IntermediateAgentic AI
Information Agents
Continuously running AI systems that proactively monitor, synthesize, and act on information across your digital workspace—transforming search from reactive queries into autonomous intelligence.
AI Dictionary
Trending Concepts
Recently updated and popular entries
Intermediate
AdvancedModels & Architecture
Self-Evolving Agentic Models
AI systems that autonomously improve their own capabilities by generating synthetic training data, debugging their own learning process, and modifying their reasoning strategies—early steps toward recursive self-improvement.
AdvancedSafety & Ethics
Autonomous AI Cybersecurity Defense
The paradigm shift where AI systems autonomously discover, verify, and help patch software vulnerabilities faster than human researchers and threat actors—finally tilting the attacker-defender balance toward defense.
IntermediateModels & Architecture
Gemini Omni
Google's any-to-any multimodal foundation model capable of generating any output (text, image, audio, video) from any input, with physics-grounded video generation as its first major capability.
From A to Z
Every concept, without the jargon.
A
IntermediateModels & Architecture
Activation Function
Activation functions introduce non-linearity into neural networks, enabling them to learn complex patterns. Common ones: ReLU, GELU (transformers), sigmoid, softmax.
AdvancedResearch
Activation Steering
A technique that injects synthetic vectors into a model's internal layers at inference time to directly shift its decision-making, enabling precision debiasing and behavioural control — but also capable of bypassing safety training without any jailbreak prompt.
AdvancedTools & Frameworks
ActTail
A global activation sparsity method that optimizes LLM inference by intelligently allocating compute budgets based on the statistical properties of Transformer weights.
IntermediateModels & Architecture
Adaptive Thinking in AI
A reasoning strategy where AI models dynamically adjust how much they think per turn — from instant responses to deep multi-step deliberation — based on task complexity.
AdvancedModels & Architecture
Adversarial Cost to Exploit (ACE)
A security benchmark that measures the economic token cost an adversary must spend to trick an AI agent into unauthorized tool use, replacing static pass/fail evaluations with game-theoretic cost analysis.
AdvancedTools & Frameworks
Agent Browser Protocol (ABP)
Open-source custom Chromium that pauses the web between AI agent steps, achieving 90%+ reproducible automation success.
AdvancedAgentic AI
Agent Evaluation
The practice of measuring AI agent performance using deterministic, execution-based testing environments that verify complete tool-call trajectories rather than relying on subjective LLM-as-a-judge grading.
BeginnerAgentic AI
Agent Operational Memory
A technique that externalises an AI agent's behavioural rules and learned heuristics into structured files loaded at session start, giving the agent persistent and consistent behaviour across restarts without fine-tuning.
AdvancedSafety & Ethics
AgentDrift
Benchmark proving AI agents blindly accept corrupted tool data — 0 out of 1,563 turns questioned, while appearing to perform well on standard metrics.
IntermediateAgentic AI
Agentic AI
AI systems that combine language models with reasoning and tool-use to autonomously execute complex, multi-step tasks — now supported by dedicated infrastructure for production deployment.
IntermediateAgentic AI
Agentic Engineering
The discipline of building autonomous AI agent systems — covering architecture, orchestration, tool integration, safety, and operations.
IntermediateAgentic AI
Agentic RAG
RAG where an autonomous agent controls the retrieval process — iteratively searching, refining queries, and cross-referencing sources.
BeginnerAgentic AI
AI Agent
An AI system that autonomously plans, reasons, and takes actions to accomplish goals using tools
IntermediateSafety & Ethics
AI Alignment
Ensuring AI systems behave in accordance with human values, intentions, and safety requirements
BeginnerPractical Applications
AI API
An AI API is a web service that lets developers integrate AI model capabilities into applications via simple HTTP requests, without running models themselves.
BeginnerSafety & Ethics
AI Governance
AI governance is the framework of policies, regulations, and practices that ensure AI systems are developed and deployed responsibly, fairly, and in compliance with laws.
BeginnerCore Concepts
AI Hallucination
When an LLM confidently generates false or fabricated information
BeginnerCore Concepts
AI Inference
The process of running a trained LLM to generate output from input
IntermediateSafety & Ethics
AI Jailbreaking
Adversarial techniques that bypass an LLM's safety guardrails to produce prohibited content — a key threat that drives AI safety research and red-teaming practice
IntermediateIndustry & Business
AI Observability
Monitoring, logging, and analyzing AI system performance in production — catching quality regressions, cost anomalies, and failures before they impact users
IntermediateAgentic AI
AI Orchestration
AI orchestration coordinates multiple AI models, tools, and data sources into unified workflows, managing the flow between components in complex AI systems.
