Tales of Industry

LLMs Are Done. What Are Agents?

Thu, 11 Sep 2025 00:00:00 +0000

Signs that the LLM paradigm has run its course.

How to Build a Production-Ready Application in Two Days

Thu, 03 Jul 2025 00:00:00 +0000

How do you build a production-ready app in only two days?
Do it four times.

Yes, that’s technically eight days! But hear me out. In my previous post I argued that Building Software Is No Longer a Nightmare, because AI coding assistants can take care of all the annoying peripheral tasks like packaging, testing, documentation, deployment, etc.

Here I want to propose that AI-assisted coding makes it easier than ever to return to the ultimate form of software engineering: “Throwaway Prototyping”. This concept was pioneered by Fred Brooks in his classic book “The Mythical Man-Month”¹, where he coined the phrase, “Plan to throw one away; you will, anyhow.” Brooks refers specifically to the first iteration of a system – the prototype. I will expand this definition to the practice of discarding code and architecture multiple times till the problem and its solution are sufficiently understood.

Building Software Is No Longer a Nightmare

Mon, 16 Jun 2025 00:00:00 +0000

Will Large Language Models make us better coders or worse? While this questions will take time to settle, one thing it clear: LLMs have finally broken down the barriers to creating and publishing open source. How many projects took off quickly with the implementation of the core logic only to run into a wall once it came to “all the other stuff”, such as:

Version control & hosting
Packaging
Publication
Testing
Documentation
UI/UX

Your experience might differ, maybe you’re a git-whiz who is cranking out command-line-foo like there’s no tomorrow. Or maybe you honed the perfect setup to scaffold and automate tasks like packaging or testing over the years. Lucky you! But the majority of developers are grinding every day through an ungodly mess of confusing git commands, dependency issues, and contradictory testing practices. And I won’t even go into the hell that is frontend development for the sake of our collective sanity ¹.

MLTO Talk: What I learned about AI Product Development

Mon, 30 Sep 2024 00:00:00 +0000

My talk at the Machine Learning Toronto (MLTO) Meetup on 30 Sep 2024.

Three Kinds of Diversity

Sat, 10 Feb 2024 00:00:00 +0000

Beyond Quotas: Nurturing True Diversity in the AI Industry for Innovation and Inclusion

AI Product Development (Part III)

Sat, 30 Jul 2022 00:00:00 +0000

How to avoid common AI pitfalls.

AI Product Development (Part II)

Wed, 01 Jun 2022 00:00:00 +0000

Build a product with the right people and process.

AI Product Development (Part I)

Sun, 08 May 2022 00:00:00 +0000

Build a product, not a model.

About

Sun, 01 May 2022 00:00:00 +0000

Who am I?

I’m Jan. In my past life as a Neuroscientist I studied how the brain learns. This peaked my interest into complex dynamical systems, what laws govern them, and how they can be improved.

After transitioning into the Data Science and Machine Learning (ML) industry, I become fascinated by the complex dynamics that play out when data, software, and humans come together to build products.

All machine learning challenges are technical challenges. All technical challenges are people challenges.

Disclaimer

Sun, 17 Apr 2022 00:00:00 +0000

None of the content on this site represents the conditions, strategies, or opinions at previous or current employers. Descriptions of individuals, organizations, and events are fictional and often exaggerated to support arguments and contrasting views.

How To Create This Blog

Sun, 20 Mar 2022 00:00:00 +0000

Getting Started with Hugo

Hugo is a static site generator (SSG). Simply put, an SSG prepares everything needed to display a website. In contrast, dynamically assembled websites obtain their assets only after they get accessed. An SSG thus minimizes the computational demand on the server, which allows providers, such as GitHub Pages and others¹, to host your static sites for free. More importantly, SSGs emanate a philosophy of simplicity-by-design stemming from a number of limitations. Being constraint forces content creators to focus on the content itself and minimizes the distractions of layout and design. SSGs are thus especially suited for blogs, landing pages, documentation and portfolio sites.

Data Science Anti-Patterns: Proliferating Data Artifacts

Tue, 16 Nov 2021 00:00:00 +0000

Use pipelines to fend off the curse of proliferating data artifacts.

Three Rules of Success for Data Science Projects

Tue, 03 Aug 2021 00:00:00 +0000

Data science projects are people projects.