Room at the bottom

One view of AI coding is that it can help you generate 10x more code or 10x more features.  As Dare and others point out, we are starting to see the failings of that idea — for many use cases, we don’t need 10x more code or 10x more features, and in fact, they might slow us down — we really need to be asking if this additional work will bring in substantially more users or revenue. 

There is another way to look at AI — it allows us to scale up, but also to scale down. Sure, you can generate 10x code, but you can also generate 1x code in 1/10th the time, at 1/10th the cost, and with 1/10th the skill base. This allows several things:

• Software for much smaller user bases — for your department, your workgroup, your family, or just for yourself. I used Claude this week to write a tax estimate app for my accountant and me, and a book review rollup app for my college roommate and me.  The cost of development is so low that I don’t care if only a couple of people use the app.  I used to do some of these types of apps in Excel, but it is way easier to do them in Claude now. 
• Ephemeral software.   Software with a limited life — for an event, a single analysis, a meeting, whatever.  You are going to throw the software away, and you don’t care because the cost to build it was low.
• Software developed by people with much less software experience or training.  Anyone can build a simple app or website now.  And if it has bugs or other issues, who cares? You are going to throw it away anyway.

All the press coverage is focused on the scale-up drama — massive capital raises, giant IPOs, competitive battles between titans of industry, and huge data center buildouts.  But there is plenty of room at the bottom, and I wonder if that won’t be the more interesting story.

Shorts

The 5 books that could save America — I’ve read Abundance, and it is great.  It cuts across liberal and conservative positions; we need more thinking like this.

The Atlas of Innovation is a nice way to understand the different mechanisms we can use to encourage R&D.  I didn’t know much about fieldbuilding; the materials science example was interesting.

Dickovers — these have become so omnipresent on the web, even on sites that are respectable and should know better.

Choose good quests — I had not stumbled across this before, but I love it.  Your life and working career are short; it is worth thinking explicitly about what quests you want to spend them on.