Happy Friday & Happy New Year!

I don't often write predictions but I think a new year is as good a time to do that as many folks wonder what the future holds.

I also don't really believe in new year's resolutions. They're a great way for folks to create ambitious goals that they feel terrible about in about three months. Steady, small, change is boring but more effective.

The main topic I want to wager a prediction on is on AI and agentic coding. I think my last email was about this as well, though more of a how-to than an email about what I think it means and how to differentiate yourself.

Here's the bottom line: AI is fast, not good.

This has held true for several years and while there is an argument to be made that is getting better, I'll retort that it is getting better slower and slower.

We are still in the position of having to double-check AI's output.

Now, speed is always something organizations and software groups obsess over, but it often doesn't work out very well. So here's the question that every company will begin to grapple with.

With AI being as fast as it is and as available as it is, are you seeing the results you expected?

My prediction: No, you aren't.

The funny thing about having AI as a tool that can work many multiples faster than a human is that it removes the conceptual need to ship things faster. The only questions left are the ones that always mattered but we always shy away from. Those questions are:

• Are we shipping the right thing?
• Are we shipping the thing right?

These are to questions that feel big enough to paralyze a lot of folks, and many organizations debate and struggle with them. One main reason for the struggle is that it seems to require sacrificing speed.

I want to lay out a few things that I'm not going to defend in this email. If you want to have a conversation about them, I'd love to.

With that, here are things that I believe that clever organizations will realize and the rest will use AI to accelerate into a metaphorical brick wall.

• Quality as something you address after-the-fact, will always cost more
• "Better requirements" has never worked and won't now
• Shorter feedback loops on quality, product, etc are key to avoiding costly mistakes
• You are already doing "What makes sense," and if that isn't going well, you'll have do something that doesn't make sense

These statements are statements that I believe will become profound truths as the speed of AI becomes normalized and the problems that come with pursuing speed become more apparent.

Sincerely,

Ryan