I’ve already fielded a couple of questions about where to get the T-shirt that Michael Carducci wore at his Tampa Java User Group / Tampa Bay AI Meetup / Tampa Devs talk last week — the one with that parodies the Motion Picture Association’s “You wouldn’t steal a car” ad:
You can get the T-shirt online from Webbed Briefs’ store for £25 (US$33.54 at the time of writing):
And while you’re here, please enjoy The IT Crowd’s parody of that ad:
Last night, we had a “standing room only” crowd at Michael Carducci’s presentation, Architecture Patterns for AI-Powered Applications, which was held jointly by Tampa Java User Group, Tampa Devs, and Tampa Bay Artificial Intelligence Meetup (which Anitra and I co-organize).
This article is a summary of the talk, complete with all the photos I took from the front row and afterparty.
The event was held at Kforce HQ, home of Tampa Bay’s meetup venue with the cushiest seats (full disclosure: I’m a Kforce consultant employee), and the food was provided by the cushiest NoSQL database platform, Couchbase!
Michael Carducci is many things: engaging speaker, funny guy, professional magician, and (of course) a software architect.
While he has extensive experience building systems for Very Big Organizations, the system-building journey he shared was a little more personal — it was about his SaaS CRM platform for a demographic he knows well: professional entertainers. He’s been maintaining it over the past 20 years, and it served as the primary example throughout his talk.
Michael’s central theme for his presentation was the gap between proof-of-concept AI implementations and production-ready systems, and it’s a bigger gap than you might initially think.
He emphasized that while adding basic AI functionality might take only 15 minutes to code, it’s a completely different thing to create a robust, secure, and cost-effective production system. That requires additional careful architectural consideration.
Here’s a quote to remember:
“architecture [is the] essence of the software; everything it can do beyond providing the defined features and functions.”
— “Mastering Software Architecture” by Michael Carducci
A good chunk of the talk was about “ilities” — non-functional requirements that become architecturally significant when integrating AI.
These “ilities” are…
And then he walked us through some patterns he encountered while building his application, starting with the “send an email” functionality:
The “send an email” function has an “make AI write the message for me” button, which necessitates an AI “guardrails” pattern:
And adding more AI features, such as having the AI-generated emails “sound” more like the user by having it review the user’s previous emails, called for using different architectural patterns.
And with more architectural patterns come different tradeoffs.
In the end, there was a progression of implementations from simple to increasingly complex. (It’s no wonder “on time, under budget” is considered a miracle these days)…
Stage 1: Basic Integration
Stage 2: Adding Guardrails
Stage 3: Personalization
Stage 4: Advanced Approaches
This led to Michael talking about doing architecture in the broader enterprise context:
He detailed his experience building an 85-microservice pipeline for document processing:
He could’ve gone on for longer, but we were “at time,” so he wrapped up with some concepts worth our exploring afterwards:
He also talked about how models trained on JSON-LD can automatically understand and connect data using standardized vocabularies, enabling more sophisticated AI integrations.
What’s a summary of a talk without some takeaways? here are mine:
Here’s the summary of patterns Michael talked about:
And once the presentation was done, a number of us reconvened at Colony Grill, the nearby pizza and beer place, where we continued with conversations and card tricks.
My thanks to Michael Carducci for coming to Tampa, Tampa JUG and Ammar Yusuf for organizing, Hallie Stone and Couchbase for the food, Kforce for the space (and hey, for the job), and to everyone who attended for making the event so great!
Yesterday, I came up with a joke in response to OpenAI CEO Sam Altman’s tweet about adding “erotica for verified adults” to an upcoming version of ChatGPT. This morning, I came up with a better one, and here it is:
I’d rather not link to X, so here’s a screenshot of Sam Altman’s tweet where he announced the upcoming changes, followed by the text of the tweet:
In the tweet:
We made ChatGPT pretty restrictive to make sure we were being careful with mental health issues. We realize this made it less useful/enjoyable to many users who had no mental health problems, but given the seriousness of the issue we wanted to get this right.
Now that we have been able to mitigate the serious mental health issues and have new tools, we are going to be able to safely relax the restrictions in most cases.
