It Started With a Simple Question

As a New York City teacher, I kept hearing the same thing about students:

They’re low-skilled.

Low-skilled readers. Low-skilled writers. Missing foundational skills. Missing prerequisite skills.

After hearing the word skills enough times, I started asking what seemed like an obvious question:

What skills, exactly?

If a student cannot understand a difficult science text, is the problem science knowledge? Vocabulary? Reading comprehension? Background knowledge? Reasoning? Or a prerequisite skill that was never fully learned?

“Low-skilled” describes a problem.

It does not tell a teacher what to teach next.

I wanted to get more precise.

Not simply: This student struggles with writing.

But: Which part of writing? Which skill? Which prerequisite? What can we actually teach?

That question became the beginning of SkillsTeacher.AI.

First, We Had to Find the Skills

The project began by researching and organizing the individual skills underneath literacy, writing, mathematics, science, social studies, and other academic subjects.

Generative AI made it possible for one teacher to investigate the problem at a scale that would have been extraordinarily difficult before. Large bodies of standards, instructional concepts, academic tasks, and subject knowledge could be examined, compared, broken apart, and organized.

The lists grew.

Hundreds of skills became thousands.

And the more specific the skills became, the more useful the idea became.

Writing, for example, is not one skill. A learner may need to gather relevant information, integrate quotations smoothly, organize an argument, connect evidence to a claim, revise clearly, or choose more precise language.

Those are different abilities.

A student can demonstrate some and still need instruction in others.

That level of precision matters because the more precisely we can identify the skill, the more precisely we can think about what to teach next.

But another problem quickly became obvious:

A list of skills isn't instruction.

A teacher can know the name of a skill and still need to know how to teach it. A student can be told to “use stronger evidence,” “improve organization,” or “make better inferences” without anyone ever explicitly teaching what those things mean or how to do them.

So the next question became:

How do we make every skill teachable?

Give Every Skill a Place of Its Own

Each skill began receiving its own instructional material.

Teacher-facing guides could explain the skill, why it matters, how it works, and how it can be taught. Tutoring scripts could explain the same idea in straightforward language that a teacher, tutor, parent, or learner could follow.

The principle was simple:

If a skill matters enough to say someone needs it, it matters enough to explain clearly.

As new AI capabilities developed, the same skill could also be presented in different forms: written explanations, tutoring dialogues, videos, visual presentations, infographics, audio, practice, and assessments.

Instead of being one item buried in a giant list, each skill could have a place of its own—somewhere it could be understood, taught, practiced, and revisited.

But as the library grew, that created another problem.

The Library Became Too Big to Browse

Eventually, the challenge was no longer a shortage of skills.

It was finding the right one.

A teacher looking at one student's essay should not have to search through hundreds of writing skills. A parent helping with homework should not need to know the educational terminology for the difficulty. A learner should not have to diagnose their own problem before they can begin learning.

So the question changed again:

What if people didn't have to search the library? What if they could simply bring the problem?

Bring the assignment.

Bring the reading.

Bring the work sample.

Or simply describe the situation.

SkillsTeacher.AI is designed to examine the actual material or situation and surface a focused set of skills that may matter.

That distinction is important.

A task may require a skill. A work sample may show evidence related to a skill. Another skill may be a likely prerequisite or a possible gap worth checking.

And sometimes there simply isn't enough evidence to know.

The system should not look at one piece of work and declare what a person can or cannot do. It can help organize the possibilities, explain why a skill may matter, and point toward what might be worth investigating next.

AI can help make the skills visible. People still decide.

The teacher knows the student. The parent knows the child. The learner knows parts of the situation no document can reveal. The professional knows the context.

AI can contribute analysis.

People bring judgment.

Then We Realized Skills Don't Stop at School

As a special education teacher, I kept encountering another concern, especially from families thinking about what happens after graduation.

Parents weren't only asking:

Will my child pass English?

They were asking:

Will my child be ready for life?

Can they make a decision? Advocate for themselves? Manage money? Organize their time? Communicate what they need? Navigate transportation? Prepare for work? Solve an everyday problem?

Those are skills too.

Teachers help develop many of them every day. Helping a nervous student stand up and speak involves more than the academic content of a presentation. Helping students choose among several options involves decision-making. Helping them manage a long assignment involves planning and executive functioning.

So the library expanded.

Not because academic skills became less important.

Because life does not organize itself into school subjects.

And students aren't the only people who need skills.

Teachers need skills to teach them. Parents and caregivers need skills to support them. Tutors and specialists need skills to intervene effectively. Principals and school leaders need skills to support teachers, families, and organizations.

Adults continue encountering new situations that require communication, planning, judgment, leadership, organization, and learning.

The same question keeps coming back:

What does this situation require you to know how to do?

What began as an academic Skills Library was becoming something larger: a system for the skills underneath school, work, family, leadership, independence, and life.

But We Don't Want AI to Do the Learning

There is an important contradiction at the center of modern AI.

AI can help people complete work faster than ever.

But completing work and learning how to do the work are not the same thing.

A student can produce a polished answer without understanding it. A person can receive a recommendation without developing judgment. Someone can repeatedly ask a chatbot what to do next without becoming more capable of deciding for themselves.

That is not what we want to build.

SkillsTeacher.AI uses AI because it can be extraordinarily useful for organizing, analyzing, retrieving, adapting, and connecting knowledge.

But the purpose of that technology is not to make people more dependent on technology.

It is the opposite.

Use AI to find the skill. Then help the human learn it.

Read. Discuss. Practice.

Try. Get something wrong. Try again.

Explain your reasoning. Apply the skill somewhere new.

And eventually, need less help.

A finished task is not the same as a learned skill.

The goal isn't simply better output.

The goal is greater human capability.

The AI Isn't the Most Important Part

That may sound strange for a company with AI in its name.

But the most important thing we are building is not a chatbot.

It is the knowledge underneath it.

The skills. Their definitions. Their prerequisites.

Their relationships, explanations, and teaching sequences.

The practice and assessment that help turn an explanation into an ability.

AI gives us a remarkable way to navigate that knowledge.

Instead of expecting a teacher, parent, learner, or professional to know the exact name of the skill they need, the person can begin with something much more natural:

The problem they actually have.

The technology helps make the connection.

The library provides somewhere useful to go.

And the human does the learning.

The underlying Skills Library does not have to begin with AI at all. People can browse it, search it, read it, teach from it, and use its resources directly.

AI makes retrieval easier.

It is not the reason the knowledge matters.

In fact, the AI interaction can sometimes be very short:

Here is the problem. Help me find the relevant skills.

Then leave the chatbot and go learn them.

One Classroom Question Became a Much Bigger Question

SkillsTeacher.AI started with a New York City teacher hearing that his students were “low-skilled” and asking:

What skills, exactly?

That led to another question:

How do we teach them?

And then:

How do we find the right ones when someone actually needs them?

Eventually, the question became much larger:

Could we organize the skills people need across school, work, family, leadership, independence, and life—and make the right ones easier to find when someone actually needs them?

That is the larger project behind SkillsTeacher.AI.

It is ambitious.

It is still being built, reviewed, tested, and improved.

And we would rather be transparent about that than pretend an AI system has every answer.

But the founding belief has remained remarkably simple:

A vague problem becomes more manageable when you can identify what needs to be learned.

So start with the work in front of you.

Start with the situation you're facing.

Get more precise.

Find the skills underneath it.

Learn what comes next.

And build the ability to need less help next time.

AI is the retrieval layer. Human capability is the goal.