A HUMAN, NOT AN ENDPOINT.

Hello, I'm
LouMM.

Engineer by training.
Curious by default.

Exploring the space between people, technology,
and “what if?”

Enter my world Say hello
Photograph of LouMM, smiling in a blue suit

Louis Maresca

Kid-hearted curiosity. Engineering degrees as enablers.

About the human
Photograph of LouMM, smiling in a blue suit

Louis Maresca

I'm LouMM. Engineering is how I turn curiosity into something useful. This is my space to explore AI, share things I'm learning, and find people who like building together. More personal details, projects, and speaking dates will be added here as this little corner of the internet grows.
MORE BELOW THE FOLD

01 / OUT IN THE WORLD

Ideas are better out loud.

Talks, meetups, and conversations that keep the curiosity going.

Start a conversation
OCTOBER212026 / UP NEXT
ON THE CALENDAR / PODCAST APPEARANCE

Intelligent Machines
Episode 893.

Joining the conversation on . The recording will be linked after publication. Until then, explore the latest episode in the podcast section.

AI & agentsIn conversationNew perspectives
See show

02 / FROM THE NOTEBOOK

Thinking, in the open.

Notes on making things, figuring things out, and staying a beginner.

A FIRST EDITION / STARTER ESSAYS
FROM THE NOTEBOOK

Building things 2 min read Starter essay

Leave room for curiosity.

The most useful question in the room might be the one nobody planned to ask.

A preview of this site's reading experience. Editorial content to make your own.

A little space for the unexpected

A project usually begins with a thing to make. A page, a tool, a conversation. The interesting part is what happens between the first sketch and the finished thing: all the questions that were not on the list.

What if this were easier? What would someone try first? Is the complicated part necessary, or have we simply stopped noticing it? Curiosity is not the opposite of discipline. It is one of the ways we find out where discipline is worth applying.

Make the question small enough to try

A good experiment does not need a grand launch. Make the smallest version that lets someone use it. Watch what happens. Keep the surprising parts and let the rest go.

Engineering gives that curiosity a way to become real. A constraint is not just a wall; sometimes it is the shape that makes the next idea possible.

Keep a little of the beginner

The more we know, the easier it becomes to reach for an answer before we have really heard the question. Leave some room for not knowing. For trying the odd idea. For asking why a familiar thing has to work that way.

This is a starter note for a growing personal site. The working principle is simple: stay curious, make something useful, and share what you learn.

✳
FROM THE NOTEBOOK

AI & agents 2 min read Starter essay

Good agents need good maps.

Before adding another tool, give your agent a clearer picture of the place it is working.

A preview of this site's reading experience. Editorial content to make your own.

Start with the room, not the robot

Imagine asking someone to help in a workshop without showing them where anything lives. Even a talented person will spend time opening the wrong drawers. An agent has the same problem when a project is full of hidden assumptions.

Clear folders, readable names, and a short description of the goal are surprisingly powerful tools. They help people and software arrive at the same understanding.

A small job has a visible finish line

Give each task a boundary. Name the files it may change, describe the behavior that should remain, and say what would count as evidence that the work is complete.

Small tasks make it easier to see a wrong turn early. They also make it easier for another person to pick up the work without reconstructing the whole conversation.

Keep the map with the work

Documentation belongs next to the thing it describes. A component with a clear purpose, a short example, and a useful test is easier to move, understand, and improve.

The goal is not a mountain of instructions. It is just enough shared context to make the next good decision obvious.

✳
FROM THE NOTEBOOK

Community 2 min read Starter essay

Build with other humans.

Better tools are wonderful. A room full of generous people is still hard to beat.

A preview of this site's reading experience. Editorial content to make your own.

Show the unfinished thing

A finished demo has a story. An unfinished demo has a conversation. Showing the part that does not quite work yet gives other people something real to contribute.

A community is useful when it makes that kind of sharing feel ordinary. Not a performance of knowing everything, but a place to compare notes.

Make the invitation easy

Explain what you are trying, what you have learned, and the question you are still carrying. Someone else may have met the same problem from a completely different direction.

Good collaboration does not always produce more code. Sometimes it produces a simpler plan, a better question, or permission to stop building the wrong thing.

✳

03 / ON THE OTHER SIDE OF THE MIC

Conversations worth keeping.

From the enterprise trenches to the next chapter of intelligent machines.

My next appearance: Oct 21 ↗
Intelligent Machines 892: Pants-based Benchmarks

INTELLIGENT MACHINES / LATEST PUBLISHED

Pants-based Benchmarks

· Episode 892

GitHub's agent revolution, with guest Martin Woodward. The latest published episode while my appearance is still ahead.

Watch or listen on TWiT ↗

NEXT UP → I'll join episode 893 on . The recording link will follow publication.

From the host's chair.

