The Education Gadfly Weekly

2025 will be remembered for a lot of bad—no, terrible—news outside of education, but inside our little sandbox of ed-reform, there’s plenty to be grateful for.

Let’s start with the nice before getting to the naughty:

Pros

• It’s the year that opinion elites woke up to America’s educational malaise, blamed it on more than the Covid pandemic, and pointed to counterexamples from Mississippi and the rest of the surging South as proof that we can do better. Here’s a sample of what I’m referring to:

  • David Brooks, “America’s New Segregation,” New York Times, August 14

  • Matthew Yglesias, “American students are getting dumber. It started before Covid, and it keeps getting worse,” Slow Boring, September 22

  • Kelsey Piper, “Illiteracy is a policy choice. Why aren’t we gathering behind Mississippi’s banner?” The Argument, September 25

  • David Brooks, “Why Are the Democrats Increasing Inequality?” New York Times, October 9

  • Idrees Kahloon, “America Is Sliding Toward Illiteracy,” The Atlantic, October 14

  • Andrew Rice, “The Big Fail: Public Education Is Crumbling,” New York Magazine, November 20

  • Megan McArdle, “The signs of educational decline are now impossible to ignore,” Washington Post, November 23

  • Jonathan Chait, “Why the Left Stopped Talking About Achievement Gaps,” The Atlantic, December 8

  This matters, I believe, because it’s the first step in getting policymakers to care about education again. And if a presidential candidate makes education reform a key issue in 2028, these opinion-makers will deserve much of the credit.




• The backlash to grade inflation, cheating, and phones in school is here (and was itself reflected in the many opinion pieces mentioned above). We learned, for example, that an overwhelming majority of teachers oppose “no zeros” policies and other grading “reforms” that result in lower expectations.




• The science of reading continued its march through the state legislatures, with even California enacting a decent law, and practitioners embracing the importance of implementation. If we can get this one right, and help millions of students learn to read effectively, everything else we’re trying to do will be a whole lot easier.




• Houston keeps crushing it under state-appointed superintendent Mike Miles, and now the Texas Education Administration is preparing to work its magic with Fort Worth’s schools, too.




• The reform movement dodged a bullet on religious charter schools. A victory for St. Isidore could have splintered the charter sector, but we were rescued by Amy Coney Barrett’s recusal. (Unfortunately, we might not be so lucky next time.)




• Both Texas and Alabama enacted historic school funding reforms, with the Lone Star State in particular demonstrating a creative path forward on teacher compensation reform via its expanded Teacher Incentive Allotment.




• Indiana made strides towards a huge breakthrough for charter schools, with plans underway for facilities and transportation in Indianapolis to be managed independently of the school district, putting charters on equal footing with traditional public schools. As Jed Wallace argues, this is an excellent “North Star” for charter advocates everywhere.




• Some large districts finally got serious about closing underenrolled schools, which is never a happy process but is essential if our system is going to focus its resources on serving students best. (Of course, plenty of other districts still have their heads in the sand.)




• The teacher shortage all but evaporated, thanks to rising unemployment for college graduates. This is the one silver lining of a weak job market, given that teachers hired during downturns like this tend to be more effective than their peers.




• The U.S. Scholarship Tax Credit in the One Big Beautiful Bill could open the door to better schools for millions of students. This one is definitely TBD, as a lot depends on implementation, which is just now getting underway. And the initiative’s lack of accountability for results is a big miss. (Same goes for Texas’s big new program.) But if lots of low-income kids in cities like New York, Denver, Los Angeles, and Chicago gain access to Catholic schools and other high-quality options, it will be beautiful indeed.

Cons

• President Trump’s immigration raids scared families and children, and contributed to our ongoing chronic absenteeism crisis.




• Kansas and other states lowered their thresholds for what it means for a student to be “proficient” in reading and/or math this year, adding to the growing list of states that are returning to the soft bigotry of low expectations and widening the “honesty gap.”




• DOGE bulldozed the Institute for Education Sciences, putting at risk our sector’s R & D infrastructure, as well as the Nation’s Report Card. Stay tuned, we hope, for our colleague Amber Northern’s recommendations on how to put Humpty Dumpty back together again.




