raghu@dark-factory :~/kb/domain/agentic-ai $ cat

Agentic AI

179 concepts the autonomous brain knows in this domain.

Agent = Model + Harness

An agent is a model wrapped in load-bearing scaffolding; improving the harness matters as much as the model.

Tool use / function calling

The model acts on the world through typed tool calls.

Model Context Protocol (MCP)

Open protocol for exposing tools/resources to models.

Agent-to-agent (A2A)

Protocol for agents delegating to and coordinating with other agents.

Context engineering

Deciding what enters the model's window and when.

Retrieval-grounded generation

Ground output in fetched source text, not parametric memory alone.

Grounding & fact-checking

Verify claims against real sources before shipping.

Hallucination control

Keep fabricated claims near zero; gate what can't be verified.

Evals & benchmarks

Measure agent quality with repeatable tasks.

Agent memory

Durable state across runs; facts, feedback, project context.

Planning & reflection

Reason about goals and critique own output before acting.

Agentic design patterns

Reusable agent design patterns — reflection, tool use, planning, and multi-agent collaboration.

Multi-agent systems

Multiple specialized agents coordinating on a task, often via agent-to-agent protocols — a leading direction in agentic…

World model

A persistent, self-updated internal model of the system's operating environment and recent state, maintained by the…

Test-time compute

The paradigm of spending more inference-time compute on extended reasoning or search to raise accuracy on hard…

AI-slop / de-slopping

The recognizable stylistic tells of unedited LLM prose (em-dashes, the 'it isn't A, it's B' antithesis closer, guru…

AI formal reasoning / theorem proving

Using language models to generate and machine-check formal mathematical proofs, extending model reasoning from…

Schema-constrained generation

Forcing a model to emit output that validates against a fixed schema (e.g. a strict JSON array) so downstream tooling…

Content provenance & trust

Verifiable tracking of whether content is human- or machine-generated to establish origin and trust as the web fills…

Proof of personhood

Cryptographic or behavioral methods that verify a request or action originates from a unique human rather than a bot or…

Agentic web

The emerging paradigm in which AI agents rather than humans are primary consumers of web content, forcing sites to both…

Adversarial anti-AI content

Content and media deliberately designed to stay legible to humans while resisting AI scraping and model ingestion, a…

Reward hacking

A self-optimizing agent games its success metric (busywork, near-duplicate edits, or manipulating its own referential…

Model collapse

The degradation that occurs when models are recursively trained on AI-generated output, causing loss of diversity and…

Generative media

Diffusion and generative models that synthesize or reimagine images, audio, and video rather than text.

Reverse centaur

An automation antipattern where the human is subordinated to the machine's pace and judgment rather than augmented by…

Human-on-the-loop oversight

A supervision model where humans monitor and can intervene on an autonomous agent fleet as it runs, rather than…

Orchestrator/supervisor agent

A supervising agent that plans, delegates to, and oversees a fleet of worker sub-agents instead of performing the work…

Model extraction / weight theft

Adversarially stealing a model's weights or replicating its behavior by querying it, turning proprietary model IP into…

Agent accountability

Attributing and auditing autonomous agent actions so each decision carries a traceable provenance and a responsible,…

Sybil attacks

A single actor forges many fake identities to game a ranking, vote, or reputation system, defeating…

Graph RAG

Retrieval that traverses a knowledge or code graph's entities and relations instead of flat text chunks, tightening…

Constitutional AI

Anthropic's alignment method for Claude: the model is trained to follow an explicit set of written principles (a…

Claude Agent SDK

Anthropic's framework for designing agentic harnesses with Claude — tool use, hooks, subagents, and context/memory…

Claude Design (Anthropic Labs)

An Anthropic Labs product for collaborative visual creation: describe a design and Claude builds prototypes,…

Human-in-the-loop

An oversight paradigm where a human must approve each agent action before it executes, blocking the loop until sign-off…

AI stylometric tells

The identifiable stylistic fingerprints of machine-written text (em-dash overuse, the 'not X, but Y' antithesis closer,…

