Skip to main content
Peninsula School District · Gig Harbor, WashingtonOpen source · free to run
Peninsula AI
Gig Harbor, Washington · since 2023

A public school district doing its AI work in the open.

Peninsula School District uses AI in teaching, operations, policy and professional learning. We publish what we build, what we decide and what we learn, including what didn’t work.

High school students working at laptops, two leaning together over one screen while a teacher listens
16Posts written by staff
45Talks and slide decks published
7Products in production
166Documents published here
01

Guidance

4 documents

The guidance our own staff work from, in plain language. Other districts are welcome to adapt it, with credit to Peninsula School District.

Guidance2025-01-30Data Security With AIFour levels of AI tools, each with a different level of data security, and what each means for working with student data.Guidance2024-08-09Peninsula School District AI Principles and BeliefsPeninsula School District's AI principles and beliefs. AI enhances and streamlines the efforts of human instructors rather than substituting them.Guidance2024-08-30Student Rights and Responsibilities Handbook LanguageHandbook language on students’ rights and responsibilities when using AI tools such as ChatGPT, in line with UDL and ethical AI use.Guidance2024-08-09Sample Syllabus Language for TeachersArtificial Intelligence (AI) Use in the Classroom: Guidelines and Responsibilities
02

Writing

16 posts

Notes from the people doing the work.

Information Sharing · Sep 29, 2026Building the Open Adaptive District: What We're Doing and Why It Matters for Public EducationSuperintendent Krestin Bahr on what Peninsula School District is building through the Open Adaptive District, and why schools need a voice in how AI is built.Krestin Bahr
Information Sharing · May 26, 2026Build Products Our Agents Can UseA call to EdTech vendors, districts and standards bodies: make products safe for district-owned agents, or districts will build the replacements.Kris Hagel
AI in Action · Dec 10, 2025Feedback in the Age of AI: More Thinking, More Learning, Less NoiseDoes including AI feedback increase the chances that students will think hard about the content and improve their future performance?James Cantonwine
03

Presentations

45 talks, workshops and board sessions, with the slides as they were presented.

Puget Sound in early morning fog, conifers on the headland, a low school building in the middle distance
04

Software we build

7 products

Built by district staff for district problems, and all open source, so another district can run them for free.

Next.js 16 · AWSPRODUCTIONAI StudioFrontier AI models for every staff member and student in the district, behind district sign-in, at about a tenth of the cost of per-seat licences.Next.js · AWSPRODUCTIONAtriumA workspace in AI Studio where staff and their agents write documents and build pages together, then publish them to the intranet or the web.Chrome extension · macOSPRODUCTIONAtrium CaptureRecord a workflow in Chrome or on a Mac and turn it into an editable step-by-step guide, saved to Atrium as a private draft — without ever capturing a password.SwiftUI · Cloud Run · GeminiPRODUCTIONLessonLensA private coaching tool for teachers — record or import a lesson and get AI feedback on specific teaching techniques that no administrator or evaluator can see.OpenClaw · AWS · Google WorkspacePRODUCTIONPSD AI AgentsA personal AI agent for district staff, with its own district Google account, that works on its owner's behalf in Google Chat — the way an assistant would.Next.js · Expo · AWSPRODUCTIONPSD EOCEmergency notification and operations for districts: start an incident or drill, notify staff, work it in a live timeline, keep an append-only record.Django · React · CeleryPRODUCTIONPSD PRR RecordsOur public records request system, built to replace NextRequest — Washington's deadlines, exemptions and audit trail enforced in code rather than by policy.
05

The Open Adaptive District

Our protocol for AI work, open for any district to adopt.

A team of three to six staff picks one problem, spends six weeks building a better way to handle it, and publishes what happened, including what didn’t work.

School staff mid-working-session around a table, a loop of sticky notes on the whiteboard behind them
  1. WEEK 1PlanPick one problem and write a one-page build plan.
  2. WEEKS 2–5DoBuild it, use it in real work, and check in every Friday.
  3. WEEK 6StudyAn AI agent drafts the wrap-up from the team’s own notes.
  4. WEEK 6SharePublish the wrap-up and make the call: keep it, stop it, or run it again.
Start here

The reference library

50 use cases from staff, 12 tools we have reviewed, and 26 summaries of outside research.

We encourage educators to use professional judgment, even if it means questioning or overriding an AI tool’s decisions, maintaining the critical human element in educational decision-making.

— AI PRINCIPLES AND BELIEFS

Built for agents to read

Every page on this site is also published as plain text an AI agent can read, in the Open Knowledge Format. Point your agent at it; no custom integration is needed.

SEARCH · LIBRARY · RSS