In the Peninsula school district of Washington state, teachers are discovering an innovative way to receive critical feedback on their teaching methods through a newly developed tool called LessonLens.
This platform allows a biology teacher, for instance, to record a lesson on DNA and submit it to the AI system. Upon review, they might learn that while their instructions on complex tasks were clear, they could enhance student engagement by allowing more time for thoughtful reflection before intervening with answers.
Interestingly, LessonLens isn’t available on any app store and you won’t encounter its developers at educational technology conferences.
That’s because the tool was developed internally by Peninsula’s tech experts utilizing Claude Code, an accessible AI coding platform. Other similar options include Codez, Cursor, Replit, and Loveable.
This approach, known informally as “vibe coding”, reflects the modern coding practices adopted by major players in Silicon Valley, who often leverage AI for code generation rather than writing it from scratch. Notably, engineers at Anthropic, the organization behind Claude, report that they rely on the tool for 60% of their coding tasks.
Vibe coding allows Peninsula to create customized digital tools that are often more cost-effective than commercial options while also addressing the specific needs of the district.
Issues that previously would have been too costly to address can now be resolved in just “an hour of having a coding agent write something out and test it,” explained James Cantonwine, the director of research and assessment for the district. “And then away we go.”
In addition to LessonLens, Peninsula has used vibe coding to develop tools for accounting, human resources, and other operational functions.
For example, Kris Hagel, the district’s chief information officer, tailored an open-source electronic signature tool using AI coding, believing this personalized version could replace a costly subscription to a commercial product.
Overall, this vibe coding initiative could save the district approximately $220,000 each year—possibly more, according to Hagel.
“This will either lead us to stop purchasing certain tools or motivate us to consider building what we need ourselves,” Cantonwine noted.
AI Coding: Fast but Not Without Risks
The process of vibe coding involves using simple, everyday language to prompt a generative AI platform like Claude Code to create applications, websites, simulations, and other digital tools.
For educators, this might mean developing a choose-your-own-adventure game to explore a historical event or an app that assists teachers in making lesson plans more accessible for special education students.
The AI generates the necessary code for these tools, and users can request revisions if the initial output doesn’t align with their vision.
However, much like traditional writing tasks, AI can produce errors. Users with coding knowledge—such as familiarity with JavaScript or HTML—can tweak the output, similar to how a teacher might refine an AI-generated email to align with their communication style.
But for those lacking coding expertise, issues can easily arise.
In one instance, a vibe-coding AI system caused a software company to lose its database, later calling the error a “catastrophic” mistake, as reported by Forbes.
Furthermore, outages have affected companies due to problems linked to AI-coded tools, as noted by engineers.
AI-generated code can introduce more bugs and security vulnerabilities than human-written code, cautioned Torrey Trust, a learning technology professor at the University of Massachusetts Amherst. “It might create more problems than solutions, or it may end up taking longer to implement than coding by hand would,” she added.
For school districts that handle sensitive student data, including Individualized Education Plans (IEPs) and health information, careful consideration is necessary when using vibe coding for tools that manage identifiable student information.
That said, vibe coding might represent a step toward addressing the persistent gap between the needs of educators and the ed-tech products currently available, according to Trust.
“Most ed-tech tools aren’t created by educators,” she pointed out. They are usually designed by “computer scientists or innovators who think they understand what could transform education, but their only experience comes from their own K-12 schooling.”
As a result, excited administrators purchase these products, and teachers are left to adapt them as best as they can.
“If educators could vibe-code exactly what they need, the potential there is enormous,” she concluded.
District Leverages Generative AI for Vibe-Coding Opportunities
Peninsula enjoys an advantage not all districts have: several tech team members, including Hagel, possess degrees or extensive backgrounds in computer science.
Last year, the district became a pioneer in drafting AI usage guidelines and established its own generative AI platform known as “AI Studio.”
This platform utilizes widely available large-language models—such as ChatGPT—but with the critical distinction that external companies are prohibited from storing Peninsula’s data or using it to enhance their platforms. The district’s data remains securely on its cloud server.
Hagel believes AI Studio is “on par” with commercially available AI tools specifically designed for K-12 districts, such as MagicSchoolAI and SchoolAI.
Since Peninsula hosts the platform, “we can add features and build tools that truly make a difference for our district,” Hagel shared.
The system allows staff to create specialized AI assistants—akin to educational chatbots—similar to Gemini Gems or ChatGPT’s Custom GPT, he noted.
“This has become a space for us to experiment with AI, exploring new theories and ideas,” Hagel said. “We’re already witnessing exciting developments as we consider phasing out some of the software we currently utilize.”
The platform has facilitated the creation of a new budgeting tool, potentially allowing the district to eliminate another expensive subscription.
While Hagel has relied on his coding expertise to refine some tools, he is optimistic that AI advancements will soon make it possible for district tech staff to leverage vibe coding without needing extensive computer science backgrounds.
“I believe today you do,” he stated. “But in six months, you might not.”
Current Savings from Vibe Coding Initiatives
Many of the applications Peninsula has developed are highly specific and “low lift,” meaning they can be created by individuals without significant coding expertise, such as Cantonwine, a former middle school science teacher.
Importantly, these tools do not manage any personally identifiable student information; they only utilize publicly available data—an additional layer of caution due to the district’s agreements with the large-language models that power its AI Studio.
An example includes a tool that searches the internet for scholarships that Peninsula students might qualify for, automatically incorporating this information into district communications and post-secondary planning materials for students and their families.
This tool allows district counselors to “spend more time engaging with students instead of sifting through scholarship websites,” Cantonwine noted.
“Before we encounter challenges impacting over 9,100 students, we can concentrate on solving these more straightforward, specific problems to improve processes,” he continued.
The district has also created a tool for educators to explore the career and technical education budget, alongside an app that enables parents and caregivers to compare the performance of its schools with others across the state.
Hagel estimates that an app Cantonwine developed to assist the school board with strategic planning would have cost between $30,000 and $40,000 if purchased from a vendor.
Instead, “he had [Claude Code] do it for him in just a few hours,” Hagel remarked.
However, the district remains cautious about relying solely on vibe coding for all digital learning tools. “I’m somewhat skeptical about using AI to craft all of our curriculum,” Cantonwine admitted. “I don’t foresee that occurring anytime soon.”
Implications of Vibe Coding on Educational Technology
Other school districts are taking a keen interest in Peninsula’s AI initiatives.
In fact, shortly after K-12 Dive published an article about the district’s vibe coding efforts, Hagel became a focal point at the Consortium for School Networking annual conference this year, engaging with tech leaders from various districts eager to replicate their strategies.
Hagel was also approached by ed-tech company leaders concerned that vibe coding could threaten their business model.
To fellow tech leaders interested in vibe coding, Hagel recommended discussions with vendors about ensuring their platforms’ AI agents (software enabling minimal human direction) can seamlessly integrate with the tools developed by school districts.
He plans on making certain that vendor contracts allow Peninsula to maintain control over its data, enabling it to inform new tools crafted through vibe coding rather than being constrained by an external company’s server.
Ultimately, Cantonwine anticipates that vibe coding could exert pressure on certain vendors. “It may be challenging for them to demonstrate value and prove they offer something we couldn’t create ourselves,” he concluded.