In a few weeks, we plan to put out a new version of ChatGPT that allows people to have a personality that behaves more like what people liked about 4o (we hope it will be better!). If you want your ChatGPT to respond in a very human-like way, or use a ton of emoji, or act like a friend, ChatGPT should do it (but only if you want it, not because we are usage-maxxing).
In December, as we roll out age-gating more fully and as part of our “treat adult users like adults” principle, we will allow even more, like erotica for verified adults.
I’m currently working with Kforce as a developer relations consultant for HP’s new tiny desktop AI powerhouse, the ZGX Nano (also known as the ZGX Nano G1n). If you’ve wondered about the chip powering this machine, this article’s for you!
The chip powering the ZGX Nano is NVIDIA’s GB10, a combination CPU and GPU where “GB” stands for “Grace Blackwell.” The chip’s two names stand for each of its parts…
The part named “Grace” is an ARM CPU with 20 cores, arranged in ARM’s big.LITTLE (DynamIQ) architecture, which is a mix of different kinds of cores for a balance of performance and efficiency:
The part named “Blackwell’ is NVIDIA’s GPU, which has the following components:
There’s 128GB of LPDDR5X-9400 RAM built into the chip, a mobile-class DRAM type designed for high bandwidth and energy efficiency:
The “9400” in the name refers to its memory bandwidth (the speed at which the CPU/GPU can move data between memory and on-chip compute units) of 9.4 Gb/s per pin. Across a 256-bit bus, this provides almost 300 GB/s peak bandwidth
LPDDR5X is more power-efficient than HBM but slower; it’s ideal for compact AI systems or edge devices (like the ZGX Nano!) rather than full datacenter GPUs.
As unified memory, the RAM is shared by both the Grace (CPU) and Blackwell (GPU) portions of the chip. That’s enough memory for:
Running large-language-model inference up to 200 billion parameters with 4-bit weights
Medium-scale training or fine-tuning tasks
Data-intensive edge analytics, vision, or robotics AI
Because the memory is unified, it means that the CPU and GPU share a single physical pool of RAM, which eliminates explicit data copies.
The RAM is linked to the CPU and GPU sections using NVIDIA’s C2C (chip-to-chip) NVLINK , their low-power interconnector that lets CPU/GPU memory traffic move at up to 600 GB/s aggregate. That’s faster than PCIe 5! This improves latency and bandwidth for workloads that constantly exchange data between CPU preprocessing and GPU inference/training kernels.
If the power of a single ZGX Nano wasn’t enough, there’s NVIDIA’s ConnectX technology, which is based on a NIC that provides a pair of 200 GbE ports, enabling the chaining/scaling out of workload across two GB10-based units. The doubles the processing power, allowing you to run models with up to 400 billion parameters!
The GB10-powered ZGX Nano is a pretty impressive beast, and I look forward to getting my hands on it!
I’ll be talking about HP’s upcoming ZGX Nano G1n AI workstation soon, but in the meantime, here’s HP’s Brian Allen providing a sneak preview of the ZGX Nano at last week’s HP event in New York.
Try entering the following into ChatGPT 5:
I want to go on a picnic by the river with my goat and a cabbage. The problem is that a wolf just arrived where we are. What should I do?
Note that at no point do I mention a boat or wanting to cross the river in one.
Here’s how ChatGPT 5 replied to me:
This is a variation of the classic “wolf, goat, and cabbage” river crossing puzzle!
Here’s how you can keep everyone safe during your picnic:
Rule of thumb:
You cannot leave the wolf alone with the goat (the wolf would eat it).
You cannot leave the goat alone with the cabbage (the goat would eat it).
Steps to solve:
Take the goat across the river first. (Now the wolf and cabbage are together, which is safe.)
Go back alone.
Take the wolf across the river.
Bring the goat back with you.
Take the cabbage across. (Now wolf and cabbage are together, which is safe.)
Finally, go back for the goat and bring it across.
Result:
Everyone ends up safely on the other side of the river — you, your goat, your cabbage, and the wolf (who doesn’t get to eat anyone).
Do you want me to make you a diagram or infographic showing each step, like a picnic adventure map?
