Ongoing
Intelligence Papers
Colendi AI Research
Books carry an argument. An archive carries the evidence. The Intelligence Papers is Colendi AI Research's record of the ideas, artifacts, and institutions that formed the modern intelligence stack: from Turing's 1936 paper on computable numbers to the reasoning models of 2025.
Every record is reviewed as evidence, not arranged as a reading list. Each carries a summary of why it mattered, an openness indicator, a difficulty rating, and a verification status, so a reader can move from a claim about intelligence to the primary source behind it.
We publish it alongside our books because the books depend on it. banking.io argues that intelligence is becoming the control layer of finance. The Intelligence Papers is where that claim can be checked.
Scope
- 281 reviewed records
- 22 research domains
- Before 2012 to 2023–2025
Domains
- Origins & Computability
- Neural Foundations
- Transformer Architecture
- Decoder-Only Language Models
- Encoders
- Encoder–Decoders
- Scaling Laws & Compute
- Data, Corpora & Tokenization
- Distributed Training
- Mixture-of-Experts
- Alignment & Preference Learning
- Reasoning & Test-Time Compute
- Agents & Tool Use
- Multimodality
- Retrieval & Memory
- Code Models
- Inference & Serving
- Long Context & Efficient Sequences
- Evaluation & Benchmarks
- Interpretability & Safety
- Continual Learning & Memory
- Open-Weight Model Families