Over the long weekend I had a couple opportunities to talk with a few friends from different fields (an accountant and a personal trainer). AI came up (as it often does these days) with both. The accountant was uneasy about how it may impact his job and the overall economy. The personal trainer less so. People are always going to want to work with a human on their physical fitness goals.

It got me thinking about my own thoughts on the impacts of AI in my field. It is undoubtedly making it easier to write code. It lowers the barrier to entry and raises the bar for what great engineers can produce. But final execution of code has always been the last stop for a project in an agency setting. What makes or breaks a project is the work at the beginning. The planning, approvals, and design. All the squishy things that don't have a defined answer that need real human judgment to make something great.

I've been reading Build by Tony Fadell, so maybe I've had product development on my mind. The idea that's stuck with me is identifying what to build. What I keep coming back to is building systems around AI tools to help pressure test and prototype ideas faster. Not to do the thinking for them, but to let people push more ideas further — far enough to spot the rough edges and decide which ones are worth refining.

In a recent interview Strauss Zelnick, CEO of Take-Two, talked about his thoughts on AI and creativity. AI is a tool built on data from the past. It, by design, is backward looking. Creativity is forward looking. Data can inform creativity, and an artificial intelligence is another tool in the creative's toolbelt to build something new.

That's why I'm not worried about AI replacing all human workers. Execution is cheaper, sure. We've had execution get cheaper before (see the Industrial Revolution) and work adapted. But knowing what to build is still on us humans.

🎥 Watch
The Man Behind Grand Theft Auto 6: Strauss Zelnick
David Senra
I have to admit. This wasn't the nuanced take I was expecting from the CEO of Take-Two Interactive. The full interview is great, but if you only have a few minutes, listen to the last 5 for Strauss Zelnick's take on AI and creativity.

📖 Read
Build: An Unorthodox Guide to Making Things Worth Making
Tony Fadell
I haven't finished this yet, but plenty of takeaways have stuck with me from the guy that helped bring the iPod, iPhone, and Nest to market. As things are easier to build, Tony Fadell's advice on building the right things seems prescient.

📰 Article
Building Pi With Pi
Armin Ronacher
One of the biggest problems facing open source projects right now is the deluge of slop issues and pull requests. LLMs don't really have a good way of parsing that slop. They'll see the bad PR and take it as fact and make their own LLM guided assumptions on top of it rather than fixing the overall system.

🎥 Watch
How to Build a Self-Improving Company with AI
Tom Blomfield (YC)
Maybe a bit of the antithesis to today's issue. Tom Blomfield, YC General Partner, talks about what a self-improving "AI-native" company actually looks like.

🔧 Tool
ZeroEntropy
I'm less interested in the specifics of the company and more interested in the training of custom models as a service. Having models trained on your own data seems like a next evolution for models. And could build nicely on the last "AI-native" company.

Thanks for reading,
Jason