"Subscribed for quick summaries of topics. Saves time. Googling and Youtube and collating info isn't convenient."
— RB
Making practical sense of software and AI.
How to use open-weight AI models
Picking a model, a harness, and where to run the thing.
What's going on with open-weight models?
How open-weight models went from a curiosity to great in 2 short years.
How do "computer use" agents work?
Why it's surprisingly challenging to let an LLM use a computer for you.
Post-Training
Post-training turns a model from a knowledgeable blob that produces rambling answers, into a helpful assistant.
RAG
Retrieval Augmented Generation (RAG) is a way to make LLMs like GPT-4 more accurate and personalized to your specific data
AI agent
An AI agent is a model plus a harness — the model decides what to do, and the harness gives it the tools and the loop to actually go do it.
Model weights
Model weights are the numbers a model learned during training — an astoundingly unsexy name for the single most important component of an AI model.
Databricks is apparently worth $100B. What do they even do?
What we should really be asking is “What does Databricks not do?”
The Modern Data Stack
The Modern Data Stack™ (MDS) is a new-ish set of tools that data teams are using to collect, transform, explore, and make use of their company’s data.
Metric
When developers monitor their applications and infrastructure to make sure things run smoothly, they set up metrics to track specific points of performance: how fast an API request gets fulfilled, the percentage of requests that fail, or even what % of total processing capacity is a server using up.
Databricks
Databricks sells a data science and analytics platform built on top of an open source package called Apache Spark.
OpenAI
OpenAI is the most popular provider of generative AI models like GPT-4.
Software Eng for Vibe Coders: Databases + Storage
Where your vibe coded app actually keeps its data, and how it might break when real users show up.
What does Turbopuffer do?
And what's going on in the category formerly known as vector databases?
SQL
SQL, or structured query language, is a type of programming language for working with databases.
Production database
The largest category of databases – both in terms of number of available options and total market size – is production databases.
Confluent
Apache Kafka is a framework for streaming real time data, and Confluent offers Kafka as a managed service.
Elastic
Elasticsearch is a popular open source database for storing and searching unstructured data.
Software Eng for Vibe Coders: On Frontends + Backends
A new series to help non-engineers build products that can handle going viral.
What are code sandboxes?
Why coding agents need a safe place to play, just like we all do.
All about Infrastructure as Code
The deceptively simple text files that help you avoid huge mistakes when shipping software.
Serverless
Serverless is a catch-all term for apps and products that don’t require you to manage your own servers.
Remote
Remote usually means a computer or server that’s not physically located with you at the moment.
Vercel
Vercel builds the AI cloud – they make it easy for engineers to deploy and run the user facing (frontend) + AI parts of their applications.
Zapier
Zapier helps business people make custom integrations between their favorite tools, without needing to write any code.
VPN
A Virtual Private Network (VPN) lets you route your internet access through a specialized server that keeps your sensitive information private.
Tokenization
Tokenization is one of the ways that backends protect sensitive information, like credit cards or social security numbers.
Okta
Okta is an enterprise-focused identity provider: they take care of managing usernames, passwords, and permissions.
Snyk
Snyk helps developers make sure that the code they're writing is secure.
"Technically has allowed me better bridge the gap between user problems and how our system works under the hood."
— Adelaide Hallowell
How to make sure anyone on the internet can't just delete your app's data, and other security considerations.
David KrevittPicking a model, a harness, and where to run the thing.
Will RaphaelsonWhere your vibe coded app actually keeps its data, and how it might break when real users show up.
David KrevittHow open-weight models went from a curiosity to great in 2 short years.
Will Raphaelson
Why it's surprisingly challenging to let an LLM use a computer for you.
Tereza Tizkova
The fast, safe, super difficult to write programming language that’s finally getting its flowers.
Will Raphaelson
If AI models are a commodity, the harness is the thing you want to own.
Paul IusztinWhat their products actually do and why they're so valuable.
Quick explanations of the concepts you see every day.

Kalshi just raised a new round at an $11B valuation. How much revenue does their prediction market really make?
Sam Schneider
What's the forward deployed engineer role, and why is every startup hiring them?
Sung Won ChungCurated collections of foundational articles to guide you through the basics of getting more technical at work.
How to understand and work effectively with AI and ML models and products.
Breaking down what ChatGPT and others are doing under the hood
How AI models learned to stop being weird and start being helpful.
GPT-3 is a Machine Learning model that generates text.
Data Science Notebooks help data teams explore data with code.
The products and business models of leading software companies.
What we should really be asking is “What does Databricks not do?”
Snowflake sells a powerful cloud data warehouse for analytics and data science teams.
OpenAI is the most popular provider of generative AI models like GPT-4.
dbt (no capitals) is a tool for transforming and organizing data in your warehouse.
How to make positive contributions to your product roadmap.
Engineering and code basics that can make you a better PM to work with.
Code is step by step directions, but for computers.
A network of computers all connected to each other and sharing information.
Cloud is how companies rent infrastructure over the internet.
You've outgrown Excel, but you've got to put that data somewhere besides your parents basement.
There are 300+ databases; what do they all do?
If you’re not a data scientist but you have questions, you want to know SQL.
Why that schema change is going to take your engineers two weeks.
Why your feature is held up by a migration and why this is actually good.
Under the hood of today's powerful AI models: how they're trained, implemented, and otherwise used.
How computers learn patterns from data — and why it’s the foundation for everything from stock prediction to ChatGPT.
Breaking down what ChatGPT and others are doing under the hood
A deep dive into how models like ChatGPT get built.
Unlike an onion, hopefully these neural network layers won't make you cry.
Your AI wrote the app. This is everything that happens after — where it runs, how it stays up, and what breaks first when real users show up. A six-part series, made with our friends at Railway.
A new series to help non-engineers build products that can handle going viral.
Where your vibe coded app actually keeps its data, and how it might break when real users show up.
How to make sure anyone on the internet can't just delete your app's data, and other security considerations.
How to make positive contributions when working with analytics and data science teams.
Product analytics is how teams instrument and analyze data about their product usage.
A deep dive into all of the tools that data teams use to do their work.
If you’re not a data scientist but you have questions, you want to know SQL.
The new set of tools data teams use to get their jobs done.