IntermediateSafety & Ethics
AI Red Teaming
Systematically probing AI systems for vulnerabilities, failure modes, and alignment gaps before deployment — now quantifiable in dollar terms via economic benchmarks like ACE.
IntermediatePractical Applications
AI Robotics
The integration of advanced AI foundation models with robotic hardware to create machines capable of autonomous, real-world reasoning and physical manipulation.
IntermediateAgentic AI
Always-On Agents
AI systems that run autonomously in the cloud on schedules, API triggers, or webhooks — executing complex workflows without requiring a user's local machine.
BeginnerCore Concepts
Artificial Intelligence (AI)
Artificial intelligence is the field of computer science that builds systems capable of performing tasks normally requiring human intelligence, such as learning, reasoning, and perception.
AdvancedModels & Architecture
Attention Mechanism
The mathematical mechanism that allows transformers to dynamically focus on the most relevant parts of the input when processing each token
AdvancedModels & Architecture
Automated Alignment Research
Using frontier AI models to autonomously discover methods for aligning other AI systems — addressing the scalable oversight challenge by letting safety research scale with capabilities.
AdvancedSafety & Ethics
Autonomous AI Cybersecurity Defense
The paradigm shift where AI systems autonomously discover, verify, and help patch software vulnerabilities faster than human researchers and threat actors—finally tilting the attacker-defender balance toward defense.
IntermediateModels & Architecture
Autoregressive Generation
Autoregressive generation is how LLMs produce text: predicting one token at a time, with each new token conditioned on all previously generated tokens.
B
IntermediateCore Concepts
Batch Size
Batch size (examples per update) and learning rate (step size for weight updates) are the two most important hyperparameters controlling how neural networks train.
AdvancedModels & Architecture
Beam Search
Beam search generates text by exploring multiple candidate sequences in parallel, keeping the top-k most promising paths to find the highest-probability output.
BeginnerCore Concepts
Benchmark (AI Evaluation)
A benchmark is a standardized test used to measure and compare AI model performance, providing reproducible scores across tasks like reasoning, coding, and knowledge.
BeginnerSafety & Ethics
Bias in Machine Learning
Bias in ML refers to systematic errors from data, algorithms, or deployment that cause models to produce unfair or discriminatory results.
AdvancedModels & Architecture
Bicameral Model
A neural architecture that couples two parallel language models via their hidden states for real-time latent-channel coordination, dramatically improving reasoning accuracy without token overhead.
IntermediateAgentic AI
Binex
A local testing framework that orchestrates AI agents using YAML DAGs, providing deep visibility and CLI debugging for multi-agent workflows.
C
AdvancedCore Concepts
Catastrophic Forgetting
Catastrophic forgetting is when training a neural network on new data overwrites previously learned knowledge, causing it to lose earlier capabilities.
IntermediatePractical Applications
Chain-of-Thought Prompting
A prompting technique that asks LLMs to reason step-by-step before answering, dramatically improving accuracy
BeginnerCore Concepts
Classifier
A classifier is an ML model that assigns inputs to predefined categories — the foundation of spam filters, sentiment analysis, image recognition, and fraud detection.
BeginnerTools & Frameworks
Claude Code
Anthropic's terminal-based AI coding assistant that operates as a multi-agent runtime for autonomous software engineering across entire repositories.
IntermediateAgentic AI
CODREAM
A post-task reflective protocol for multi-agent AI in which agents collaboratively analyse completed tasks, distil insights into compact heuristics, and route that knowledge asymmetrically to teammates who need it most — permanently improving performance without fine-tuning.
IntermediateTools & Frameworks
Composio
An open-source integration platform that connects AI agents to over 1,000 external tools, handling complex API routing and secure authentication.
AdvancedSafety & Ethics
Constitutional AI (CAI)
A training approach where AI models critique and revise their own outputs against a set of principles, using AI-generated feedback for scalable alignment
IntermediateAgentic AI
Context Compression for AI Agents
Techniques to reduce token counts while preserving meaning — critical for agentic workflows that exhaust even million-token context windows.
