Writing

Essays, notes, lessons.

What I'm learning, in public.

Curated long-form on AI, finance, distributed systems, and forward-deployed engineering — sorted newest first.

Essays · 20

2026-05-02

Writing my Neovim config from scratch

A two-part Medium series on building a Neovim setup from a blank init.lua — bootstrap, options, keymaps, lazy.nvim, then the plugin layer that makes it a real editor.

ToolsEngineeringNeovim

2026-04-28

The Complete AI Hardware Stack — Layer by Layer

From silicon atoms to generated tokens — every layer that makes modern AI inference possible, and where the bottlenecks hide. With deep dives on SRAM vs HBM, co-packaged optics, and EDA.

2026-04-26

Why TypeScript Strict (Even for a Personal Site)

The argument for TypeScript strict mode in projects that 'don't need it,' from someone who just migrated a personal site.

softwareTypeScript

2026-04-26

Why This Site Exists: O-1 Evidence Hub

The site is a personal portfolio, but it is also a deliberately built canonical evidence hub for an O-1 extraordinary-ability visa application.

Meta

2026-04-15

The FDE Feedback Loop

Forward deployed engineering compounds when each engagement makes the next one easier. The loop has four steps.

AIVenture

2026-04-10

Claude Code as an Operating System

Why I think of Claude Code as an OS for engineering work — skills as installed apps, hooks as services, MCP as syscalls.

softwaretools

2026-03-09

AI Hardware Stack Deep Dive

From TSMC N3 shortages to HBM constraints — what the silicon shortage means for frontier AI.

AIHardware

2026-02-14

Why Forward Deployed Engineering Beats Consulting

The unit economics of FDE vs. traditional consulting, and why Palantir cracked it before anyone else.

VentureAI

2026-01-18

Aggregation Theory in 2026

Ben Thompson's aggregation theory, applied to AI dev tools, CRM, and the next category of platform consolidation.

financestrategy

2025-12-04

Optimizer Intuition: SGD → Adam → Lion

The 60-second mental model for the optimizer ladder you actually need to know.

2025-09-01

SpatialDINO Lessons: 3D SSL on Lattice Light-Sheet Microscopy

A detailed log from building SpatialDINO — a 3D self-supervised vision transformer for label-free segmentation and tracking of subcellular dynamics in lattice light-sheet microscopy.

2025-08-30

Convergence Intuition: Why Adam Works

The geometry of why per-parameter learning rates dominate vanilla SGD on transformers.

2025-06-10

FSDP vs Tensor Parallel: Which Sharding Strategy

When to reach for FSDP, when to reach for tensor parallelism, and why the answer depends entirely on your interconnect.

2025-03-15

Raft in Five Minutes

Raft is Paxos that you can actually explain at a whiteboard. The mental model in five minutes.

2024-10-22

RocksDB Write Amplification

Why your RocksDB-backed database writes 10× the bytes you asked it to, and what the levers are.

2024-09-01

Why I Left Physics

After my first paper in chemical physics I didn't pursue a PhD. Here's the honest reason.

physicscareer

2024-08-12

Flink Checkpointing — What Actually Goes Wrong

A field guide to the failure modes I hit during my BU thesis on adaptive Flink checkpointing.

2024-05-02

Adaptive Checkpointing in Apache Flink

A summary of my undergraduate thesis on driving checkpoint cadence from backpressure signals.

Distributed Systems

2023-08-21

A Merkle Tree POC in Rust — ZeroSync, 2023

A from-scratch Merkle tree in Rust over JSON records, built as a proof-of-concept for tamper-evident sync at ZeroSync. SHA-256, deterministic field hashing, and what the borrow checker forced me to learn.

2022-11-15

Supercritical Fluids: First Paper, in Plain English

What we actually did in the 2022 Journal of Chemical Physics paper, without the jargon.

physicsresearch