Semantic search, not keyword matching
Traditional search finds exact words. If you search "order block" you only get results containing "order block." But ICT often explains the same concept using different language across different lectures.
ICT Decoder uses semantic search — it understands the meaning behind your question, not just the words. Every transcript has been broken into searchable sections and indexed so the system understands what each part of each lecture is actually about. When you ask a question, it finds the most relevant sections across the entire archive, even if ICT used completely different words than your question.
This is why you can ask "how should I handle a losing trade mentally?" and get results from lectures where ICT discusses journaling, emotional discipline, and revenge trading — even though none of those sections contain the words "losing trade."
Knowledge Base Free
The core Q&A experience. Ask a question in plain English and get an answer built entirely from ICT's own words.
How it works
Your question is matched against thousands of indexed transcript sections. The system finds the 8 most relevant sections from across the entire archive. An AI model then reads those sections and writes a clear, structured answer. Every claim is tied to a specific source, and every source links to the exact YouTube video and timestamp.
The AI is explicitly instructed to never invent information. If the transcripts don't contain an answer, it says so. This is fundamentally different from asking ChatGPT about ICT — that model will happily hallucinate trading concepts. ICT Decoder can only tell you what ICT actually said.
Deep Analysis Pro
For questions that span multiple lectures and years of teaching. Deep Analysis retrieves more source material and produces a structured, multi-section breakdown.
How it differs from Knowledge Base
Where Knowledge Base retrieves 8 sections, Deep Analysis pulls 20 transcript sections from across the archive. This wider net captures nuance that a standard query might miss — connecting how ICT explained a concept in Core Content versus how he refined it in the 2022 Mentorship versus what he said most recently.
The synthesis model (Claude Opus) is also significantly more capable, producing organized sections with headers, cross-referencing ideas across different lectures, and delivering the kind of structured analysis you'd expect from spending hours watching multiple videos yourself.
Hidden Gems Elite
This is not a chatbot. Hidden Gems is an autonomous, multi-step research agent that plans its own investigation, executes it, checks its own work for contradictions, and then delivers a comprehensive report. You watch each step happen in real time.
Decomposition
HaikuA planning model reads your question and breaks it into 2-4 targeted sub-queries. Each sub-query is designed to search a distinct aspect of your question. For example, "How does ICT's use of order blocks relate to his liquidity concepts?" becomes separate searches for order block mechanics, liquidity theory, and the intersection of the two.
Iterative Retrieval + Analysis
SonnetEach sub-query runs its own semantic search against the full archive. For every set of results, a dedicated analysis model extracts findings, identifies key quotes, and flags gaps. This happens sequentially — each step's results are visible to you as they complete. Sources are deduplicated across steps so the final report doesn't repeat itself.
Contradiction Detection
SonnetAfter all sub-queries complete, a separate model reviews the compiled findings and looks for conflicts. Did ICT say different things about the same concept in different years? Has his teaching evolved? This step catches nuance that a simple Q&A would miss entirely — and it's something no other trading education tool does.
Final Synthesis
OpusThe most powerful model (Claude Opus) receives all findings, the contradiction analysis, and up to 20 deduplicated source sections. It weaves everything into a comprehensive, structured report — with inline citations linking to exact YouTube timestamps. The result reads like a research paper on your specific question, grounded entirely in what ICT actually said.
What makes this different
There are other ways to search for ICT content. Here's why none of them do what this does.
Asking ChatGPT about ICT
- No access to ICT's actual transcripts
- Hallucinated concepts presented as fact
- No source citations or timestamps
- Can't distinguish ICT's words from internet noise
- No awareness of how ICT's teaching evolved
ICT Decoder
- Searches 567 actual transcripts semantically
- Only states what the transcripts contain
- Exact quotes link to the second ICT speaks; topic references link to the relevant section
- Only ICT's words — no third-party interpretation
- Contradiction detection tracks evolving views