numbers and watch backpropagation compute every gradient. Operations: + - * / **, .tanh() .sigmoid() .relu() .exp() .log(). The output is L (or the last result). Engine in the style of micrograd by Andrej Karpathy (MIT).
Value
+ - * / **
.tanh() .sigmoid() .relu() .exp() .log()
L
Arrows go forward (inputs → output). Blue = inputs, orange = output. grad = how much the output changes per unit change of that number.
Numerical gradient = (L(x + h) − L(x − h)) / 2h with h = 10⁻⁶: nudge one input and measure how the output moves.