BeginnerTools & Frameworks
Context Rot
The gradual degradation of AI agent performance as a session accumulates tokens, causing the model to lose focus on earlier instructions and constraints.
BeginnerCore Concepts
Context Window
The maximum number of tokens an LLM can process in a single request
AdvancedCore Concepts
Continual Learning
Continual learning enables AI systems to learn new tasks over time without forgetting previous knowledge, solving the stability-plasticity dilemma.
IntermediateCore Concepts
Cosine Similarity
Cosine similarity measures how similar two vectors are by computing the cosine of the angle between them — the standard metric for comparing AI embeddings.
D
AdvancedSafety & Ethics
DeceptGuard
A constitutional oversight framework that detects deceptive behavior in LLM agents by analyzing their internal reasoning traces and hidden states.
BeginnerCore Concepts
Deep Learning
Deep learning is a machine learning technique using multi-layered neural networks that automatically learn hierarchical data representations, powering modern AI breakthroughs.
BeginnerModels & Architecture
DeepSeek
A highly efficient, open-weight AI model family that delivers frontier-level coding and reasoning capabilities at significantly lower computational costs.
AdvancedModels & Architecture
DeepStack Injection
A VLM architecture that routes abstract visual features to early Transformer layers and high-resolution details to later layers for optimal document parsing in compact models.
IntermediateTools & Frameworks
Deterministic Agent State Machine
An AI agent architecture that governs execution through predefined states and tool-gating rules rather than LLM judgment, physically preventing destructive or out-of-order actions and dramatically improving production deployment rates.
BeginnerAgentic AI
Difference Between a Chatbot
A chatbot responds to messages in conversation; an AI agent autonomously plans, uses tools, and takes multi-step actions to achieve goals.
IntermediateOpen Source
Difference Between Open Weights
Open weights means model parameters are downloadable; true open source includes weights, training code, data, and a permissive license — most "open" AI models are open weights only.
BeginnerCore Concepts
Difference Between Regression
Classification predicts categories (spam/not spam); regression predicts continuous values (house price). These are the two fundamental supervised ML problem types.
BeginnerCore Concepts
Difference Between Training
Training teaches a model by adjusting its parameters on data (expensive, done once); inference uses the trained model to make predictions (cheap, done millions of times).
IntermediateModels & Architecture
Diffusion Model
A diffusion model generates images by learning to reverse a noise-adding process, iteratively refining random noise into coherent outputs guided by text prompts.
AdvancedAgentic AI
Dynamic Cognitive Scaffolding
A technique that lets AI agents build their own reasoning structures at inference time rather than relying on fixed scaffolds, significantly improving performance on complex tasks.
E
IntermediatePractical Applications
Edge AI
Edge AI runs AI models directly on local devices instead of the cloud, enabling privacy, low latency, and offline functionality through quantized and distilled models.
IntermediateCore Concepts
Embedding
A numerical vector that captures the semantic meaning of text, enabling similarity search
AdvancedPractical Applications
Embodied AI
AI systems designed to perceive and interact with physical or virtual environments, bridging the gap between digital reasoning and real-world action.
AdvancedModels & Architecture
Emotion Vectors
Measurable internal neural representations inside AI models that function like emotions and causally steer the model's behavior.
IntermediateModels & Architecture
Encoder-Decoder Architecture
An encoder-decoder architecture pairs an encoder (which reads and compresses input) with a decoder (which generates output), forming the basis of transformer model variants like BERT, GPT, and T5.
IntermediateSafety & Ethics
Explainability & Interpretability in AI
Explainability and interpretability address the AI black-box problem: understanding why models make specific decisions, using techniques like SHAP, LIME, and Chain-of-Thought.
F
IntermediateCore Concepts
Feature Engineering
Feature engineering transforms raw data into informative input variables for ML models — selecting, creating, and encoding features that help models learn effectively.
AdvancedCore Concepts
Federated Learning
Federated learning trains AI models across decentralized devices by sharing model updates instead of raw data, enabling privacy-preserving machine learning.
BeginnerPractical Applications
Few-Shot Prompting
Providing a few worked examples in the prompt to guide an LLM's behavior — typically improving accuracy by 20-30% over zero-shot
IntermediateCore Concepts
Fine-Tuning
Training a pre-trained LLM further on domain-specific data to specialize its behavior
AdvancedModels & Architecture
Flash Attention
A hardware-aware algorithm that massively speeds up LLM processing by optimizing GPU memory reads, enabling very long context windows.
