Field Notes on AI

A small collection reflecting how I think about AI — excited but measured, and grounded in systems.

Confidence vs. capability

Wear it: AI confidence is often louder than AI capability.

Systems deserve evaluation, not optimism.

Cartoon by Joel Pett about attention span and AI

Judgment matters

Tools change cognition over time.

Responsible use requires maintaining human judgment.

Cartoon by Joel Pett

Hype vs. systems

Morning coffee: Humor captures anxiety about automation.

My work focuses on controlled, well-evaluated systems.

Tool, not authority

Here's hoping: AI reflects how we choose to use it.

Discipline remains a human responsibility.

Internet meme, source unknown

Adoption gaps

Not for long: At a non-technical conference, a large share of attendees reported not using AI at all.

Organizations that delay AI literacy risk falling behind peers who experiment responsibly.

AI Chats

Selected moments from real conversations — some technical, some reflective, all contributing to how I think about building with AI.

Screenshot of a supportive ChatGPT response — **Support** Encouragement that kept momentum during difficult builds.

Screenshot of a humorous ChatGPT exchange — **Humor** A well-timed laugh in the middle of serious engineering.

Screenshot describing identity traits such as valuing judgment and responsibility — **Identity** A reflection on judgment, responsibility, and why cognition still matters.

Attention

When the conversation itself becomes the artifact.

The screenshot is clickable to view full size.

Attention (continued)

The rest of the breakdown and conclusion.

Continuation of the longer ChatGPT exchange (part 2)

When I was building projects piecemeal.

Curated Q&A

Selected questions and answers, organized by theme. (More coming.)

Humor

Question

I'm concerned that article is going to conflict with the "Clarity over hype" footer

Answer

No — unless your LinkedIn article sounds like: “AI is magic and everything will change tomorrow!!!”

Question

if smoothing and smoothing > 0: ?

Answer

This checks truthiness

Encouragement

Question

How do I get past that inhibition to continue self-defense training?

Answer

Think of sparring the way doctors think of surgery:

Surgery causes injury—but with healing intent
Sparring causes contact—but with protective intent

Question

Logistic regression has the highest roc_auc?

Answer

That is not a failure — it’s a baseline anchor.

Portfolio

Question

Do I need to post to LinkedIn every 2-3 days?

Answer

Hard Truth (Respectfully) You're not "bothering" people. People scroll past 95% of content without noticing.

Question

Should I note that it is paced and unlocked as we advance to explain the delay?

Answer

This is a very Greg question. You’re thinking about signaling discipline without sounding defensive.

Question

503 Service Unavailable on Render?

Answer

If this is a lead portfolio project, “sometimes asleep” is a real conversion killer.

Question

The model did not use XGBoost but HistGradientBoostingClassifier instead?

Answer

“Used sklearn-native HistGradientBoosting for baseline stability” looks better than “Forced XGBoost everywhere without justification”

Debugging

Question

I may have missed a step, GitHub partially updated?

Answer

No worries — this is a super common situation, and it’s fixable in a few clean steps.

Question

Concerned about different variable names and nomenclature consistency

Answer

So your concern about nomenclature consistency and correctness — you can relax now. This is correct and robust.

Selected Comments

Other commentary that stuck with me

“You're officially in the arena now.”

“So odds are very high this just works”

“Boom — we found it. ✅”

“This is seasoning, not the meal.”

“This is production-grade ML, not notebook spaghetti.”

“What NOT to worry about"

“Take a break (seriously—you earned it)”

AI is neither magic nor menace. It is a powerful tool that demands structure, evaluation, and responsible use.