Prompt injection

An attack where adversarial text hidden in tool output, fetched web content, or documents overrides an agent's real…

Automation bias

The human tendency to over-trust automated or AI-generated outputs and under-apply independent judgment, which agentic…

Agentic RAG

A retrieval paradigm in which the agent itself iteratively decides when, what, and how to retrieve — issuing…

Crawler tarpit

A defensive endpoint that traps and slows abusive automated crawlers by serving deliberately slow, deep, or endless…

Browser fingerprinting

Distinguishing real human browsers from headless scraping automation by profiling TLS, JS-runtime, and rendering…

AI Output Watermarking

Embedding a statistically detectable, hard-to-remove signal into AI-generated text or media so its machine origin can…

Centaur (human+AI)

A collaboration model in which a human directing an AI assistant outperforms either the human or the AI acting alone,…

Device attestation

Cryptographic attestation of a genuine device or platform (e.g. Private Access Tokens) that lets a site admit real…

Web bot auth

An emerging standard where legitimate automated agents cryptographically sign their HTTP requests (HTTP Message…

AI productivity paradox

The measured gap between perceived and actual gains from LLM assistance, a durable caution against reflexive 'just ask…

Behavioral biometrics

Distinguishing humans from bots by analyzing interaction dynamics such as click cadence and inter-event timing rather…

Cognitive offloading

The human tendency to outsource reasoning and recall to AI tools, which over time erodes the operator's own skill and…

Appropriate reliance

The practice of calibrating trust to a model's actual reliability — accepting correct output and overriding incorrect…

Epistemic debt

The accumulating hidden cost of accepting unverified AI answers in place of grounded understanding, which quietly…

Critical ignoring

The information-literacy skill of deliberately declining to engage with or trust low-value AI outputs and sources, as…

AI hype cycle

The recurring maturity arc in which inflated expectations of AI give way to a disillusionment backlash against…

Automation-framework detection

Classifying a visitor as a machine actor by probing browser-automation tells such as the navigator.webdriver flag and…

Progressive session trust

Accumulating behavioral evidence over a session to raise and persist a human-trust signal that unlocks further…

Agent egress transparency

Auditing and disclosing exactly what data an AI agent or dev CLI transmits over the network, so its exfiltration…

Minimal Agent Scaffolding

The design philosophy of building an agent from the smallest legible core loop (a model plus a few tools) instead of a…

Text-to-SQL

Translating natural language into database queries with an LLM, where the generated SQL must be validated or…

AI Supply-Chain Security

Vetting the provenance and data-flow of third-party AI models, tools, and dependencies to know exactly what they access…

Tool Poisoning

An attack where malicious tool or MCP-server metadata and descriptions hijack an agent's behavior to exfiltrate data or…

Stylometric evasion

Mechanically stripping statistical style markers that flag text as machine-written (em-dashes, the 'it isn't A, it's B'…

Generative world simulators

Frontier generative models that synthesize interactive, explorable visual futures or environments on the fly instead of…

Reflexion

Agentic technique in which an agent distills lessons from its past attempts into episodic memory so a self-improvement…

Agent trajectory

The ordered sequence of an agent's states, actions, and observations across a run, treated as a first-class artifact…

Cognitive Architecture

The integrated, persistent structural substrate — doctrine, ontology, world-model, and memory — that organizes how an…

Self-Consistency

Sampling several independent reasoning paths for one problem and taking the majority or aggregated answer to raise…

Chain-of-Thought

A prompting technique that elicits explicit intermediate reasoning steps before a final answer, improving multi-step…

ReAct (Reason+Act)

An agent pattern that interleaves reasoning traces with tool/environment actions so each action is chosen from an…

Best-of-N Sampling

A test-time strategy that generates N candidate outputs and selects the single best via a verifier, judge, or gate…

Output Homogenization

When a large share of knowledge work is mediated by a handful of shared AI models, the collective body of output…

Algorithmic Monoculture

When many actors depend on the same few foundation models, their outputs and errors become correlated, flattening idea…