• Too many schools embraced dubious instructional uses of AI—at the very time we’re seeing the disastrous results of incorporating phones and other devices into our classrooms.




• Socialists are now in charge in NYC and Chicago. Maybe Zohran Mamdani will surprise us reformers and embrace charter schools and other smart policies, but if Chicago is any indicator, the Big Apple is in for a rough four years.

It’s close, but to my eye, that adds up to more light than darkness for 2025—and reason for hope as we enter 2026.

This year, I am conducting the AI in Action Learning Tour to see student-facing AI solutions in classrooms and to learn about what the product developers are wrapping around the large language models in the design of the apps. I am doing this because I believe, though educators will always be instrumental in teaching and learning, student-facing AI can make it easier for teachers to be great. Earlier this year, I shared early patterns from what I am learning and committed to sharing more as I go.

After completing another round of product conversations, I see more clearly three defining instructional choice points that I predict will drive trends in the year to come.

1. Product designers either presume teacher autonomy or design for common implementation at a district or school level. The products that presume teacher autonomy are largely selling to teachers with freemium models, then encouraging districts to purchase enterprise accounts. Teacher-facing tools include functions like lesson planning, text selection, and materials development. The student-facing sides of these products are designed for flexible use-cases. Other products presume that districts are giving teachers direction about what materials they should be using, with day-by-day or week-by-week pacing expectations. Their products are designed to become part of that model. They are selling, not classroom-based, but district-wide solutions. Products are designed to fit into a defined instructional model; they intentionally do not add features that would encourage modification. Currently, more AI products on the market are designed around teacher autonomy than common implementation.

2. Product designers either design features to “meet students where they are” or anchor to grade-level instruction. Many educators and parents hold the belief that good teachers target instruction to the student’s current academic level. Fueled by reports like the Opportunity Myth and analysis of the difference between acceleration and remediation, many state and district leaders have been trying to replace that belief with the idea that all students can meet high expectations and will make more academic progress if they’re given access to grade-level instruction. Product developers seem to subscribe to one belief or the other—those designed to support teachers in “meeting students where they are” offer tools like text releveling. Those designed to provide grade-level instruction steer away from such choices and launch all instruction from a grade-level starting point. There are currently more products that try to “meet students where they are” than anchor to grade-level instruction.

3. Products are either built to do a single instructional job or designed to have multiple coherent touch points. Some products try to do one job well, usually across grades and subjects—for example, nailing exit tickets. Other solutions try to help with a full stack of instructional jobs—for example, using Tier I instruction and assessment with teacher PL in middle school ELA. Most products start with a lean design but end up adding more functionality.

In 2026, I predict that student learning gains will correlate with the quality and specificity of instructional leadership and teacher support more than product selection; however, it will be a year that shapes AI’s future role in teaching and learning. Here is what I suspect we will see:

1. I predict that the teacher autonomy tools will win the early adoption surge but will begin to lose market share to products designed for district-wide use. It will be much harder for teacher autonomy tools to prove impact than common use solutions—and the more comprehensive instructional model-based tools will start to show outcomes more clearly. AI has some superpowers in analysis and personalization that can improve coherence when used across classrooms. Common applications can also give leaders visibility into what students are doing and learning across the school or system. Districts will only be able to leverage those insights if everyone is using the same products in similar ways. Ultimately, districts hold the power of the purse. I suspect we will see more muscular district guidance on what tools they expect teachers to use and which tools they don’t want to see influencing instruction.

2. I predict that existing curriculum, assessment, and intervention providers will start offering AI-powered features. I sense that  many companies have been waiting to see how AI unfolds, but they’ll begin increasing pressure to implement their own AI solutions—as well as a sense of possibility about how AI can improve their services. They will have a leg up with an existing client base, but I don’t know how much districts will trust evolutions versus purpose-built products. I predict some will allow for the “meet students where they are” functionalities, while others will resist demand for these features and this, eventually, becomes a flash point in selection.

3. I predict that instructional super products will begin to emerge that transcend categories (e.g., serving as curriculum, assessment, and intervention with embedded PL all in one). I predict that district and state leaders will get most excited about the way these superproducts can improve coherence, reduce the volume of testing, and displace cost categories rather than just adding more log-ins or budget line items.

