Skip to content

Essays

  • 4 min read

It Takes Two Neurons To Ride a Bicycle

Source: Fermat's Library — annotated version
Date Published: 2004 (NIPS 2004)
Author: Matthew Cook (California Institute of Technology)

TL;DR

A classic paper from NIPS 2004 demonstrating that a remarkably simple two-neuron network can control a bicycle in a desired direction. The network achieves accurate long-range goal following while stability issues dominate short-run behavior — not by explicit design, but as a natural consequence of how the network controls the bicycle.

  • 3 min read

Using AI to Write Better Code More Slowly

Source: Using AI to Write Better Code More Slowly
Date Published: 2026-05-25
Author: Nolan Lawson

TL;DR

Contrary to the dominant narrative that AI coding tools are "slop cannons" for generating low-quality code quickly, Nolan argues that LLMs are extremely flexible and can be used just as effectively to write high-quality code more slowly. The key technique is multi-model PR review — running several different LLM agents to find bugs, then carefully prioritizing and validating their findings.

  • 2 min read

Webwright: A SWE-Style Browser Agent Framework

Source: Microsoft/Webwright on GitHub
Date Published: 2026-05-26
Author: Microsoft Research

TL;DR

Webwright is a radically simple browser agent framework that achieves state-of-the-art results on long-horizon web tasks. Its key insight: separate the agent from the browser, treating the browser as a disposable environment the agent spawns while developing a program. No multi-agent systems, no graph engines, no plugin layers — just a terminal, a browser, and a model.

  • 3 min read

America Keeps Falling Uphill — Beijing's Big Blind Spot

Source: The Washington Post — David Ignatius
Author: David Ignatius
Date: May 21, 2026

TL;DR

Washington Post columnist David Ignatius offers a post-summit analysis of relative US-China strength, framed as an imagined intelligence briefing for China's Ministry of State Security. While the Trump-Xi summit appeared to favour Beijing on optics, Ignatius argues that US structural advantages — a resilient and adaptable economy, deep alliance networks, and unmatched innovation capacity — are consistently underestimated by Chinese planners. America has a habit of "falling uphill" in ways Beijing's strategic models fail to capture.

  • 2 min read

Digital Design & Computer Architecture — Cache Design, Management, and Tradeoffs

Source: YouTube — Onur Mutlu Lectures
Course: Digital Design and Computer Architecture, ETH Zürich, Spring 2026
Lecturer: Professor Onur Mutlu
Lecture: L23 — Cache Design, Management, and Tradeoffs
Course URL: safari.ethz.ch/ddca/spring2026/

TL;DR

Lecture 23 of Onur Mutlu's Digital Design and Computer Architecture course at ETH Zürich covers cache memory design principles: how caches are organised, the tradeoffs between capacity, associativity, block size, and latency, and management strategies for maximising hit rates. The lecture is part of a comprehensive Spring 2026 series spanning ISA, microarchitecture, memory hierarchy, and advanced topics.

  • 2 min read

China's $200B Equity Outflow: Brad Setser on Capital Flight and the $1T Question

Source: Brad Setser on X
Date: May 25, 2026
Author: Brad Setser (Senior Fellow, Council on Foreign Relations; former US Treasury economist)

TL;DR

In a thread reacting to reports of $1T total capital outflows from China, Brad Setser highlights that equity outflow was ~$200B — large but not unmanageable. Relative to GDP, it is far less severe than the capital flight from South Korea during the Asian Financial Crisis. The thread also quotes Richard Casey's reporting on China tightening capital controls despite maintaining a large current account surplus and what Beijing considers an undervalued currency.

  • 3 min read

China's Housing Slump: Signs of Bottoming Out, or Another Lull?

Source: The New York Times — Keith Bradsher
Author: Keith Bradsher
Date: May 21, 2026
Location: Beijing and Yancheng, China

TL;DR

After a 38% plunge since 2021, average existing home prices in China's tier-1 cities (Beijing, Shanghai, Shenzhen, Guangzhou) ticked up 2% from February through April. But with 90 million empty or unfinished apartments nationwide, analysts are split: this could be a genuine bottom, or another "dead cat bounce" before the next leg down. Middle-class savings have been devastated — families who poured their life savings into property now face 30%+ losses. Previous stabilisations during the multi-year crash proved short-lived.

  • 3 min read

Figma's FigCache: Next-Generation Data Caching at Scale

Source: Figma Blog
Author: Figma Engineering

TL;DR

As Redis became critical-path infrastructure at Figma, it hit scalability and reliability limits — connection bottlenecks, thundering herds, observability gaps, and data pollution risks. Figma built FigCache: a stateless, RESP-wire-protocol proxy acting as a unified Redis data plane. The architecture features a decoupled frontend/backend design, configuration-driven routing via Starlark, and custom extensions to the Redis protocol. Since its rollout to Figma's main API service in H2 2025, the caching layer has achieved six nines of uptime.

  • 4 min read

The Hall–Héroult Process: How Aluminium Went From Precious Metal to Ubiquitous Commodity

Source: Wikipedia — Hall–Héroult Process
Focus: Economic impact

TL;DR

The Hall–Héroult process is the industrial method for smelting aluminium: alumina (Al₂O₃ from bauxite) is dissolved in molten cryolite and electrolysed at 940–980°C. Before its simultaneous 1886 discovery by two 22-year-olds — Charles Martin Hall (USA) and Paul Héroult (France) — aluminium was more expensive than gold or silver, reserved for emperors. The process collapsed production costs, transformed Al from a precious metal into a dirt-cheap commodity, and enabled the modern age of aviation, construction, packaging, and electronics. But it remains energy-intensive (~15.4 kWh/kg) and a major CO₂ source (~12.7t per ton Al).

  • 4 min read

Huawei's τ Scaling Law: Reframing Semiconductor Competition from Geometry to Time

Source: Global Semi Research (Substack)
Date: May 25, 2026

TL;DR

Huawei's announcement of Tau (τ) Scaling Law is analysed as a strategic move to reframe semiconductor competition from geometric scaling to time-domain optimisation — a battlefield that plays to Huawei's strengths given restricted access to cutting-edge EUV lithography. The τ framework spans 12 orders of magnitude (transistor ps → system s). Its first major engineering implementation is LogicFolding in the Kirin 2026 chip, achieving +55% transistor density and +41% energy efficiency at a fixed process node — no new lithography needed.