RAIN — Week of October 26, 2026

CRAI News

Please welcome Justin Thornbladh, our new Partnerships & Business Development Lead! Justin has spent more than 20 years in senior leadership roles in telecommunications, enterprise technology, and strategic alliances. At CRAI, he'll lead our partnerships work, building relationships with organizations that can help us get practical AI tools and training into more rural communities. Justin lives in Durango, so this work is close to home for him. Join us in welcoming him to the team!

Ask Amy

After discovering the limits of Gemini, I think I’m finally going to give Claude a try. Which paid version would you suggest I use? I use AI for building classroom exercises on how to use AI and for doing some consulting secondary research.    Loretta in Texas

Congratulations Loretta on being such a good user that you know you have reached the limits of the technology. Gemini is a great tool especially inside of the Google environment. For what you’re describing, Claude Pro at about $20 a month is the one to try. I prefer to keep my subscriptions on monthly billing so I have the freedom to move to another model easily.

With a paid plan you have access to more than 5 project spaces which can hold your course context so you aren’t re-explaining your class setup every time. You also get the Research tool. For those readers who don’t know, secondary research is research done on material that already exists. Think gathering up published reports, market data, trade press and academic papers, government statistics and then writing a report.

The Research tool in Claude investigates a question in depth. It runs multiple searches that build on each other and decides what to look into next based on the findings. The best part is that the output comes back with citations attached. You will still need to verify them all before you use them.

The difference from ordinary web search is that search answers the question you asked. Claude’s Research tool breaks your question into sub-questions, searches each one, reads what comes back, notices the gaps, and searches again. It’s the difference between one lookup and having a bazillion tabs open while you track down this line of thinking and then the next.

AI tools and models have different “talents”. To Loretta and to all my readers, find the tool that works best for the work you are doing. Don’t be afraid to experiment and jump to a different model. If Claude doesn’t resonate then give Chat GPT a try.

Until next week,

Amy

Every column starts with a reader question. Yours can be next: ruralai.org/rain

Stories of the Week

Farm robots aren't replacing workers as quickly as expected

A UC Davis study of 16 fruit operations in California and Washington found AI fruit-counting systems, maturity sensors and robotic harvesters still depend on human judgment. One experimental robotic harvester ran at under a third of a crew's speed with twice the cull rate.

Growers should budget for retraining, not headcount cuts: platform operators at one California orchard earned 15 to 25 percent more than in their prior roles. Farms that let workers help evaluate equipment before purchase had smoother rollouts.

Viewpoint: AI data centers are putting America's farmland and water at risk

Kansas rancher Brandi Buzzard debunks a viral claim that a Texas data center stopped calf births, then lays out the documented case: Pew finds 67 percent of planned US data centers are headed to rural areas, and a hyperscale AI campus can use 5 million gallons of water a day.

In Virginia, data center developers outbid young farmers for acreage, and a quarter of centers sit in water-stressed regions. About $260 billion in projects have been canceled or blocked this year, largely through local petitions and permit fights.

Nathan Leamer: Can America Build AI and Rural Broadband at the Same Time?

Fiber prices have climbed as much as 50 percent this year as hyperscalers lock up supply, including Meta's $6 billion Corning deal, just as $42.45 billion in BEAD construction gets underway. CoBank warns smaller rural operators may not get materials at workable prices.

Leamer wants NTIA to spend BEAD non-deployment dollars training splicers and line crews, with the industry already short 28,000 construction workers and 30,000 technicians. Sen. Capito's permitting bill gets a vote after the Nov. 3 elections and covers energy, not fiber.

AI resources and classroom lessons

Three NSF-funded Advanced Technological Education centers are giving community colleges free AI curricula, mentorships and faculty training, including 19 online sessions running through mid-November. The National Applied AI Consortium has reached 3,500 educators at 480 colleges since 2024.

Marion Technical College's "Empowering Rural Communities" project runs free AI workshops for small business owners; one skeptic built a customer service system that saved his company $200,000. It has also launched four AI courses and a micro-certificate.

CSU taps AI to predict hailstorms days in advance

Colorado State researchers are pairing an NVIDIA Earth-2 nowcasting model with a machine learning system trained on nine years of US storm data to forecast hail days out instead of 30 minutes. Hail-StormScope targets two to three hours; the longer-range model covers one to eight days.

Colorado and Wyoming sit in a hail hotspot where storms shred crops and have pushed insurers to raise rates statewide. Funding runs through the ASCEND Engine, a 10-year federal award worth up to $160 million.

Opinion: States Don't Have to Solve the AI Education Equation Alone. Tennessee Shows How

Only 43 percent of 29 surveyed state education agencies offer districts hands-on AI implementation help such as procurement or vendor vetting, and just 11 percent have reached what researchers call the integrator stage. Tennessee leaned on SCORE, a statewide nonprofit, to fill the gap.

SCORE helped fund a Coursemojo literacy pilot that grew from one district to seven districts, 20 schools and more than 4,000 students. Without an intermediary, small and rural districts get stuck vetting AI vendors alone while better-resourced systems experiment.