4. I predict that some districts, teachers, and parents will opt out of AI entirely. The pressure against screen time is real, and the risks of cheating that come with access to any browser will likely lead to bifurcation of interest and use. I think some parents will choose schools based on technology use or disuse. I predict this will become a factor in teacher job interest—some teachers will favor districts that give them access to AI tools, and others teachers will favor districts where they can teach “old school.” Ultimately, I do not think this will be a large group because the benefits of strong tools to student learning and the teacher job will be so evident and significant.

My record of predictions is wrong as often as it is right. Either way, I look forward to learning more together in 2026. I wish you and your families a healthy and happy holiday season.

Editor’s note: This was first published by Instruction Partners.

For my first four years in the classroom, I taught English as a second language to Somali refugees and Latin American immigrants in a small city in Wisconsin. I offer my experience teaching these children not as an omniscient viewpoint but as an informative sliver of a perspective.

Which immigrants should we let in or remove? On what criteria? What benefits should we provide or not? While important, answering those questions is not this article’s focus.

Instead, I’m interested in how many immigrants interact with our institutions once they’re here. What do they learn about our country? How do we shape them or not? For many young immigrants, their first experience with America is the public education system, an experience that is terribly flawed.

The primary issue was chaos. Students were in constant conflict across cultures. They expressed verbal hatred toward one another, and that animosity spread to violence in the hallways. I still recall talking one Hispanic student down from choking a Somali student unconscious. You don’t quickly forget the sound of a 12-year-old gagging and wheezing.

Unfortunately, the district administration was too sheepish to enforce order. Who were they to impose their cultural norms? What were the optics of suspending a refugee? They instead tried various restorative circles and listening sessions to build community. At one listening session, a mother said that her children feared other students, yet seemingly no one was ever punished.

The tribal hatreds from their home countries didn’t suddenly disappear once they came to my classroom at the end of the hall. Warring factions threatened to disrupt the school. And without school leaders enforcing rules, these conflicts festered. Instead of softening, the hatreds hardened.

The kumbaya strategies failed, and so these refugee students walked into a Hobbesian building where only strength and clannishness could protect them from bullying and fights. What is their primary, formative experience with America then? A public education system flush with resources — they previously attended small shack schools in refugee camps — but terribly violent. A place where rules went unenforced and strong-arm tactics were tolerated.

The second biggest issue I encountered, which led to the first, was the progressive pedagogical mindset among the teachers. Grammar instruction, memorization of new vocabulary, rote practice, and explicit, systematic phonics instruction were shunned. Should we really be forcing them to learn English, anyway?

Instead, we encouraged students to learn content in their first language. All reading and writing was “taught” through exploration and self-expression. For spoken language, such “learning by osmosis” can work, but it fails dramatically when learning how to write. As a result, these children didn’t learn English, didn’t learn any other academic content, and were forever on the peripheries of the school. Without English, they were always “other.”

Third observation: Immigrant students received a very negative perspective on their adoptive country. More than an honest account of history, they heard a narrative that cast America’s story as one of slavery, impoverishment, and evil. Any suggestion that gratitude is an appropriate emotion, that America is a great country, or that they should adopt the ideals in our Constitution was anathema.

It’s important to note that many of these were deeply troubled students who had come from traumatic and unstable circumstances. While that observation may offend some, it’s a reality that schools cannot ignore. They have to reckon with it whether they like it or not. I had many students whose parents were brutally murdered by cartels or whose fathers were missing limbs from a warlord’s machete.

As a result, these students were more prone to violent outbursts and volatile emotions. It made a school full of them difficult to manage, let alone any efforts at acculturation or assimilation. It was even harder as new students came to our classrooms unexpectedly in the middle of the school year, often with little English and sometimes no paperwork to confirm their age.

One veteran teacher contested these policies. She is herself a Latin American. She saw that students needed explicit instruction, that they needed to learn English first, that they needed strong behavioral systems, and that America was a good country. But eventually, the stressors of teaching in this environment and her unheeded calls for improvement overwhelmed her. She checked into a mental health institution sometime around spring break and never came back.