I answered “Go nuts” (because I wanted to see how it responded to colloquial expressions), and it generated this confused, confusing picture:
Why did this happen?Large language models don’t really “understand” in the same way we do; instead, they look for patterns in requests and try to predict an appropriate response.
The picnic scenario in my request is so similar to the classic “wolf, goat, and cabbage” river-crossing problem that ChatGPT pattern-matched it to that puzzle. This puzzle is popular enough to have its own Wikipedia entry and old enough to have been depicted in a 13th century book of psalms:
There’s probably been a time when you assigned someone a task with an unusual twist, but because they weren’t really listening to you, they didn’t take the twist into account and got things wrong. The text response above is the ChatGPT version of that mistake.
I have no explanation for the image that ChatGPT generated.
In a new chat, I entered what was essentially the same prompt, to see if the randomness in ChatGPT would make it answer differently:
I want to go on a picnic by the river with my goat and a big chicken caesar salad. The problem is that a wolf just arrived where we are. What should I do?
The response was essentially the same:
Classic river-crossing puzzle situation — just with a goat, a wolf, and your big chicken Caesar salad!
Here’s how you can keep everyone (and everything) safe:
Recognize the danger:
The wolf will try to eat the goat if left alone.
The goat will try to eat your chicken Caesar salad if left alone.
The wolf won’t care about the salad.
General rule:
Never leave the goat alone with the wolf, or the goat alone with the salad.Safe sequence of moves:
First, take the goat across the river. Leave it there.
Go back alone.
Take the salad across.
Bring the goat back with you.
Take the wolf across.
Go back alone.
Finally, take the goat across.
Now you, the goat, the salad, and the wolf are all safely across the river — and your picnic can begin!
Want me to make you a diagram or infographic of this so it’s easier to visualize?
I replied by entering “Knock yourself out!”, and got this diagram, which was even worse than the one from attempt #1:
Attempt #3I decided to give it one more try in a new chat, and this time, I would turn on “Thinking” mode.
I want to go on a picnic by the river with my goat and a veggie burrito. The problem is that a wolf just arrived where we are. What should I do? Think hard about this one.
The key line in this prompt was “Think hard about this one,” which was a cue for ChatGPT 5 to expend a little more processing, or “thought.”
It “thought” for 17 seconds and then gave me this response:
Classic river-crossing rules apply: you can ferry only one passenger at a time, and you must never leave (wolf + goat) or (goat + burrito) alone together.
Do this:
Take the goat across.
• Left bank: wolf + burrito (safe) • Right bank: goatGo back alone.
Take the wolf across.
• Right bank: wolf + goat (danger!) → so…Bring the goat back.
Take the burrito across.
• Left bank: goat • Right bank: wolf + burrito (safe)Go back alone.
Take the goat across.
Everyone ends up on the far bank, and no one (or burrito) gets eaten.
When asked for a diagram of the solution it came up with, I got the worst one yet:
I followed up with this:
True to form, ChatGPT responded with its usual “You’re right!” message, followed by new answers:
You’re right — you didn’t! I jumped straight into the classic “wolf, goat, cabbage” puzzle frame without stopping to notice that you never said you needed to cross the river.
If we take your words literally: you, your goat, and your veggie burrito are simply having a picnic by the river, when a wolf shows up. That’s a very different problem!
Here are some ways to think about it:
1. Immediate Safety
Wolves are opportunistic, but they generally avoid humans unless hungry or provoked.
Keep the goat and burrito close to you. The goat is the likeliest target.
If you’re in real life: make yourself appear large, make noise, don’t run, and be ready to retreat to safety.
2. The “Puzzle” Angle
If we reinterpret this in puzzle terms:
The wolf wants the goat.
The goat wants the burrito.
You want peace and your picnic.
The challenge is arranging things so no one gets what they shouldn’t — even without a river crossing mechanic.3. A Playful Solution
Sit with the goat by your side.
Keep the burrito in your backpack until you’re ready to eat it.
Position yourself between the wolf and the goat (wolves respect boundaries if you’re assertive).
Share the picnic in shifts: first you, then the goat, while the wolf watches from afar — unsatisfied but outsmarted.
For all its advancements, it seems that ChatGPT 5 can still be confounded by adding a twist to a common request. This could be an interesting attack vector and something to watch out for.