Books | Free AI Books by Dweve

A beginner's map to the ideas, tools, and trade-offs behind modern AI. What AI really is, the mathematics it runs on, what learning means, and the path from classical machine learning through neural networks, transformers, and large language models to ethics and the future. Written for a curious reader who wants the shape of the field, not the headlines. Last verified May 2026.

Classical machine learning, built by hand

The workhorses behind decades of AI, worked end to end. The book covers linear and logistic regression, regularisation, trees and forests, gradient boosting, support vector machines, neighbours and probabilities, neural networks, clustering, and dimensionality reduction. It closes with the leak-proof pipeline from raw data to a production model.

A history of artificial intelligence, told through its people

The history of AI, from mechanical minds and forgotten founders to the birth of the field, its winters, the connectionist underground, the deep learning explosion, and the transformer era. The book arrives at today's major debates and the choices ahead, with the past in view.

Binary neural networks, from first principles to deployment

Binarised neural networks, Tsetlin machines, logic gate networks, and sub-bit compression. The binary imperative, the genesis of BNNs, the architecture wars, learning beyond gradients, the hardware story, what worked and what failed, and where the research goes next. The efficiency frontier of modern AI, in one volume.

A practical guide to data-parallel performance

A practical guide to SIMD, SIMT, compiler behaviour, graph IRs, memory traffic, benchmarking, and kernel design. It follows data-parallel work from the mental model and roofline through architectures, tools, runnable kernels, and shipping.

An investigation into European digital sovereignty

A living report on the gap between digital-sovereignty language and the systems, contracts, vendors, cloud dependencies, and policy choices that decide actual control.

A systems investigation into Dutch road safety

This report looks at traffic safety: where AI can help, where automation claims become dangerous, and what evidence public infrastructure should require before systems touch roads.

Eight published works, free to read in English and Dutch. The books build understanding from first principles. The reports follow evidence through public questions where easy claims are not enough. No account is required.

The reading guides are one series of fifteen volumes, from a citizen’s first map of the field through to reinforcement learning and multi-agent systems. Four are published and free to read in the browser, and they do not assume an order: begin with the question you actually have. The rest are being written, and each lands the same way, in the browser, in English and Dutch, with nothing to install.

Ground truth is the broadest starting point. It explains the vocabulary, mathematics, learning methods, modern model families, and social questions in one route through the field.

The other three books go deeper without assuming an order. The Builder works through classical machine learning. The Archivist follows AI's history through its people. The Essential examines binary and sub-bit approaches.

Four volumes published of fifteen, each available in English and Dutch.

I want the history behind today's debate

Every route opens directly in the reader.

Start anywhere. Each book stands on its own.

SIMD is for readers who want to work close to execution. It connects the mental model and roofline to x86, ARM, other instruction sets, GPUs, compiler behaviour, language support, and practical kernels. It is written for a reader who already writes code and wants the measurement that goes with it, from the roofline down to a kernel you can run.

The opening chapters establish what data-parallel execution can and cannot speed up. Architecture chapters then compare the routes available across CPUs, GPUs, and related hardware.

The later chapters move into compilers, language mechanisms, kernel design, synthesis, and shipping. Runnable examples and kernel listings keep the discussion tied to work a reader can inspect.

One published engineering field guide, available in English and Dutch.

The appendices include complete kernel listings.

One operation, many lanes, measured against the machine

CPU, GPU, and NPU paths are covered in the browser reader.

The report shelf is separate from the guides. Each report takes a public question, defines the system around it, and follows the evidence through law, institutions, infrastructure, incentives, and practical consequences.

The Sovereignty Illusion examines who controls Europe's digital systems. The Compliance Illusion compares EU AI Act duties with the compliance evidence published by the market. Both follow the contracts, institutions and decisions behind the headline claim.

Traffic Safety in the Netherlands studies road safety as a system, including evidence, human behaviour, institutions, policy choices, and the limits of automation claims. Fourteen chapters ask what evidence public infrastructure should require before systems touch roads.

Three published investigations, all available in English and Dutch.

Who has practical control over Europe's digital systems?

What would make Dutch roads safer as a whole system?

The claim is the beginning, not the conclusion

All three reports include their references in the reader.

Published reports with their evidence trails intact

Open a report to read its chapters, timeline, glossary, and bibliography.

Deep learning for vision, sequences, generation, and use

Transformers, LLMs, alignment, and reasoning models

Data governance and responsible handling

Deploying and operating AI in production

Reinforcement learning and multi-agent systems

Published volumes open in the reader. The rest are in the works.

Nothing to install, no account, no download. Pick a question, open the volume, and read it in English or Dutch. The reports sit alongside the guides and read the same way.

The guides assume no order. Start where your question is, not at volume one.

No install, no account, no download. The reader opens on the chapter.

Every published volume is written in English and Dutch.

The EU AI Act arrived with obligations, deadlines and fines. This dossier reads the law against what the market actually does: who is covered, who is exempt, what the transparency and systemic-risk duties require in practice, and where enforcement has teeth. Twenty-one chapters, every claim sourced.

Sources, claims and verification for every figure, 1.9 MB