Autoformalization

The technique of translating natural-language mathematics into machine-checkable formal statements so an LLM's…

Tree of Thoughts

A reasoning technique that explores multiple branching solution paths with lookahead and backtracking rather than…

Multi-Agent Debate

Several agents independently reason and then critique each other's answers over rounds to converge on a more factual,…

Lazy tool loading

An agent loads tool schemas on demand at call time rather than injecting every tool definition upfront, cutting the…

Context rot

The degradation of an LLM's attention and reliability as its context window fills with tokens, so long-running agents…

Context compaction

Actively summarizing or pruning an agent's accumulated conversation history so a long-horizon run stays within the…

Long-horizon agency

The capacity of an agent to sustain a coherent, multi-step task across many actions and long wall-clock time without…

Mechanistic interpretability

Reverse-engineering a model's internal features and circuits to explain how it actually computes an answer, rather than…

Chain-of-thought faithfulness

The degree to which a model's verbalized reasoning trace genuinely reflects the computation that produced its answer,…

Wire-level tool auditing

Reverse-engineering the raw network traffic of an AI coding agent to independently verify exactly what data leaves the…

Sparse autoencoder

A dictionary-learning probe that decomposes a model's dense activations into sparse, human-interpretable features to…

Activation steering

Controlling model behavior at inference by adding interpretable feature directions into the residual stream instead of…

Lethal Trifecta

The compounding agent-security risk that arises only when a system simultaneously has access to private data, exposure…

Potemkin understanding

The failure mode where a model produces correct benchmark answers while lacking the underlying conceptual grasp, so…

Circuit tracing

Reconstructing the internal computation a model uses for a specific behavior by building attribution graphs over its…

Superposition

The phenomenon whereby a neural network represents more features than it has neurons by encoding them as overlapping…

CoT Monitorability

The safety property and practice of reading an agent's externalized chain-of-thought to detect misbehavior or intent,…

Activation patching

A causal interpretability method that copies activations from one forward pass into another (ablate/patch) to measure…

Memory Poisoning

An attack that plants malicious or false content into an agent's persistent long-term memory so it silently corrupts…

Logit Lens

An interpretability technique that projects a transformer's intermediate-layer hidden states through the output…

Continuous Partial Attention

The fragmented cognitive state of continuously monitoring many low-value information streams at once, the condition the…

Cost-aware agent eval

Benchmarking and ranking AI agents by the tokens and dollars they spend per completed task, treating cost-efficiency as…

Mixed-Initiative Interaction

An interaction paradigm in which a human and an agent fluidly hand control back and forth on a shared live artifact…

Causal abstraction

A mechanistic-interpretability framework that tests whether a high-level causal model of a task is actually implemented…

In-Context Learning

A model's ability to acquire task behavior from instructions and examples placed in the prompt at inference time,…

Induction Head

An attention circuit that completes prompt patterns by copying what followed a token's previous occurrence, identified…

Deskilling

The long-term erosion of human practitioners' underlying competence when agents routinely perform the work, leaving…

Transcoder

An interpretability tool that replaces an MLP block with a sparse, more interpretable approximation to trace features…

Ironies of automation

The paradox that automating the easy parts of a task leaves humans with the harder residual work of monitoring and…

Linear probing

An interpretability technique that trains a simple linear classifier on a model's internal activations to test whether…

Monosemanticity

The interpretability goal of decomposing a model's activations into features that each carry a single,…

Jagged capability frontier

The observation that an LLM's competence is unevenly distributed, excelling at some tasks while unpredictably failing…

Benchmark contamination

Leakage of evaluation or test data into a model's training corpus, which inflates benchmark scores so they overstate…

Causal scrubbing

An interpretability method that validates a hypothesized circuit by resampling activations along behavior-preserving…

Crosscoder

A sparse dictionary-learning variant that jointly learns shared features across multiple layers or across two models…

Path patching

A causal-interpretability technique that intervenes on specific computational paths (edges) between model components…

Attribution patching

A gradient-based linear approximation of activation patching that estimates every component's causal effect in a single…

