3 Crazy Happenings in Northwest Arkansas, Including Artificial Intelligence

I cited a statistic about AI teaching children that I could not source, hedged it on air, and said it anyway. A year later I have gone looking for that number and cannot find it. That is the section that matters here, so let me get the two easy ones out of the way first.

1. Maple Street is finished, and they narrowed it rather than widened it

The project is real and larger than I described. It is the Maple Street Improvements project, a joint effort between the City of Fayetteville, the University of Arkansas and the U.S. Department of Transportation, funded largely through a federal Safe Streets and Roads for All grant, with city and university money on top. Total cost is around $33.5 million for about a half mile between the Razorback Greenway and Garland Avenue.

I said they were "expanding the road." They did the opposite. The design right-sizes the travel lanes — makes them narrower — and uses the space for a greenspace buffer, better crosswalks and a two-way protected bike path. That is a road diet, not a widening, and the distinction is the whole point of the project. If you drive Maple expecting more lanes, you will be surprised.

My timeline was right. Phase I, Gregg to Leverett, ran through summer 2025 and finished by year end. Phase II, Garland to Leverett, ran through summer 2026 and reopened to traffic around the start of August, before students came back. Some sidewalk work continues into fall. I was also right that they schedule around the academic calendar.

One small fix: I called Maple "the runway for all the fraternities." It is more accurately sorority row — Alpha Chi Omega, Phi Mu, Delta Gamma, Pi Beta Phi, and Alpha Phi all line it.

2. The cycling accelerator is real, and much smaller than I said

The Arkansas Global Cycling Accelerator exists. I overstated its size by roughly two orders of magnitude.

I said they were "inviting so many hundreds of companies across the world." The first cohort was 10 companies, selected from 64 applications across 15 countries. A second cohort ran in spring 2026, also 10 companies — from Belgium, Canada, France, Norway, the UK, and the US — with a Demo Day at Compton Gardens on June 10, 2026, timed to the NBDA Summit and Bentonville Bike Fest.

It is run by the University of Arkansas's Greenhouse Outdoor Recreation Program together with Startup Junkie, funded by a $125,000 Arkansas Economic Development Commission grant plus matching Walton Family Foundation money. The Ledger is the demo-night venue rather than the program's home, which is a fair thing to be loose about since the first demo night was in fact there in May 2025.

A cohort of ten companies is still a real accelerator. It just is not hundreds, and a founder deciding whether to apply deserves the actual odds.

On whether Bentonville is genuinely a cycling-industry cluster: it is. Vittoria, Rapha, Allied Cycle Works, and Bentonville Bicycle Company all have operations here, the U.S. National Mountain Bike Team is based here, and the city runs a cycling business incubator. I could not confirm whether any accelerator alumni have opened permanent Bentonville locations, so I will not claim they have.

3. The AI education claim — I could not find the study

Here is what I said. The Heartland Summit had announced an AI teaching initiative, and I explained why I thought it mattered by citing "a study out of Texas" where each student had a personal AI, there were very few teachers, and results showed "a 63% increase of knowledge base and quickness of learning versus actually doing normal teacher stuff."

I did add, immediately: "whether they're true or if there's a special settlement there, I'm not sure." I want credit for that hedge and I do not think it is enough. A number said out loud travels further than the caveat attached to it.

First, what was actually announced. The organization is **Heartland Forward — I called it "the Harland Institute" — a Bentonville-based policy group founded in 2019 and backed by Walton family money. At the 2025 Heartland Summit they announced a partnership with an edtech company called Stemuli to build a gamified platform teaching **AI literacy and prompting skills to rural students and entrepreneurs across the heartland.

That is a meaningfully different thing from what I described. It teaches students how to use AI. It is not AI replacing teachers. I collapsed those two ideas together, then reached for a statistic about AI replacing teachers to justify my enthusiasm for a program that actually teaches AI literacy.

Second, the 63 percent. I cannot find it. The study I was almost certainly half-remembering is Alpha School in Austin, Texas, and its "2 Hour Learning" model. Alpha publishes claims like "2x faster learning," "2.6x average MAP growth" and "99th percentile" — different numbers, different framings, and none of them 63 percent. So the figure I gave appears to be a garbled version of a claim that itself needs scrutiny.

Third, those claims are not independent research — and this is the part I should have known before repeating any of them. They come from Alpha's own internal analysis of its own students' NWEA MAP test data. No peer review. No external audit. The critiques are substantial and come from serious people. The National Education Policy Center and others have flagged reliance on conditional growth percentiles rather than absolute score gains — a metric where a high-performing student who barely improves can still post a big number — along with very small samples for some flagship claims, no published underlying data, and reports of prizes tied to test performance, which inflates scores without necessarily reflecting learning.

There is also the obvious confound. Alpha charges between $10,000 and $75,000 a year, with most campuses around $40,000. Justin Reich of MIT's Teaching Systems Lab has pointed out that comparing those students against the national MAP norm — mostly public school children in far harder circumstances — is not a fair comparison. Motivated families paying $40,000 a year would likely outperform the national average at a school with no AI in it at all.

Fourth, a local fact I did not have. Unbound Academy, the operator using the Alpha model, applied for a charter in Arkansas and was rejected by the State Board of Education in 2024. Board member Tina Moore said the application "did not demonstrate a viable plan, relevant experience or readiness to provide a high-quality learning environment for Arkansas students." Arkansas was one of five states to deny it. There is no Alpha-model school operating in this state.

Fifth, my actual argument was backwards. I said the appeal for rural areas is that "they can't always get the talent to go to nowhere Arkansas... but AI doesn't really care" — that AI works precisely where adults are scarce. The independent evidence points the other way, though I want to be careful not to replace one confident number with another. The research I can find on AI tutoring reports effects broadly in the range of human tutoring rather than dramatically beyond it, and the strongest results consistently come from blended models — AI handling diagnostics and practice while a human provides coaching and motivation. One Google DeepMind study found AI-assisted human tutors performed about as well as unassisted human tutors, which is AI helping teachers rather than substituting for their absence. What I could not find anywhere is independent evidence that fewer adults produces better outcomes. That was the load-bearing part of my argument, and it is the part with nothing under it.

The teacher shortage I was reacting to is entirely real. Arkansas classifies a district as high-need when 20 percent or more of its teachers are unlicensed or leaving, and roughly one in twelve Arkansas teachers is unlicensed or teaching outside their certification, concentrated in rural districts and the Delta, especially in secondary science, math and foreign languages. The state raised minimum teacher pay from $36,000 to $50,000 in 2023, which narrowed the gap without closing it. DESE publishes the shortage district list every year.

So the problem is genuine and the interest in AI as part of the answer is reasonable. What I owe you is the difference between "this is a promising tool that independent research says works best alongside teachers" and "a study showed 63 percent." Only one of those is true, and I said the other one.