AI builder · Educator · Former Meta

I was evaluating AI at Meta before ChatGPT made it headline news.

I'm Eli Allen. For five years at Meta I built software that improved and evaluated the AI they were working on, before and after the large language model hype. Since then I've used AI every day to write software, automate the boring stuff, and build learning platforms that give students real practice for real jobs.

That's why people listen when I talk about working with AI: I didn't arrive with the hype. I was already there, testing what it could and couldn't do.

What I'm working on now My story in three eras ↓ AvailaBooks.com ↗
5 yrs
At Meta, building AI improvement & evaluation software
Pre + Post
ChatGPT: worked on AI on both sides of the LLM moment
105
Public repositories on GitHub, active since 2019
2023
Master's in Information Systems Management, BYU

The story

Three eras, one through-line: does it actually work?

Tap an era. Each one shaped how I judge AI today, not by demos, but by whether it holds up when real people depend on it.

Why listen to me on AI?

Three kinds of credibility, not just opinions

Most AI advice comes from one angle. Mine comes from evaluating it, shipping with it, and teaching people to use it.

What I'm working on now

Simulations, grading, and teaching people to actually use agents

Filter by what you care about. Everything here is current, real work, not a roadmap.

Muse Simulator ↗

Simulation

An interactive simulator of the Muse AI app, built with AI assistance for my course. Students tap through realistic setups, including a “Your First Run” onboarding simulation added Nov 3, 2026, and practice working with an agent before they have to do it for real.

github.com/eaallen/muse-simulator · Most recent project

AvailaBooks ↗

Platform

Home of my free online textbook Personal Productivity with JavaScript, and the platform I'm building beyond it: interactive simulations and knowledge checks instead of passive videos, so students learn by doing.

availabooks.com · Free textbook + learning platform

Automated grading for educators

Grading

The bulk of my recent work: helping instructors auto-grade Word and Excel student work, including transcript grading and Excel simulators. I maintain @availabooks/simmy on npm, an embeddable Excel simulator with a graded task pane.

npm: @availabooks/simmy · Word & Excel auto-grading

Prometheus Excel / SQL Educator

Grading

I promote the Prometheus Excel/SQL Educator, a MyEducator resource co-authored by Gove Allen and Nick Ball, to college instructors who need students to get genuine spreadsheet and SQL practice, with grading that scales.

MyEducator resource · For college instructors

“Get the Help You've Always Wanted”

Live course · Dec 2026

My live course on working with AI agents, hands-on with Muse, taught through AvailaBooks in December 2026. Three live classes, office hours, and a simple promise: you leave with an agent set up and doing real work for you, with you approving what matters.

Through AvailaBooks · December 2026

Speaking & USU Eastern proposal

Teaching

I spoke at the virtual “AI for Educators” conference run by the Close the Gap Foundation on September 26, 2026. Separately, I'm proposing a one-semester course at USU Eastern, Personal Productivity with JavaScript, built around using AI to build tools for technical work.

Virtual conference · Close the Gap Foundation + proposed course at USU Eastern

What I believe

Why I'm excited, and what I'm building toward

Open each one. This is the point of view behind the work above.

Reading and watching only go so far. I'm most excited about AI giving students simulated experiences (a spreadsheet to fix, an app to navigate, an agent to direct) that prepare them for real jobs, safely and repeatedly, before the stakes are real. The Muse Simulator and the Excel simulator are both bets on exactly that.
Since Meta, I've used AI extensively to write software and automate repetitive, boring work. The gain isn't just speed. It's attention. When grading, drafting, watching an inbox, and routine follow-up are handled, teachers and small operators get their time back for judgment, relationships, and craft.
Building evaluation software at Meta taught me to distrust a good demo. Good AI work means checking outputs, knowing where a system fails, and keeping a human in charge of what sends, spends, or counts. That's the habit I teach: draft, review, approve, then automate with confidence.
I'm really trying to make the world a better place with this. That's why the textbook is free, why the course is hands-on instead of theoretical, and why I keep building tools for educators. If AI is going to change work, the people who learn it by doing, not just hearing about it, will be ready.
“I didn't get into AI when it got popular. I was building the tools that measured it, and now I'm building the tools that teach it.”

If you're an educator, a small operator, or just someone who wants an agent doing real work for you, start with the free textbook at AvailaBooks.com, try the Muse Simulator, or follow the work on GitHub and LinkedIn.