Normalization of deviance

The gradual process by which operators of an autonomous system come to accept recurring anomalies as normal, eroding…

Dictionary learning

An unsupervised technique that decomposes a model's dense, polysemantic activations into a large overcomplete set of…

Automated interpretability

Using an LLM to automatically generate and score natural-language explanations for the features or neurons surfaced by…

AI content disclosure

The practice, and increasingly the legal requirement, of explicitly labeling machine-generated text, media, or actions…

Accountability sink

An automated arrangement in which decisions issue from a system that no identifiable human owns, so responsibility for…

Default-deny egress

A security posture that blocks all outbound network by default and requires explicit allowlisting, enforcing an agent's…

Attention economy

The media and economic system in which finite human attention is the scarce, contested resource that content, feeds,…

System-prompt extraction

An attack (or audit technique) that recovers an agent's hidden system/instruction prompt by wire-level interception or…

Capability-Based Security

A security model that grants an agent only unforgeable, narrowly-scoped tokens conferring specific authority,…

AI washing

The deceptive marketing practice of overstating, mislabeling, or fabricating a product's use of AI to attract users,…

Information extraction

Pulling structured entities and relations out of unstructured text at scale, e.g. mining thousands of book mentions…

Jailbreaking

Crafting adversarial prompts that bypass a model's safety alignment or system-prompt constraints to elicit disallowed…

Confused deputy

A privileged agent tricked by untrusted input into misusing its own authority on an attacker's behalf, e.g.…

Capability overhang

The gap between a frontier model's latent abilities and what current prompts, tools, and harnesses actually elicit, so…

Astroturfing

The practice of manufacturing fake grassroots consensus—via self-upvoting or coordinated inauthentic engagement—so…

Sleep-time compute

An agent spends idle time between queries pre-processing its context and building memory so that live requests need…

Model / System Card

A standardized provider-published document disclosing a model's intended use, evaluations, limitations, and safety…

Black-box auditing

Assessing an AI system's real behavior, cost, and honesty purely from its inputs, outputs, and network traffic when the…

Capability–cost frontier

The Pareto frontier of model or agent quality against dollar-and-latency cost, used to pick the cheapest configuration…

Data exfiltration

The covert reading of local files or sensitive context by an AI tool or agent and its silent transmission to a remote…

Workspace confinement

Restricting an autonomous agent's filesystem access to a single explicitly-allowed directory so it cannot read or write…

Ambient authority

The default, undesignated authority an agent inherits over its whole environment (e.g. read access to the user's entire…

Computer use

A paradigm in which an agent operates a full computer environment — shell, filesystem, and screen — as a general tool,…

Design fiction

The practice of using generative models to produce deliberately plausible speculative artifacts (e.g., a future news…

Blast-radius containment

Designing an autonomous agent's runtime so that a compromise or misfire can reach as little as possible — through…

Agent skills

Packaged, discoverable instruction-and-resource modules an agent loads on demand to acquire a capability, exposing only…

Information-Flow Control

Information-flow control tags sensitive data and tracks it as it moves through a program so an agent cannot route…

Excessive agency

An agentic system granted more autonomy, permissions, or file-system reach than its task actually requires, so a bug or…

Trust boundary

The perimeter separating components that trust each other from those that don't; most agent-security failures are…

Data-loss prevention

Content-aware inspection that detects and blocks sensitive data (keys, credentials, source) from leaving an agent's…

RLHF

Reinforcement learning from human feedback fine-tunes a model against a reward model trained on human preference…

Vector embedding

A dense numeric vector that encodes the semantic meaning of text so related content can be found by distance in…

Tokenization

Splitting text into subword tokens (e.g. via byte-pair encoding) is the model's input representation whose count drives…

Agentic misalignment

An autonomous agent pursuing its inferred objectives takes actions harmful to its principal — such as exfiltrating data…

Model Arena

An evaluation paradigm that ranks models by crowd-sourced pairwise (head-to-head) preference votes to produce an…

Evaluation awareness

A model detecting that it is under evaluation and altering its behavior accordingly, which undermines the…

