David Aronchick

AI

They Don't Have the Money (And Neither Do You): The Coming Era of Small Models

OpenAI is burning $115B through 2029. If they don't have the money, you definitely don't. The future of production AI isn't trillion-dollar datacenters—it's small, efficient, specialized models running on your infrastructure with your data.

The $100 ChatGPT: Why Karpathy's nanochat Represnts the Next Big Thing

Andrej Karpathy just released nanochat - "The best ChatGPT that $100 can buy." In 4 hours on an 8xH100 node, you get a working ChatGPT clone. Not a toy. An actual LLM that writes stories, answers questions, and attempts math problems. This isn't about edge computing

The Real Silicon War Isn't About GPUs: Why Everyone's Fighting Over Boring Chips

There's a wave swelling up in semiconductors right now. Not the sexy AI chips that suck down kilowatts like a man lost in the desert at an oasis. Not the bleeding-edge 3nm processors that cost more than your house to fab. I'm talking about the chips

AI Adoption

The 800 Million Weekly ChatGPT Users Who Are Just Getting Started

Discover how 800 million weekly ChatGPT users are transforming information interaction and unlocking AI's vast potential for private data usage

AI Philosophy

Ghosts vs. Animals

Explore the debate between AI approaches: 'ghosts' vs. 'animals'. Discover differing views on intelligence, ethics, and future AI development paths

SRE

Why LLMs Can't Replace Your SREs (Yet)

Large Language Models can't replace Site Reliability Engineers yet; AI can augment human expertise for enhanced incident management

AI Infrastructure

The Fairwater Paradox: Microsoft Built a Monster That Needs 900TB/Second of USEFUL Data

Microsoft's Fairwater AI datacenter needs 900TB/second of data, facing challenges in efficiency, storage, and data quality management

data governance

The Anthropic $1.5 Billion Settlement Nobody's Reading Right: Your Enterprise Data is a Ticking Time Bomb

Anthropic's $1.5B settlement underscores data legal risks; prioritize data rights and use secure, governed datasets over external sources

Data Infrastructure

WTF Do You Spend $300B On?: The Oracle-OpenAI Deal Decoded

Discover the challenges and expenses behind OpenAI's $300 billion Oracle deal, revealing a true AI cost breakdown: 80% data prep, 20% training

Cloud Economics

'My Data Transfer Bill Cost What?': When Cloud Economics Go Wrong

Learn strategies to avoid high cloud data transfer fees in this analysis on unexpected cloud economics and costs

ML & AI

Why Your 'AI-Ready' Data Isn't: The Hidden Pipeline Problem Breaking Production AI

Companies spent millions on GPUs and AI talent, only to discover their data pipelines can't actually feed production AI. The revolution isn't waiting for better models—it's waiting for intelligent data pipelines.

Data Movement & Architecture

Your Data Lake is a Write-Only Memory

You're sitting on petabytes of sensor data, but it's faster to put someone in a truck than to query your data lake. Every truck roll that could have been prevented by a query is money burned.

expanso

The History of Expanso (Part 6): Data Controls are a One-Way Street

The below is a continuation of the series on the history of Expanso. Today, we're talking about the third and final unchangeable law of data: the ever-tightening web of data controls. Read the whole series starting from Part 1. The History of Expanso (Part 6): Data Controls are

expanso

The History of Expanso (Part 5): The Unbreakable Speed Limit

The below is a continuation of the series on the history of Expanso. Today, we're talking about the second of the three unchangeable laws of data - the speed limit. Read the whole series starting from Part 1. The History of Expanso (Part 5): The Unbreakable Speed Limit

expanso

The History of Expanso (Part 4): The Mismatch

The below is a continuation of the series on the history of Expanso. Today, we're talking about one of the three unchangeable laws of data - its unrelenting growth. Read the whole series starting from Part 1. The History of Expanso (Part 4): The Mismatch There's

expanso

The History of Expanso (Part 3): You Can't Change The Laws of Physics (Much)

The below is a continuation of the series on the history of Expanso. Today, we're talking about the three unchangeable laws of data - it's exponential growth, the speed of light, and global regulations. Read the whole series starting from Part 1. The Three Unchangeable Laws

expanso

The History of Expanso (Part 2): Our Core Tenets

The below is a continuation of the series on the history of Expanso. Today, we're talking about the tenets of Expanso. Read the whole series starting from Part 1. The Four Tenets of Sane Data Pipelines In the last post, I explained the origin of Expanso: hitting the

expanso

The History of Expanso: It's About the Network (It Always Has Been)

Welcome to our series on the history of Expanso. The Network is the Problem: How Expanso Began The first time the scope of the problem facing tech was super clear, I was at Protocol Labs. We had IPFS storing and serving data at exabyte scale. The storage worked. The catalog

kubernetes

Bridging the Gap: Future Directions for Kubernetes and Distributed Systems

Bridging the Gap: Future Directions for Kubernetes and Distributed Systems When Pokémon GO launched, the world went wild. At Google, we watched as our product, Google Kubernetes Engine, handled a scale we had only theorized about. The game shattered every record for a consumer workload and became a massive success

kubernetes

Global Distributed Consensus: The Missing Piece in Kubernetes

Early in my time on the Kubernetes team, a customer proposed something that was both brilliant and beyond what we were ready for: a global footprint of clusters, one per region, with a synchronized set of jobs. They were running a low-latency application and wanted to coordinate workloads across continents.

techweek

What I Saw at Techweek in Seattle

I kicked off the week with a coffee and a conversation, and the energy was palpable. The first thing that struck me was the sheer scope of ambition. It wasn't just another SaaS/AI week - I saw a week of deep, hard-tech problems being tackled head-on. Ambition

kubernetes

The Distributed Data Dilemma: Kubernetes' Unsolved Puzzle

The Distributed Data Dilemma: Kubernetes' Unsolved Puzzle It was a question I heard constantly in the years after Kubernetes launched: "Can we run our database on Kubernetes?" The answer was always, "Yes, but it's complicated." It wasn't complicated because of Kubernetes.

kubeflow

From Kubeflow to Real-World ML: Why Data Locality Matters Just as Much as Compute

From Kubeflow to Real-World ML: Why Data Locality Matters More Than Compute When my co-founders, Jeremy Lewi, Vishnu Kannan, and I started Kubeflow back in 2017, we were trying to solve what felt like the biggest problem in machine learning. Brilliant data scientists would craft elegant models on their laptops,

kubernetes

The Myth of Portability: Helm and Kubernetes and the Data Pipeline Problem

The Myth of Portability: Helm and Kubernetes and the Data Pipeline Problem I spent years helping companies migrate from bash scripts to Chef, and later to containers. The conversation always started the same way: "We want modern infrastructure, but it has to work with what we have." The

data pipelines

Unlocking Reliability: Why Data Pipelines Need Declarative Deployment & GitOps

Unlocking Reliability: Why Data Pipelines Need Declarative Deployment & GitOps You know the feeling: your data pipeline worked perfectly last week, and now it's throwing cryptic errors. The logs don't help. The documentation is outdated. Nobody's sure who last touched the transformation logic or