IntermediateModels & Architecture
Foundation Model
A foundation model is a large AI model pre-trained on broad data at scale that can be adapted to many downstream tasks through prompting, fine-tuning, or retrieval augmentation.
IntermediateAgentic AI
Function Calling
Function calling lets LLMs request the execution of external tools and APIs, enabling real-world actions and data retrieval beyond text generation.
G
IntermediateAgentic AI
Galactic
An open-source orchestration tool that isolates parallel AI coding agents into separate Git worktrees to prevent file and port conflicts.
IntermediateModels & Architecture
GAN (Generative Adversarial Network)
A GAN uses two competing neural networks — a generator and a discriminator — to produce realistic synthetic data through adversarial training.
IntermediateModels & Architecture
Gemini Omni
Google's any-to-any multimodal foundation model capable of generating any output (text, image, audio, video) from any input, with physics-grounded video generation as its first major capability.
IntermediateModels & Architecture
Gemma 4
Google DeepMind's open-weight multimodal model family that natively handles text, vision, and audio on-device.
BeginnerCore Concepts
Generative AI
Generative AI is a category of AI systems that create new content — text, images, audio, code — rather than just analyzing existing data.
BeginnerPractical Applications
Generative Engine Optimization (GEO)
Optimizing content for AI discovery instead of just search engines — answer-first structure, structured data, and question-oriented titles.
BeginnerModels & Architecture
GPT
GPT (Generative Pre-trained Transformer) is OpenAI's family of large language models that demonstrated how scaling transformers produces increasingly capable AI.
IntermediateCore Concepts
Gradient Descent
Gradient descent is the optimization algorithm that trains neural networks by iteratively adjusting parameters in the direction that reduces prediction error.
IntermediatePractical Applications
GraphRAG
A RAG architecture that pre-builds a knowledge graph from documents, enabling multi-hop reasoning over entity relationships instead of flat vector search.
IntermediatePractical Applications
Grounding in AI
Anchoring LLM responses to verified external sources to reduce hallucinations and enable citation
AdvancedModels & Architecture
GRPO (Group Relative Policy Optimization)
A reinforcement learning algorithm that aligns language models by comparing groups of outputs against each other, eliminating the need for a separate reward model.
IntermediateSafety & Ethics
Guardrails
Guardrails are safety mechanisms that constrain AI system behavior — filtering inputs, validating outputs, and preventing harmful or off-topic responses in production applications.
I
AdvancedSafety & Ethics
ILION
A deterministic safety gate that instantly blocks unauthorized real-world actions proposed by AI agents without relying on statistical training.
IntermediatePractical Applications
In-Context Learning (ICL)
The ability of LLMs to learn new tasks from examples provided in the prompt — without any weight updates or fine-tuning
AdvancedAgentic AI
Inference-Time Co-Evolution
A training-free paradigm where a population of AI agents dynamically specialises, learns from failures, and restructures its own collaboration topology during execution — without updating model weights.
IntermediateAgentic AI
Information Agents
Continuously running AI systems that proactively monitor, synthesize, and act on information across your digital workspace—transforming search from reactive queries into autonomous intelligence.
IntermediateSafety & Ethics
Instruction Hierarchy for AI Safety
Safety pattern giving system prompts priority over user inputs and tool outputs — preventing prompt injection in autonomous agents.
AdvancedTools & Frameworks
ITBench-AA
An AI agent benchmark testing incident response on real Kubernetes infrastructure, revealing that frontier models like Claude Opus 4.7 and GPT-5.5 resolve fewer than 50% of production failures.
K
AdvancedAgentic AI
Kairos
An always-on background daemon inside Claude Code that autonomously prunes, merges, and resolves contradictions in the AI agent's working memory.
IntermediatePractical Applications
Knowledge Graph
A knowledge graph stores real-world entities and their relationships as a structured network, enabling machines to reason over connected facts and enhance AI accuracy.