TWiET / THE 2023 ARCHIVE

These are my four most recent TWiET host credits, verified against TWiT's archive. Original publication dates are preserved.

04 / THE INTERNET IS STILL INTERESTING

Worth opening another tab.

A small collection of rabbit holes. Bring your curiosity.

COLLECTED, NOT ALGORITHMICALLY SERVED.A CONTINUOUS THREAD. DRAG TO EXPLORE →

05 / A DIFFERENT KIND OF HELLO

A little less form. A little more conversation.

Meet LouBot. A small, browser-local experiment in making the internet more personal.

Ask-LouMMEXPERIMENT / 001

Your browser.
Your conversation.
No server in the middle.

AI is optional. Explore the site without downloading a thing.

Local console controls

LouBot

LOCAL AI / PRIVATE INBOX
Your browser. Your conversation.
Explore before you enable AI

Start with a question.

These are curated prompts, not AI replies. Choose one to draft a question; no model runs or downloads until you enable it.

Enable local AI to send. Drafting a question downloads nothing.

Local AI is off. No model has been downloaded by this console.

Small model. Public context.

OR FIND ME IN THE CONVERSATION

06 / SHOW YOUR WORK

A window into the work.

Not magic. Not a black box. Just a few small steps, made visible.

SCRIPTED, NOT LIVE

How a small agent-assisted task might unfold. Prerecorded-style examples, not real command execution. No shell, files, or processes are accessed.

The agent workbench

A small request. A bounded plan. A change you can inspect.

Waiting for the visual terminal.
1 / 3

Make a small feature. Step 1: Understand the request. Paused.

Choose a script: demo feature, demo review, or demo ship.

Read the complete walkthrough transcript
Understand the request
$ agent "add a reading-time label"
Read the article template and content schema.
Keep the change local. No new dependencies.
Plan: count words, round up, show the label.

Make one focused change
$ agent apply --scope reading-time
+ readingMinutes = max(1, ceil(words / 220))
+ Add a quiet label below the article title.
Leave the article content and layout alone.

Verify the edges
$ npm test -- reading-time
Example result: empty article -> 1 min
Example result: 221 words -> 2 min
Review the diff. A human decides what ships.

Tools do the repetition. People make the calls.

Signal garden

INK / 002

A little order. A little noise.
An original, living character field.

Waiting for the visual terminal.
A quiet moment.

Original browser artwork · Ink-rendered · 8 fps maximum

The agent workbench · scripted demo
All three scripted walkthroughs, in plain text

These are illustrative, prerecorded-style sequences. No commands execute, no files change, and nothing deploys. Read every step here—with or without JavaScript.

Make a small feature

A small request. A bounded plan. A change you can inspect.

Understand the request
$ agent "add a reading-time label"
Read the article template and content schema.
Keep the change local. No new dependencies.
Plan: count words, round up, show the label.

Make one focused change
$ agent apply --scope reading-time
+ readingMinutes = max(1, ceil(words / 220))
+ Add a quiet label below the article title.
Leave the article content and layout alone.

Verify the edges
$ npm test -- reading-time
Example result: empty article -> 1 min
Example result: 221 words -> 2 min
Review the diff. A human decides what ships.

Review an agent change

Read the intent, challenge the edges, keep the human in the loop.

Read the scope
$ git diff --stat
Example diff: 2 files, one form validation.
Question: does the fix preserve valid input?
Read the patch before trusting the summary.

Challenge the happy path
$ agent review --focus edge-cases
Check empty input and whitespace-only input.
Finding: trim before checking the length.
Request a focused regression test.

Make the decision
$ npm test -- form-validation
Example result: valid input still accepted.
Example result: blank input gets a hint.
Suggested outcome: ready for human review.

Ship a static site

Check the build, inspect the output, publish with intention.

Check before building
$ npm run check && npm test
Example checks: types, links, focused tests.
No server or secret belongs in the browser.
Confirm the page still reads without JS.

Inspect the static output
$ npm run build
Example output: HTML, CSS, lazy JS assets.
Preview small screens and keyboard paths.
Check reduced motion and loading states.

Publish deliberately
$ agent summarize --release
Write a short summary and rollback note.
A person approves the hosting deployment.
This demo publishes nothing. Your move.

07 / FIND YOUR PEOPLE

Better, together.

The best part of building?
The people you meet along the way.

For the makers, the question-askers, and the “let's try it” people. These are the spaces this corner of the internet is growing toward.

AI builders

Experiment. Share. Build something useful.

TAKING SHAPE

Agents Anonymous

Agent-curious? You're in good company.

TAKING SHAPE
In the meantime, find me on GitHub
STAY CURIOUS

  1. MAKE THINGS.

  2. hello, world.

  3. LOOK AGAIN

  4. WHAT IF?

  5. HUMANS FIRST.

  6. “The good part
    is figuring
    it out.”

A still orbit. Use the arrows to explore.