Your Agent Has Two Untrusted Inputs
AI agents have two untrusted inputs: code the model writes and tool output it reads. One now has a real WASM sandbox; the other, MCP tool poisoning, does not.
AI & TechnologyWriting
Thoughts on design, development, AI infrastructure, and building products.
Engineering Philosophy: Kyunghyun Cho
Kyunghyun Cho let models learn their own control structure -- gated memory in the GRU, soft alignment in attention -- and paired it with honest, reproducible science.
Engineering & CraftEngineering Philosophy: Joanna Rutkowska
Joanna Rutkowska built Qubes OS on reasonable paranoia: you cannot make software bug-free, so assume compromise, distrust the infrastructure, and isolate the blast radius.
Engineering & CraftEngineering Philosophy: Raj Reddy, Technology for the Bottom Billion
Raj Reddy pioneered machines you can talk to -- speech recognition, the blackboard model -- then aimed the most advanced AI at the world's poorest, not its richest.
Engineering & CraftEngineering Philosophy: Geoffrey Hinton, Conviction Over Fashion
Geoffrey Hinton bet on brain-like neural networks through two AI winters when the field mocked them -- conviction over fashion, intuition over formalism.
Engineering & CraftApple's Translation Framework: Free, On-Device, and Sharper Than It Looks
Apple's Translation framework: free on-device translation via translationPresentation and TranslationSession, plus the offline-download edges demos skip.
AI & TechnologyEngineering Philosophy: Radia Perlman
Radia Perlman invented the Spanning Tree Protocol and built networks that heal themselves -- loop-free, self-stabilizing, and correct even when nodes lie. Robustness as a first principle.
Engineering & CraftEngineering Philosophy: Satoru Iwata
Satoru Iwata was a programmer who became Nintendo's president without ever stopping being a gamer -- technical mastery in service of the player's joy.
Engineering & CraftEngineering Philosophy: Salvatore Sanfilippo (antirez)
Salvatore Sanfilippo built Redis around one principle -- do less. A small, single-threaded, in-memory store you can hold in your head, written in C for humans to read.
Engineering & CraftEngineering Philosophy: Werner Vogels
Werner Vogels, Amazon's CTO, built the cloud on one assumption: everything fails all the time. Design for failure, minimize blast radius, and stay available through it.
Engineering & CraftEngineering Philosophy: Demis Hassabis, Solve Intelligence to Solve Everything
Demis Hassabis built general intelligence from first principles -- games as the proving ground, neuroscience as inspiration -- then pointed it at protein folding and won a Nobel.
Engineering & CraftThe iOS 26 Widget Surface: One App Intent, Many Places
iOS 26 widgets, Control Center controls, and Live Activities are all App Intents surfaces. One intent powers buttons, controls, and Live Activity actions.
AI & TechnologyExternal Publications
Technical writing published at Introl
Local LLM Hardware Pricing Guide 2025
Comprehensive hardware recommendations and cost analysis for running large language models locally.
H100 vs. H200 vs. B200: Choosing the Right NVIDIA GPUs for Your AI Workload
GPU selection guide comparing NVIDIA's latest datacenter accelerators for different AI workloads.
TPU Architecture: Complete Guide to Google's 7 Generations
Deep technical dive into Google's Tensor Processing Unit evolution from TPUv1 to TPUv5.
GPU Memory Pooling in Multi-Tenant Kubernetes
Resource sharing strategies for GPU clusters in containerized environments.
Ray Clusters: Distributed AI Computing Infrastructure
Guide to building and managing distributed AI computing with Ray framework.
DeepSeek V3 Open Source AI Cost Advantage
Analysis of open source LLM economics and DeepSeek's competitive positioning.
NVIDIA Vera Rubin GPU 600kW Racks 2027
Future datacenter power requirements and NVIDIA's next-generation GPU roadmap.
SMR Nuclear Power for AI Data Centers
Small modular reactor solutions for powering next-generation AI infrastructure.
DeepSeek MHC Architecture Breakthrough
Technical analysis of DeepSeek's Multi-Head Compression architecture innovations.