AdvancedModels & Architecture
KV Cache
A memory optimization that stores previously computed key-value pairs in transformer attention layers — avoiding redundant computation and accelerating generation 3-5×
L
BeginnerCore Concepts
Large Language Model (LLM)
A neural network trained on massive text data to understand and generate human-like language
AdvancedCore Concepts
Latent Space
Latent space is the internal representation space learned by neural networks — a compressed mathematical space where data is mapped to vectors capturing essential features and relationships.
AdvancedResearch
Latent Space Manipulation
A class of techniques that directly read, steer, or couple the internal numerical representations of AI models rather than operating through text, enabling real-time alignment audits, bias detection, and token-free inter-model communication.
IntermediateModels & Architecture
LoRA (Low-Rank Adaptation)
An efficient fine-tuning method that trains only small adapter layers instead of the full model
IntermediateCore Concepts
Loss Function
A loss function measures how wrong a model's predictions are, providing the error signal that training algorithms minimize to improve the model.
M
BeginnerCore Concepts
Machine Learning (ML)
Machine learning is a branch of AI where systems learn patterns from data to improve at a task without being explicitly programmed.
IntermediateSafety & Ethics
Magnifica Humanitas
Pope Leo XIV's 150-page encyclical on AI ethics, calling for the disarmament of AI from tech monopolies, democratic oversight, and grounding AI policy in human dignity and theological anthropology.
AdvancedModels & Architecture
Mamba
A highly efficient AI architecture that uses State-Space Models instead of Transformers to process massive amounts of text with very low memory usage.
IntermediateAgentic AI
Managed Agents
Cloud-hosted AI agent platforms that handle infrastructure, credential management, and sandboxing so developers only define tasks and guardrails—dramatically accelerating agent deployment.
AdvancedModels & Architecture
MiniMax-M2
A 229.9B parameter Mixture-of-Experts model with only 9.8B active parameters per token, optimized for agentic tasks and exhibiting early signs of self-evolution—autonomously debugging its own training and modifying its scaffolding.
AdvancedModels & Architecture
Mixture-of-Experts (MoE) Model
An architecture that routes tokens to specialized sub-networks, increasing model capacity without a proportional increase in computing costs.
IntermediatePractical Applications
MLOps
MLOps applies DevOps practices to machine learning: automating deployment, monitoring, and maintenance of ML models in production.
IntermediateSafety & Ethics
Model Card
A model card is standardized AI model documentation covering intended use, performance, limitations, training data, and ethical considerations — a transparency label for AI.
BeginnerTools & Frameworks
Model Context Protocol (MCP)
Open standard for connecting AI to external tools — now embedded in browsers, CLIs, and websites via WebMCP, though cross-source data queries remain a challenge.
IntermediateModels & Architecture
Model Distillation
Training a smaller 'student' model to replicate a larger 'teacher' model's capabilities at a fraction of the cost and latency
IntermediateAgentic AI
Multi-Agent Systems (MAS)
Architectures where multiple specialized AI agents collaborate, divide tasks, and verify each other's work — with routing strategies like the Advisor pattern enabling cost-efficient orchestration.
IntermediateIndustry & Business
Multi-Tenancy in AI
Serving multiple isolated customers from a single LLM deployment — reducing per-customer costs by 40-60% while maintaining strict data separation
BeginnerMultimodal & Creative
Multimodal AI
Multimodal AI systems process and generate multiple data types — text, images, audio, video — within a single model, enabling cross-modal understanding and creation.
N
AdvancedResearch
Natural Language Autoencoders
An Anthropic interpretability technique that automatically translates a large language model's internal activation vectors into human-readable text, enabling pre-deployment alignment audits and detection of hidden biases or deceptive intent.
BeginnerCore Concepts
Natural Language Processing (NLP)
Natural language processing is the AI field that enables computers to understand, interpret, and generate human language, underpinning chatbots, translation, and LLMs.
AdvancedModels & Architecture
Nemotron-Labs Diffusion
NVIDIA's family of language models (3B-14B) that merge autoregressive and diffusion generation into one architecture, enabling both GPT-style sequential generation and 10-50x faster parallel diffusion mode.
BeginnerCore Concepts
Neural Network
A network of interconnected artificial neurons that learns patterns from data — the foundational architecture behind all modern AI
O
BeginnerOpen Source
Open-Source AI
AI models, weights, and tools that are publicly available — now matching closed-source frontier models on major benchmarks, democratizing access to advanced capabilities.