I have nothing but sympathy for these students. They are innocent of their country’s lawlessness and don’t deserve the hardships they’ve endured. We cannot return to a forced assimilation that beats students for speaking their home language, but we cannot then spurn helping them adopt different behavioral norms, learn a new language, or develop an appreciation for their new home. To do so only compounds their hardships.

Editor’s note: This was first published by the Washington Examiner.

In the 1990s, Sweden overhauled its public education system. Previously run almost entirely by local governments, public schools could now be operated either by districts or by private providers—similar to U.S. charter schools. Unlike the United States, Sweden’s compulsory schooling lasts 10 years, followed by an optional three-year high school program. Students rank their preferred high schools and are admitted based on their grade 7–9 GPA rather than residential zoning. By 2023, nearly half of Stockholm high school students attended charter-like schools, and roughly 80 percent of them were enrolled in for-profit providers.

In a new study, Petter Berg of the Stockholm School of Economics examines long-run economic outcomes for Stockholm’s high school students who attended for-profit or non-profit charter schools. He links data from roughly 20,000 students who enrolled in 158 high schools between 1995 and 2008 with administrative data on their graduation, college enrollment, field of study, and labor earnings at ages 23 and 30.

Using a value-added regression framework that controls for observable differences in students’ prior achievement and family background, Berg estimates three effects on long-term outcomes: school value-added (school-wide features such as teacher quality), program value-added (market-wide returns associated with different fields of study), and the school-program match (how well a specific school delivers a certain program). To understand why students choose certain schools, he also analyzes rank-ordered school applications from 2013–2015, which reveal how location, programs, and admissions constraints shape demand for for-profit and non-profit charters.

The topline finding is that Swedish charter schools reduce on-time high school graduation by 2.3 percentage points and reduce college enrollment by 3.2 points. Charter attendance also lowers long-run earnings by 1.11 percentiles (roughly 2 percent) relative to traditional public schools, with non-profit charters producing larger losses (1.5 percentiles) than for-profits (0.8–0.9 percentiles).

The mechanisms underlying these losses differ sharply by sector. For-profit charters losses are driven by weak school value-added. Despite offering higher-return vocational programs, cost-cutting practices such as hiring younger, less-credentialed teachers reduce overall quality and diminish student earnings. For non-profit charters, earnings reductions stem from a disproportionate focus on programs, such as art, that yield lower labor-market returns than the academic or vocational tracks offered in public schools or for-profit charters. When program type is held constant, non-profits yield similar or even slightly better school-value added compared to traditional public schools.

Given these outcomes, why do students choose charters? Berg finds that choices are largely driven by location and program availability rather than effects on long-term outcomes, with students showing little willingness to travel farther for higher-value schools. Low-GPA students often end up in for-profits not by choice but because GPA-based admissions and capacity limits in nearby public schools leave them with fewer options. Though top-tercile students typically attend their first or second choice school and bottom-tercile students attend their fifth or sixth, low GPA students in for-profits enroll in higher-return programs than their public-school peers. The reverse is true for high-GPA students. As a result, for-profit charters are nearly as effective as public schools at boosting long-run earnings for students lower GPA students but are significantly worse for top performers.

So what are the implications for charter schools in the United States?

Sweden and America’s charter sectors differ in countless ways. Far fewer American charter schools are run for profit compared to the Swedish system. Most U.S. charters use lotteries for admission, whereas students in Sweden are admitted largely based on GPA. In fact, Swedish “high schools” may have more in common with America’s for-profit colleges as they do with U.S. charter networks like KIPP or IDEA.

Some important questions are difficult to answer. Would students in non-profit arts schools have chosen these low-return programs without expanded choice policies? If not, school choice reforms may be reshaping long-term career trajectories by increasing access to programs with poor labor-market returns. If students would have selected these programs regardless, then the issue may not lie with non-profit charters themselves but with program demand.

The Swedish case underscores that school choice must be paired with careful attention to program quality and outcomes, a major lesson for the U.S. charter sector.

SOURCE: Petter Berg, “Schooling for Profit: Long-run Effects of Private Providers in Public Education,” Stockholm School of Economics (November 2025).