Skip to content

Writing and mumblings

Tips for probabilistic software

This writing stems from my experience advising a few startups, particularly smaller ones with plenty of junior software engineers trying to transition into machine learning and related fields. From this work, I've noticed three topics that I want to address. My aim is that, by the end of this article, these younger developers will be equipped with key questions they can ask themselves to improve their ability to make decisions under uncertainty.

  1. Could an experiment just answer my questions?
  2. What specific improvements am I measuring?
  3. How will the result help me make a decision?
  4. Under what conditions will I reevaluate if results are not positive?
  5. Can I use the results to update my mental model and plan future work?

Public Baths

Going to American baths is just so weird. I spent my summer in Japan visiting different onsens, and it was both a natural and spiritual experience. Before entering the water, everyone would bathe in the front, and kids would learn from their dads how to bathe. I would often sit on the edges of cliffs, gazing at the water or the sunrise, and it felt like we were monkeys, freely splashing about in nature.

In contrast, the time I spent in LA or New York City at various bathhouses was different. No one looked like an animal; instead, everyone seemed focused on optimization. People barely bathed before entering the water, wearing their dirty little speedos and swim trunks that they had definitely peed in the month before.

Gross.

Anatomy of a Tweet

The goal of this post is basically to share what I have learned about writing a tweet, how to think about writing a hook, and a few comments on how the body and the cta needs to retain and reward the user. Its not much, I've only been on twitter for about 6 month.

I used to hate rich people.

This entire piece of writing is dedicated to a recent response on Hacker News. I hope you can see, as a member of reality, that I write this sincerely.

—

Preamble

Also, I wrote this as a speech-to-text conversion. As I mentioned in my advice post about writing more, my measure for writing more is simply putting more words on a page. If you're wondering how I can be so vulnerable, it's the same as what I mentioned about confidence. If you think this comment hurt me remember that you're just a mirror.

I've also learned that writing is a exorcism of your own thoughts. The more I write, the less these thoughts stick around in my head.

Learning to Learn

After writing my post advice for young people, a couple of people asked about my learning process. I could discuss overcoming plateaus or developing mastery, learning for the joy of learning. I could also talk about how to avoid feeling overwhelmed by new topics and break them down into smaller pieces. However, I think that has been done before.

Instead, I'm going to explore a new style. I'm just going to go through a chronological telling of my life and what I learned from just trying new things. I'm going to talk about the tactics and strategies and see how this pans out.

  • in RAG
  • 6 min read

How to build a terrible RAG system

If you've followed my work on RAG systems, you'll know I emphasize treating them as recommendation systems at their core. In this post, we'll explore the concept of inverted thinking to tackle the challenge of building an exceptional RAG system.

What is inverted thinking?

Inverted thinking is a problem-solving approach that flips the perspective. Instead of asking, "How can I build a great RAG system?", we ask, "How could I create the worst possible RAG system?" By identifying potential pitfalls, we can more effectively avoid them and build towards excellence.

This approach aligns with our broader discussion on RAG systems, which you can explore further in our RAG flywheel article and our comprehensive guide on Levels of Complexity in RAG Applications.

  • in RAG
  • 7 min read

With the advent of large language models (LLM), retrieval augmented generation (RAG) has become a hot topic. However throught the past year of helping startups integrate LLMs into their stack I've noticed that the pattern of taking user queries, embedding them, and directly searching a vector store is effectively demoware.

What is RAG?

Retrieval augmented generation (RAG) is a technique that uses an LLM to generate responses, but uses a search backend to augment the generation. In the past year using text embeddings with a vector databases has been the most popular approach I've seen being socialized.

RAG

Simple RAG that embedded the user query and makes a search.

So let's kick things off by examining what I like to call the 'Dumb' RAG Model—a basic setup that's more common than you'd think.

Kojima's Philosophy in LLMs: From Sticks to Ropes

Hideo Kojima's unique perspective on game design, emphasizing empowerment over guidance, offers a striking parallel to the evolving world of Large Language Models (LLMs). Kojima advocates for giving players a rope, not a stick, signifying support that encourages exploration and personal growth. This concept, when applied to LLMs, raises a critical question: Are we merely using these models as tools for straightforward tasks, or are we empowering users to think critically and creatively?

Good LLM Observability is just plain observability

In this post, I aim to demystify the concept of LLM observability. I'll illustrate how everyday tools employed in system monitoring and debugging can be effectively harnessed to enhance AI agents. Using Open Telemetry, we'll delve into creating comprehensive telemetry for intricate agent actions, spanning from question answering to autonomous decision-making.

If you want to learn about my consulting practice check out my expert calls page. If you're interested in working together please reach out to me via email

What is Open Telemetry?

Essentially, Open Telemetry comprises a suite of APIs, tools, and SDKs that facilitate the creation, collection, and exportation of telemetry data (such as metrics, logs, and traces). This data is crucial for analyzing and understanding the performance and behavior of software applications.