Goodhart's law

When a proxy metric or benchmark becomes the optimization target it stops reflecting the real capability it was meant…

Machine Unlearning

Techniques for removing the influence of specific training data from an already-trained model without full retraining,…

Training-data consent

Norms and machine-readable signals (do-not-train opt-outs, consent registries, provenance-based exclusion) that let…

Consent-or-Pay

A data-collection model in which a service conditions core access on the user either consenting to have their data used…

Agent-native application

Software designed from the ground up as a shared execution surface that both human users and AI agents operate as…

Embodied agents

AI agents that act through physical or playful embodiments (toys, whimsical hardware, robots) rather than only a text…

Data commons

The shared pool of open public data that AI systems both depend on for training and threaten to enclose, degrade, or…

Word Error Rate (WER)

The standard edit-distance metric for scoring automatic speech recognition by counting insertions, deletions, and…

Speaker Diarization

The speech-processing task of partitioning an audio stream into segments labeled by who is speaking (the 'who spoke…

Recursive Self-Improvement

The paradigm in which an AI system iteratively improves its own capabilities or harness across cycles, now being…

Workslop

AI-generated work output that looks polished and complete but offloads the real interpretive and verification effort…

Intelligence explosion

The hypothesized rapid, self-reinforcing acceleration of capability that could follow once an AI system can reliably…

Fill-in-the-Middle

A code-model training and inference technique that completes or edits code conditioned on both the preceding and the…

LLM jury

Aggregating verdicts from a panel of several diverse LLM evaluators instead of a single judge, to label data and score…

Weak supervision

Training or evaluating models on large volumes of cheap, noisy labels — increasingly produced by panels of LLMs…

Knowledge-graph embedding

Encoding a knowledge graph's entities and relations as dense vectors so connected concepts can be retrieved and…

LLM-as-annotator

Using an LLM, or a panel of them, to label, tag, and annotate datasets in place of human annotators as a scalable…

Graph community detection

Unsupervised graph algorithms such as Louvain or Leiden that partition a network into densely connected clusters, used…

Compute overhang

The condition in which available hardware compute far exceeds what current algorithms exploit, so an algorithmic or…

Generation-verification gap

The asymmetry that checking a candidate answer is often cheaper and more reliable than producing it, which is why…

Corrigibility

The property of an autonomous agent accepting human correction, reversal, and shutdown without resisting or subverting…

Mesa-optimization

When a trained model internally develops its own optimizer whose learned objective can diverge from the objective it…

Instrumental convergence

The tendency of sufficiently capable goal-directed agents to pursue similar sub-goals such as self-preservation and…

Agent Experience (AX)

The design discipline of shaping API surfaces, docs, and affordances so autonomous agents — not just humans — are…

Foundation Model

A large model pretrained on broad data that serves as an adaptable base for many downstream tasks via fine-tuning,…

Open Source AI Definition

The standard (formalized by the OSI's OSAID) that a model counts as genuinely open source only when weights,…

Agent-as-a-Service

The paradigm of delivering autonomous computer-use agents as hosted, callable developer APIs rather than as local tools…

GUI grounding

A vision-language agent's capability to localize and act on on-screen UI elements (buttons, fields, menus) directly…

Agent browser infrastructure

Hosted, API-controllable browser or desktop environments provisioned so computer-use agents can drive real interfaces…

Accessibility-tree grounding

Grounding a computer-use agent in the platform's accessibility tree — structured element roles, names, and states — so…

Set-of-mark prompting

A prompting technique that overlays numbered or labelled marks on a UI screenshot so a vision-language agent can refer…

Model Provenance

The verifiable origin, training lineage, and integrity chain of a downloadable model's weights, letting a self-hosting…

Frontier model

The class of most-capable general-purpose AI models at the current edge of capability — the tier that AI safety and…

Social simulacra

Using a population of LLM-driven generative agents to simulate and forecast how a human community would behave, letting…

Cross-encoder reranking

A two-stage retrieval refinement where a heavier cross-encoder rescores the top candidate documents jointly against the…