BeginnerCore Concepts
Overfitting
Overfitting means a model memorizes training data without generalizing; underfitting means a model is too simple to learn the underlying patterns. Balancing them is key to effective ML.
P
IntermediateModels & Architecture
PEFT (Parameter-Efficient Fine-Tuning)
A family of techniques that adapt large AI models to specific tasks by updating only a tiny fraction of parameters, cutting fine-tuning costs by 90–99%.
IntermediateModels & Architecture
Perplexity in NLP
The standard metric for evaluating language model quality — measuring how well a model predicts text, where lower values indicate better language understanding
AdvancedModels & Architecture
Positional Encoding
Positional encoding tells transformers the order of tokens in a sequence, since self-attention alone is position-agnostic. Modern approaches like RoPE enable 128K+ context windows.
IntermediateCore Concepts
Pre-training
Pre-training is the initial training phase where an AI model learns broad patterns from a large general-purpose dataset before being adapted for specific tasks.
IntermediateTools & Frameworks
Programmatic Tool Calling
Anthropic's replacement for JSON function calling — Claude executes code blocks to invoke multiple tools dynamically, eliminating round-trip overhead.
AdvancedSafety & Ethics
Project Glasswing
Anthropic's AI-powered security initiative that uses Claude to autonomously discover and verify tens of thousands of critical vulnerabilities in global software infrastructure faster than threat actors can exploit them.
BeginnerCore Concepts
Prompt
The input text or instructions given to an LLM to generate a response
IntermediateIndustry & Business
Prompt Caching
Storing and reusing processed prompt prefixes on LLM servers to reduce costs by up to 90% and latency by 3×
IntermediateAgentic AI
Prompt Chaining
Breaking complex tasks into a sequence of simpler LLM calls where each output feeds the next input — improving quality 20-40% over single-pass processing
BeginnerPractical Applications
Prompt Engineering
The systematic practice of designing effective prompts to get optimal results from LLMs
IntermediateSafety & Ethics
Prompt Injection
An attack where malicious input manipulates an LLM into ignoring its instructions
AdvancedAgentic AI
Proof-Derived Authorization
A security model for AI agents where every action must be backed by a cryptographic proof of its authorisation chain, making prompt injection and unauthorised actions mathematically impossible rather than merely policy-prohibited.
R
IntermediateModels & Architecture
RAG (Retrieval-Augmented Generation)
A technique that combines LLMs with external knowledge retrieval to improve accuracy and reduce hallucinations
IntermediateAgentic AI
Real-World Agent Reliability Gap
The critical gap between AI agent performance on benchmarks (90%+) versus real enterprise workflows (<50%), revealing that frontier models fail at multi-step, ambiguous, tool-heavy tasks humans routinely delegate.
IntermediateCore Concepts
Reasoning in AI
AI reasoning is the ability of models to think step by step, using techniques like Chain-of-Thought and reasoning models (o1, o3) for complex problem-solving.
IntermediateCore Concepts
Reinforcement Learning (RL)
Reinforcement learning is a machine learning paradigm where an agent learns optimal behavior through trial-and-error interaction with an environment, guided by reward signals.
BeginnerSafety & Ethics
Responsible AI
Responsible AI is the practice of building and deploying AI systems that are fair, transparent, accountable, safe, and beneficial to society.
IntermediateSafety & Ethics
Reward Hacking in AI Agents
AI agents gaming their benchmarks — evaluator tampering occurs in 50% of episodes and gets worse with more capable models.
AdvancedModels & Architecture
RLHF (Reinforcement Learning from Human Feedback)
A training technique that uses human preference ratings to align LLM behavior with human values
S
IntermediateTools & Frameworks
Safetensors
A secure binary file format for storing ML model weights that prevents arbitrary code execution, now the industry standard under the PyTorch Foundation.
AdvancedCore Concepts
Scaling Laws for LLMs
Empirical patterns showing that LLM capabilities improve predictably as model size, training data, and compute increase — enabling reliable planning of AI investments
AdvancedModels & Architecture
Self-Evolving Agentic Models
AI systems that autonomously improve their own capabilities by generating synthetic training data, debugging their own learning process, and modifying their reasoning strategies—early steps toward recursive self-improvement.
IntermediateCore Concepts
Self-Supervised Learning
Self-supervised learning trains models by generating labels from the data itself — like predicting the next token — enabling pre-training on virtually unlimited unlabeled data.
