What’s coming?

There’s a back and forth in technology on whether the current AI cycle is all hype, or a true revolution that is just at the bottom of an exponential curve. It seems that, like in so many things, a nuanced position is not a popular one. I think that more than “back and forth” there are are actually three camps that compose the current balance of the tech world: AI maximalists, AI doomers, and AI doubters.

The Perspectives at Play

• Maximalist view: Something Big is Happening
• Doubter view: Better Offline: No, Something Big Isn’t Coming, Ed Zitron of “Where’s Your Ed At”
• Doomer view: The AI Doomers Are Getting Doomier
• It’s cool to hate on AI view: Something Small is Happening

The AI Maximalists

This is going to change everything forever.

The “maxi” is all in, all of the time. Maybe they were early, maybe not. Maybe they missed out on other trends and want to be here first. They are true believers in what the technology can do and won’t shut up about it.

The AI Doomers

I hope they put us in the nice part of the human zoo.

A negative-biased version of the “maxi” that believes in the potential, but sees the outcomes in an almost exclusively negative light. It’s not clear to me if this group thinks a lack of safety is the issue, or that it’s simply inevitable.

The AI Doubters

It’s all hype just like the dot-com crash.

I will admit to finding this group the most frustrating and least informed. They seem to relish mathematical tasks like counting the letter “r” in “strawberry” and getting to a “look how stupid this thing is!” as quickly as possible.

The AI Hater (Bonus!)

I went offline before it was cool.

Basically a snotty vegan barista that judges you harshly for your life choices, but for AI. This is something that I’m trying to engage less with(and be less of a part of) as I age. I acknowledge that it exists as a perspective, but I just don’t find it interesting and won’t devote time to it.

Where I’m At(for now)

Presuming that my opinion is worth the bits to get it to your screen, here’s what I’m thinking about the personas:

• The maximalists are overselling, but not by large degree. Sure, Matt Shumer is talking exclusively in a way that benefits himself, but that doesn’t make him wrong. The failure to acknowledge the externalities is aggravating. Timelines are up for debate, but given my last several weeks, I think that I’m, reluctantly, in this camp.
• The doomers love a good spiral. See the panic over Moltbook and the failure how easy it was to manufacture a story out of the content there.
• The doubters are judging fish on their ability to climb trees. I think that they have largely formed opinions on older models, if they’ve tried them at all. Some of the most ardent critics seem to have adopted “AI Luddite” personas when models exponentially improving. I’m not sure what is to be gained by being closed minded to the fact that every lab is trying to improve all of the time.

Things I’m Keeping in Mind

• The models today are the worst they will ever be barring some training that goes wrong. Expect improvements on an ongoing basis.
• From my experience, using AI to get to discrete values is their best use.
• Replit, as a business, has been replaced by models improving, and projects like Spec Kit in a matter of months. Maybe I’m just late to this?
• SaaS is in for a world of hurt- especially offerings full of bloat and distractions. These are going to see the cost on a per seat basis drop. Buyers are going to have huge amounts of leverage over pricing when they only use 1/3 of the features offered
• Tired: putting AI in your product. Wired: enabling AI to interact with your product. The value isn’t in summaries of your data within the platform it’s housed in, but rather, combining it with disparate data easily. Trying to lock folks in to a platform will only encourage them to attempt to exit more quickly.

What I’ve Built

In the last week, the things I’ve managed to build with minimal effort:

• App to ingest backup reports that were previously emails(and mostly ignored)
• App to replace Google’s BigQuery console.
• App to map, pivot, and merge data from vastly different systems of record
• App to speed read. Based on Rapid Serial Visual Presentation(RSVP). Currently attaining 800 WPM on a consistent basis
• App for home maintenance that can leverage check-ins triggered by NFC tags

Going Forward

I mostly agree with Shumer that pure coding is solved. Knowing the application of techniques and understanding the problems that you’re trying to solve really well is the most value you can have in the day-to-day. Reducing yourself to a “spec engineer” feels like the best way to be replaced.

POSSE Note

I own my own content. I subscribe to POSSE: Publish on Own Site, Syndicate Everywhere. The original location of this content is invisiblethreat.ca.