The Freeze The System Bought

On Monday, July 8, Chancellor Kamar Samuels of New York City Public Schools sent an email to the principals of the country's largest school system. The message did not announce a program or a partnership. It asked the principals to stop buying software, at least for the moment. Central staff would review new purchase orders closely, and, Samuels wrote plainly, some would be "delayed or denied."¹ The exemption was narrow. Software required for mandated services or the opening of school. Everything else was to wait.

The email is the most concrete action New York has taken in a fifteen-month debate about artificial intelligence in schools. In March, the Department of Education released its first draft of AI guidance, and the reaction was quick and unhappy. In May, Samuels told the City Council that the department had "missed the mark."² The revised guidance was promised for June. June came and went. On June 9, twenty-nine members of the fifty-one-seat council, a coalition running from Republicans to Democratic Socialists, sent a letter to Samuels and Mayor Zohran Mamdani calling for a two-year moratorium on AI in schools.³ A parent-and-educator petition asking for the same has since collected more than four thousand signatures.⁴ Monday's email is what a chancellor does when the policy is not ready and a school year is coming.

The Freeze Underneath The Freeze

There is a familiar reading of a purchasing freeze. A city grew afraid. The pressure won. The tools are out and the pencils are back. That reading is available, and it is not entirely wrong. But look one step closer. Samuels has not banned anything. He has bought his own staff time to see. Officials have already begun surveying principals about which AI systems their schools use. They are building what the department calls a scorecard for teacher-facing AI products, and a more rigorous review process for any new tool.¹ The freeze is not the destination. It is the room where the review can happen.

For a system of roughly nine hundred thousand students, that is not a small piece of furniture. It is the difference between purchasing software and understanding what it does. In the years since ChatGPT, most large districts have bought first and read the fine print later, or asked their principals to. New York's July 8 email is a mundane sentence with an old-fashioned meaning underneath. Wait until we know.

What The Vendors Were Selling, And What The Schools Were Buying

The council members' letter is worth reading not for its politics but for its inventory of harms. Student data privacy, first. Then cognitive development. Creativity. Mental health. The environmental cost of the models themselves.³ It is a longer list than most district AI conversations get to, and it points to a problem the sales cycle rarely surfaces. A school buys software by the feature. It lives with software by the effect. The features are in the demo. The effects show up in the child six months later, often too small to name.

A rigorous review process, in the sense Samuels seems to mean, is a system that can name them. Not a compliance checkbox. A living record of what actually happens between a child and the tool. What did a fifth grader draft on her own before she asked? What did she paste in without reading? Where did the writing become hers again, and where did it stop being hers at all? These are not questions a purchase order can ask. They are questions a workspace can answer, if it is designed to remember.

What A Scorecard Could Be For

That is where we spend our days at Koan. Not building another AI companion to sell to a district, but building the ordinary layer underneath the writing where the process leaves a trace worth reading. A pause before a prompt. A revision. A place where the child noticed the sentence was not hers and took it back. If New York's next scorecard has a line for whether a tool makes the student's thinking visible to her teacher tomorrow morning, we would like the score to be honest. That is the only kind of purchasing decision worth making.

The Monday email will not solve the debate that produced it. The revised guidance is still not out. The moratorium petition is still growing. Samuels' team still has an unresolved argument with almost every parent group in the city. What the freeze buys is a smaller and older thing. A moment to look at what a school year is about to purchase before it purchases it. In a country that has spent two years chasing a technology in a hurry, that moment is not a retreat. It is a rare piece of educational discipline.

If a district could see, on any given Monday, exactly what its software was teaching the children to do without it, what would it want to keep, and what would it quietly return?