Articles

transistor · bell-labs

The Thinking Machine Chronicles #0010: The Transistor: The Tiny Device That Ended the Vacuum Tube Age

In December 1947, three physicists at Bell Laboratories assembled a device smaller than the palm of a hand that could amplify electrical signals without vacuum, without heat, without any moving parts. The transistor did not just replace the vacuum tube: it redefined the physical limits of what a computer could be.

eniac · early-computers

The Thinking Machine Chronicles #0009: ENIAC: Eighteen Thousand Vacuum Tubes and the Dawn of Electronic Computing

On 14 February 1946, the US Army unveiled a machine the size of a tennis court that could do in one second what took a human mathematician twenty hours. ENIAC was not the first electronic computer: Colossus was: but it was the first one the world got to see, and it changed what people believed was possible.

memex · vannevar-bush

The Thinking Machine Chronicles #0008: As We May Think: Vannevar Bush Dreams of the Thinking Machine's Memory

In July 1945, the man who had run American science through the entire war published a vision of a machine that would extend the human mind itself: not a calculator, but a memory, a library, and a thinking aid all in one. Nobody could build it yet. But everyone who mattered read it.

von-neumann · architecture

The Thinking Machine Chronicles #0007: The Blueprint of Every Computer Ever Built: Von Neumann Architecture

On 30 June 1945, a Hungarian-American mathematician circulated a twenty-three-page document that described, for the first time, a machine in which program instructions and data live in the same memory. Every computer built since has followed that blueprint.

colossus · early-computers

The Thinking Machine Chronicles #0006: Colossus: The Secret Electronic Brain That Cracked Lorenz

In 1943, a Post Office engineer named Tommy Flowers built the world's first large-scale programmable electronic computer in a condemned London factory. It remained secret for thirty years. This is the story of Colossus.

neural-networks · mcculloch-pitts

The Thinking Machine Chronicles #0005: The First Artificial Neuron: McCulloch and Pitts Describe a Logical Brain

In 1943, a neurophysiologist and a self-taught logician published a four-page paper that founded computational neuroscience and planted the seed for every neural network ever built. Warren McCulloch and Walter Pitts showed that a network of binary threshold neurons could compute any logical function: and therefore, in principle, anything a brain could compute.

turing · cryptography

The Thinking Machine Chronicles #0004: The Machine That Won a War: Turing's Bombe and the Breaking of Enigma

In 1940, Alan Turing designed the Bombe: an electro-mechanical device that exploited a fatal mathematical flaw in the Enigma cipher. By the war's end, over 200 Bombes were running at Bletchley Park, reading the Wehrmacht's mail in near real time.

zuse · z1

The Thinking Machine Chronicles #0003: Gears and Logic: Konrad Zuse Builds the World's First Programmable Computer

In 1938, a German civil engineering student named Konrad Zuse assembled the world's first programmable computer from sheet metal and discarded film strips in his parents' living room. This is the story of Z1: mechanical binary arithmetic before transistors existed.

shannon · boolean-logic

The Thinking Machine Chronicles #0002: Electricity Learns to Think: Shannon's Boolean Algebra for Circuits

In 1937, MIT master's student Claude Shannon showed that Boolean algebra, an abstract 19th-century mathematics, maps perfectly onto electrical switching circuits. Every digital computer ever built rests on this 71-page thesis.

subq · sparse-attention

Inside SubQ: How a Fully Sub-Quadratic Sparse-Attention Architecture Actually Works

Standard transformers compare every token to every other token — that's O(n²) compute. SubQ's Sparse Attention Architecture (SSA) finds which of those relationships actually matter and attends only to those, achieving near-linear scaling. This article digs into how that works, why all previous attempts compromised on quality, and what SubQ's benchmark numbers tell us about whether they've finally cracked it.

turing · computation

The Thinking Machine Chronicles #0001: The Paper That Started Everything: Turing's "On Computable Numbers"

In 1936, a 24-year-old Cambridge postgraduate named Alan Turing solved Hilbert's Entscheidungsproblem and, in doing so, described a theoretical machine that is the ancestor of every computer ever built.

machine-learning · neural-networks

How Neural Networks Actually Learn: Backpropagation Explained

Most tutorials hand-wave over backprop. This one walks through the actual math — gradients, the chain rule, and why any of it works — with code you can run.

machine-learning · transformers

Understanding Transformers: The Architecture Behind Modern AI

Attention mechanisms, positional encoding, why the 'query-key-value' metaphor actually makes sense, and how a transformer processes a sentence — step by step.

machine-learning · optimization

Gradient Descent: The Algorithm That Trains Every Model

From the basic idea of walking downhill on a loss surface, to momentum, Adam, learning rate schedules, and the practical choices that actually matter when training models.

bert · python

Fine-Tuning BERT for Text Classification with Python and HuggingFace

A practical walkthrough of fine-tuning bert-base-uncased on a sentiment dataset using the HuggingFace Trainer API — tokenisation, training loop, and evaluation in under 60 lines.

nlp · transformers

BERT and the Attention Mechanism: How Transformers Actually Work

What makes BERT different from earlier NLP models, how masked language modelling works under the hood, and why bidirectional attention changes everything.

spring · java

Spring Framework: Why Dependency Injection Changed Java Development

Rod Johnson published the ideas behind Spring in his 2002 book and released the framework in 2003. Here is what dependency injection solved and why it made Java applications dramatically easier to test.

hibernate · java

Hibernate: Mapping Java Objects to Database Tables

Hibernate arrived in 2001 and solved the object-relational impedance mismatch that entity beans had failed to solve. Here is what it looked like to use it and why it changed Java persistence.

struts · java

Struts: MVC for Java Web Apps Before Spring MVC

Apache Struts brought the Model-View-Controller pattern to Java web development in 2000. Here is what building a web application with Struts looked like and why Spring MVC eventually replaced it.

ejb · j2ee

EJB Entity Beans: Persistence Before Hibernate Made It Bearable

Entity beans were the official J2EE approach to object persistence. Here is why they were painful to use and what we did instead.