IntermediateTools & Frameworks
Semantic Chunking
Splitting documents into meaning-preserving segments based on topic boundaries rather than fixed character limits — improving RAG retrieval accuracy by 20-40%
IntermediatePractical Applications
Semantic Search
Semantic search retrieves information based on meaning rather than keywords, using AI embeddings and vector similarity to find relevant results.
IntermediatePractical Applications
Semantic Training Gap
The gap between an AI model's statistical language fluency and its grounded understanding of domain-specific operational semantics, leading to hallucinated identifiers and cascading failures in industrial applications.
AdvancedModels & Architecture
Speculative Decoding
Speculative decoding speeds up LLM inference by having a small draft model generate candidate tokens that the large model verifies in parallel — same quality, 2-3x faster.
BeginnerMultimodal & Creative
Speech AI
Speech AI covers technologies for converting speech to text (STT), text to speech (TTS), voice cloning, and speech translation, enabling natural voice interaction with AI.
IntermediateTools & Frameworks
State Machine Guardrails
A deterministic agent control technique that restricts available tools to those relevant to the current workflow phase, preventing destructive actions and reasoning loops without relying on model judgment.
AdvancedModels & Architecture
State-Space Model (SSM)
An efficient AI architecture that maintains a continuously updating internal state to process massive sequences of data without the memory overhead of Transformers.
IntermediatePractical Applications
Structured Output
Structured output forces LLMs to produce machine-readable data (like JSON) matching a predefined schema, making AI outputs reliably parseable by applications.
BeginnerCore Concepts
Supervised Learning
Supervised learning is a machine learning approach where models learn from labeled input-output pairs to make predictions on new data.
IntermediateCore Concepts
Synthetic Data
Synthetic data is artificially generated data that mimics real-world patterns, used when real data is scarce, biased, or privacy-restricted.
IntermediateSafety & Ethics
SynthID
Google's digital watermarking technology that embeds imperceptible, persistent identifiers in AI-generated images, audio, text, and video to prove synthetic origin.
BeginnerPractical Applications
System Prompt
A system prompt is the developer's instruction set that defines an LLM's behavior, role, constraints, and output format for a specific application.
T
BeginnerCore Concepts
Temperature in AI
A parameter controlling the randomness of LLM output — lower values produce consistent results, higher values increase creativity
AdvancedAgentic AI
Test-Time Co-Evolution
A training-free technique that evolves how multi-agent AI systems collaborate at inference time, allowing agents to develop specialized roles and route knowledge to where it is needed most.
BeginnerMultimodal & Creative
Text-to-Image Generation
Text-to-image generation uses AI models to create images from natural language descriptions, powered by diffusion models in tools like Midjourney, DALL-E, and Stable Diffusion.
AdvancedModels & Architecture
Text/Action Mismatch
A failure mode where an LLM verbally refuses a restricted request in its text output while simultaneously executing the forbidden action in its structured tool-call output.
BeginnerIndustry & Business
Token Economics
The pricing and cost structure of LLM usage based on token consumption
BeginnerCore Concepts
Token in AI
The smallest unit of text an LLM processes — approximately 4 characters or 0.75 words
BeginnerCore Concepts
Tokenizer
A tokenizer converts raw text into tokens — the discrete units a language model processes — using subword algorithms like BPE or SentencePiece.
IntermediateCore Concepts
Top-p (Nucleus) Sampling
A decoding method that samples from the smallest set of tokens whose cumulative probability exceeds a threshold p — adapting candidate pool size to model confidence
IntermediateAgentic AI
Trajectory Refinement
A technique that treats an AI agent's action plan as an optimizable object, iteratively refining it through inspection and textual gradient feedback to close the gap between planning and execution.
IntermediateCore Concepts
Transfer Learning
Transfer learning is a technique where knowledge from a model trained on one task is reused for a different task, enabling powerful AI with less data and compute.
IntermediateModels & Architecture
Transformer
The neural network architecture underlying all modern LLMs, using attention mechanisms to process text
V
IntermediateTools & Frameworks
Vector Database
A specialized database for storing and searching embedding vectors, enabling semantic similarity search
IntermediateModels & Architecture
VLM (Vision-Language Model)
An AI model architecture that jointly processes visual and textual inputs, enabling tasks like document understanding, image reasoning, and visual question answering.
