Jonatan Matajonmatum.com
conceptsnotesexperimentsessays
© 2026 Jonatan Mata. All rights reserved.v2.1.1
Concepts

Artificial Intelligence

Field of computer science dedicated to creating systems capable of performing tasks that normally require human intelligence, from reasoning and perception to language generation.

growing#ai#machine-learning#deep-learning#llm#neural-networks#foundation-models

What it is

Artificial intelligence is the field of computer science that seeks to create systems capable of performing tasks that traditionally require human intelligence: reasoning, learning, perceiving, generating language, and making decisions.

It's not a new concept — the term was coined in 1956 at the Dartmouth conference — but the convergence of three factors transformed it in recent years: massive amounts of data, accessible compute power (GPUs, TPUs), and advances in neural network architectures, particularly the Transformer introduced in 2017.

Layers of the current ecosystem

Foundation Models

Foundation models are large-scale neural networks trained on massive amounts of unlabeled data. They're called "foundational" because they serve as a base for multiple tasks without requiring complete retraining.

Examples: GPT-4, Claude, Gemini, Llama, Mistral.

Their key characteristic is emergence: capabilities that weren't explicitly programmed but arise from training at scale, such as chain-of-thought reasoning, cross-language translation, or code generation.

Large Language Models (LLMs)

A subset of foundation models specialized in processing and generating text. They use the Transformer architecture with attention mechanisms that allow them to capture long-range relationships in text sequences.

Current LLMs don't just generate text — they can follow complex instructions, maintain context in long conversations, and use external tools when configured to do so.

Generative AI

The application of foundation models to create new content: text, code, images, audio, video. It's the most visible layer for end users and the one that has driven massive adoption since 2022.

Interaction paradigms

The way humans interact with AI systems has evolved rapidly:

ParadigmMechanismMain limitationExample
PromptingNatural language instructionsDepends on model knowledgeChatGPT, Claude
RAGQueries external sources before respondingQuality depends on retrievalPerplexity, enterprise systems
Tool UseInvokes APIs, databases, or servicesRequires defining available toolsFunction calling, MCP
AgentsReasoning + memory + tools in multiple stepsOrchestration complexity and safetyCopilot Workspace, Devin

Each paradigm builds on the previous one. AI agents represent the current frontier.

Practical considerations

  • Hallucinations: LLMs generate plausible but not necessarily correct text. All output must be verified.
  • Context window: models have a token limit they can process. This directly affects how long a conversation or document can be.
  • Cost: more capable models are more expensive per token. Model choice must balance capability with budget.
  • Latency: text generation is sequential (token by token). For real-time applications, this is a limiting factor.

Why it matters

AI has moved from a research field to an engineering tool. Understanding its layers — from foundation models to agent frameworks — enables informed decisions about what to build, what to buy, and where to invest learning effort.

References

  • Attention Is All You Need — Vaswani et al., 2017. The paper that introduced the Transformer architecture.
  • On the Opportunities and Risks of Foundation Models — Stanford CRFM, 2021. Foundational report on large-scale models.
  • Anthropic — Creators of Claude and the Model Context Protocol.
  • OpenAI — Creators of GPT-4 and pioneers in commercial generative AI.

Related content

  • AI Agents

    Autonomous systems that combine language models with reasoning, memory, and tool use to execute complex multi-step tasks with minimal human intervention.

  • Model Context Protocol (MCP)

    Open protocol created by Anthropic that standardizes how AI applications connect with external tools, data, and services through a universal interface.

  • Neural Networks

    Computational models inspired by brain structure that learn patterns from data, forming the foundation of modern artificial intelligence systems.

  • Takeaways: Why 2026 Is the Year to Build a Second Brain

    Key takeaways from Nate B Jones' second brain series — from the original 8 building blocks to Open Brain (Postgres + MCP), the two-door principle, and the implementation gap.

  • Takeaways: The Renaissance Developer — Dr. Werner Vogels

    Key takeaways from Dr. Werner Vogels' final keynote at AWS re:Invent 2025, where he presents the Renaissance Developer framework and argues why AI will not replace developers who evolve.

  • Takeaways: The Adolescence of Technology

    Key takeaways from Dario Amodei's essay on civilizational risks of powerful AI and how to confront them.

  • Building a Second Brain in Public

    Chronicle of building a second brain with a knowledge graph, bilingual pipeline, and agent endpoints — in days, not weeks, and what that teaches about the gap between theory and working systems.

  • Strands Agents

    Open source SDK from AWS for building AI agents with a model-driven approach. Functional agents in a few lines of code, with multi-model support, custom tools, MCP, multi-agent, and built-in observability.

  • Spec-Driven Development

    Development methodology where the specification is written before the code, serving as a contract between teams and as the source of truth for implementation.

  • Semantic Search

    Information retrieval technique that uses vector embeddings to find results by meaning, not just exact keyword matching.

  • Prompt Engineering

    The discipline of designing effective instructions for language models, combining clarity, structure, and examples to obtain consistent, high-quality responses.

  • Maturity Models

    Structured frameworks for progressively assessing and improving organizational capabilities, from CMMI to modern approaches like DORA and simplified models.

  • llms.txt

    Proposed standard for publishing a Markdown file at a website's root that enables language models to efficiently understand and use the site's content at inference time.

  • Large Language Models

    Massive neural networks based on the Transformer architecture, trained on enormous text corpora to understand and generate natural language with emergent capabilities like reasoning, translation, and code generation.

  • Knowledge Graphs

    Data structures representing knowledge as networks of entities and relationships, enabling reasoning, connection discovery, and semantic queries over complex domains.

  • Fine-Tuning

    Process of specializing a pre-trained model for a specific task or domain through additional training with curated data, adapting its behavior without starting from scratch.

  • AI Evaluation Metrics

    Frameworks and metrics for measuring AI system performance, quality, and safety, from standard benchmarks to domain-specific evaluations.

  • AI Safety

    Field dedicated to ensuring artificial intelligence systems behave safely, aligned with human values, and predictably, minimizing risks of harm.

Concepts