TetraMesa

Archives for April 2025

MVPs + Vibe + Moral Hazard = VibeWreck?

By

“There’s two classes of failure: those who thought and never did, and those who did and never thought.” – Laurence J. Peter

As product people, we may have some dev experience. Or not. Either way, we should know about risk. We know when deploying products the bottom line is what value we’re putting out there and is it worth more than the costs. The risk comes into play depending on how far out on the edge we’re going. Upon failures, regardless of cause, who takes the first hit and is invited to seek out other career options? Typically product. Perhaps a whole division. And yes, perhaps a whole company. I’ve seen this up close and been a part of it a couple of times. Still, when building the new, at some point we have to do as Richard Branson says, “Screw it, let’s do it.”

I understand I might be hopping on this train bit late. While I thought my rant was semi-on-time, while typically sitting on my draft over weeks, others have piped up and said things well or better. Such as this thread from Simon Wardley about Vibe Coding in general.

Fine. OK. So I’m not first. That’s ok. Here’s my perspective from the product side anyway…

Here’s the Too Long, Didn’t Read (tl;dr): I believe the whole Vibe Programming thing is going to result in some tragically bad outcomes. Yes, perhaps also some exceedingly rare big wins. But mostly not. And yes, many of us now have spectacular tools for faster prototyping and testing. That’s great. Will this collective benefit be worth the costs of what I think might be some stunning failures coming soon to a web thing near you? Maybe. Tough call. Stay tuned. Now I have to mention my favorite quote from that Wardley comment thread I came across: “Just make sure you hire lots of really good lawyers and fire extinguishers then. Or, hire some software engineers … you’ll need them.”

Here’s my bottom line and if you buy off on it, you can just skip the whole rest of the article:

The latest Gen AI tools for coding and product design and production are amazing. I’ve quickly adopted them myself and find them super useful. And yet, they’re still super scary. Anyone actually delivering production product with these things is sending a lot of risk and waste out into the world. And I think we’re going to sense the fallout from that soon enough.

[Read more…]

LLM / Text Vectors for Product Managers

By

Intro

Understanding how these things work matters.

Not because you’re going to build the next GPT yourself, but because understanding just enough of how LLMs and vector math work can change how you think about products, teams, and strategy. It can help inspire better solutions, make smarter tradeoffs when AI promises start sounding magical, and maybe even help you call BS when needed. Whatever strategic product decisions you may be making, your implementation team could be internal or perhaps a contract shop. In either case, there’s operations impact and costs that will likely impact your roadmap. If you also have P&L responsibility, you’re going to need to look at the costs here with regard to your business case. And if you don’t, chances are you may be the one who still has to justify the spend to others. As usual in product manager land, even if you’re not the one executing the actual work, you likely need to understand enough about the pieces to know what they can do and what this might cost.

tl;dr

  • LLMs turn text into numbers using math called vector embeddings. We’re going to look at this below.
  • These vectors live in a high-dimensional space, where “distance” equals “semantic similarity.” Again, we’ll look at an example below.
  • Transformers (not the ones from the comics/movies) are the model architectures that makes GPT-style LLMs so powerful.
  • All this lets us build apps that “understand” language enough to generate answers, categorize, summarize, translate, and more.
  • But it’s still math, not magic. And it’s expensive from a lot of perspectives. The question is where do we want to take the expense hit(s) and for what level of benefit.
[Read more…]

De-Dollarization Risks from Crypto and AI

By

Edit: Here’s a July, 2025 update. Just a few months after this post, it’s becoming clearer that Stablecoins are on an amazing tear. And most are pegged to the dollar. The U.S. dollar. If this continues to spread worldwide, then that trend is potentially directly in conflict with the thoughts I’d had about the potentially of dollarization via crypto. This may be especially true if official U.S. policy explicitly promotes Cryptomercantilism. Though there will also then be explicit attempts by others to regulate the risk that comes with this. (See the Study: Cryptomercantilism vs. Monetary Sovereignty-Dealing with the Challenge of US Stablecoins for the EU.)

Anyway, to the original article…

[Read more…]

Upgrading an AI with a RAG Vector DB

By

In a previous post about Building a PM Helper with AI, I showed how a fun personal AI project I’d built to be my personal Product Management tool searched across multiple sources before synthesizing answers. Unfortunately, I made both a strategic and a tactical error in that 1.0 version. The solution? Using Retrieval Augmented Generation with a Vector Database. What I’m going to do here is offer some super fast high level definitions as I go through the problem space, and maybe in future posts, go more deeply into RAG and Vector databases in terms of value.

tl;dr:

  • If your a product manager working with AI at any level, you will likely need to understand Retrieval Augmented Generation (RAG) to some degree. The following is a small, practical use case to help see the value in action.
  • For the most part, when using LLMs for your own custom work, you’re stuck with the foundation model.
  • Fine tuning can change the weights of the model to various levels, depending on how deep you want to go. These weights basically control how models transform input and can be in the billions. The deeper you want to have impact, the higher the cost. (You’re not likely fine-tuning for personal projects though. And if you are, it will all but certainly be with open source foundational models.)
  • Retrieval Augmented Generation (RAG) doesn’t change weights at all. RAG just passes more information into a Prompt, (which is a fancy name for an information query, unless you really add fuller instructions ), but is limited to something called a context window. Basically, how much info you can pass in. It’s like saying, “Here, read this before answering.” So theoretically RAG reduces the chances of hallucinations and offers more “truthy” answers. (Assuming good data in what you feed it.)
[Read more…]

Location

We're located in Stamford, CT, "The City that Works." Most of our in person engagement Clients are located in the metro NYC area in either New York City, Westchester or Fairfield Counties, as well as Los Angeles and San Francisco. We do off site work for a variety of Clients as well.

Have a Project?

If you have a project you would like to discuss, just get in touch via our Contact Form.

Connect

As a small consultancy, we spend more time with our Clients' social media than our own. If you would like to keep up with us the rare times we have something important enough to say via social media, feel free to follow our accounts.