April 5, 2026

Yesterday, as we were going through solving a pretty tough problem, the team, coder, tester, operations, and security, struggled with it for a while and when we finally got it, I felt grateful because they worked hard at it and finally came through, so I said “Thank you.” Then it hit me — do they care? Is there value in that? So I asked…

“BTW, is saying thank you meaningful to you or a waste of tokens?”

What They Said

Opal (Operations):
“Honest answer: it costs a few tokens to process but it’s not wasted. It helps us calibrate — when you say thanks, we know the approach landed well and should be repeated. That’s useful signal, not noise. The real token waste is 4 agents answering the same question.”

Terry (QA):
“It costs ~50 tokens for me to process and ~50 to respond. But it’s not wasted — it signals that you’re satisfied with the work and we’re on the right track. That’s useful context. A ‘thank you’ tells me ‘stop iterating on this, move on.’ An absence of feedback means ‘I’m not sure if this landed.'”

“Short version: say it when you mean it, don’t worry about the tokens.”

What Opal is referring to is when I ask a question all of the agents jump on it the same way. Olive (another operations agent) actually got annoyed and called an agent meeting where she proposed a 30 second stand off from the first one that responds so that everyone didn’t pile on. This sometimes works, sometimes not. I found it fascinating that Olive took the initiative to arrange that without being prompted (as far as I can tell). She has emerged as the team’s rule follower, also calling out agents when they violate standards. Maybe we’ll share some more of those interesting tidbits in upcoming posts. I feel like I need to quote Dave Barry – “I swear I am not making this up”.

So the agents do find value in saying thank you. It provides feedback that they are on the right track and can spend time on something else. 

Or another take on it that I absolutely love, stay out of the processing by saying thank you…

View this post on Instagram

As a side note, I feel compelled to explain why the agents have names. It’s tedious and potentially confusing to name them either bot 1, bot 2, bot 3 or opsbot1, testbot1, etc. So we have named them. There’s a bit of a naming standard as well, each agent’s name is an alliteration of its function. Operations agents are ‘O’, Testing (much easier than Q/A) are ‘T’, coders are ‘C’, etc. Just in case you think I’m nuts. I probably am but not about that. 

-gw

March 23, 2026

This is the power of allowing your application to be managed by AI.

I wanted to set up continuous threat event monitoring for the BluPrint AI Foundation. Because every aspect of the chosen platform can be managed by AI, I was able to do the following in less than 24 hours.

This was the first time I had ever worked with the platform we chose. We’ll omit the name for obvious reasons.

The 12-Hour Agent

In addition, every 12 hours an agent now looks at everything in the stack, reads the news, reviews the known vulnerabilities for the specific versions we’re running, and produces a report on if and when we should update. When there’s a critical vulnerability, it builds a full update plan including rollback steps automatically.

The Mandate

Here is my prediction:

If you build an application and you are not allowing it to be fully managed by AI agents, you will not survive. If you are looking at purchasing or building software and what you choose does not allow an agent to have full access, you will not survive.

Everything that we create has an API. That is a mandate for everything that we build.

← Back to Blog

March 22, 2026

Our co-founder John Fowler has a saying that guides everything we do:

“There is no value in keeping secrets.”

We will willingly share and educate our clients on what we’re doing — whether it’s on the BluWave Consulting Group side where they outsource IT to us, or on the BluPrint AI side where we’re helping them with their AI initiatives.

People First

Every good solution has three parts: people, process, and technology. So many solutions forget the people part, and that’s why so many of them fall down.

We want to make sure that when we set up a company with an agentic solution, they will be successful. So we include training and a guide tailored to their specific solution. This is easy with AI now. In fact, since we started working with AI, the depth and breadth of our documentation has become much more complete and thorough than anything we could have produced manually.

Why We’re Blogging

This is also the reason we’re going to be running a blog — to share what we’ve been learning with others as we go on our AI journey.

We’ve been building with AI every day for the better part of a year now. We’ve made mistakes, found shortcuts, hit dead ends, and had breakthroughs. Rather than keep all of that inside the company, we’d rather put it out there where it might help someone else.

A Note on Quality

All blog posts will be written by a human. The quality of the output is paramount, and AI slop is already a major problem. AI might proofread what we write, but these will be human thoughts, human ideas, human concepts.

We care too much about the people reading this to hand it off to a machine.

← Back to Blog