A curriculum map of AI-powered student projects aligned to Texas TEKS — from Kindergarten through Grade 12.
Vibe-coding means describing what you want in plain language and letting AI generate it — then iterating, evaluating, and improving the result. For students, it's not a shortcut. It's a thinking scaffold.
At each grade band, the complexity shifts: in K–2, the teacher holds the prompt and the class shapes the output together. By high school, students write specs, deploy working apps, and defend every AI-assisted decision publicly.
Every project below requires students to prompt → evaluate → revise. AI produces the first draft. Students are responsible for everything that ships.
Students describe what they need in precise language — the first act of computational thinking.
AI output is never accepted at face value. Students fact-check, annotate, and identify errors.
Students improve the output using content-area knowledge — the real learning moment.
At every grade band, work is shared — with the class, a community partner, or the public.
How vibe-coding evolves from guided exploration to professional-grade production
The teacher controls the prompt. The whole class shapes the output together. No student logins required — just a projected screen and curious minds.
Kids dictate a 3-sentence story using weekly sight words. AI generates it with illustrations. Students illustrate a favorite part by hand.
Students say a word, ask AI to make a rhyming poem, then illustrate one stanza. Class compares which rhymes feel natural vs. forced.
Kids describe a scene ("a house made of squares and a triangle roof") → AI or tool generates it. Students label each shape and its attributes.
Students dictate: "Show me 7 frogs." AI generates image + equation. Class builds a counting book, one number per page.
Kids tell AI what today's weather looks like; it writes a weather report they read aloud and record. Teacher uses a fill-in prompt template to scaffold.
Students narrate each stage of a butterfly or frog life cycle → AI generates one sentence + image per slide. Students sequence and label.
Kids describe a job; AI writes what that person does and generates an image. Class builds a "Community Helpers" book together.
Each student picks a Texas symbol (bluebonnet, mockingbird, Lone Star), prompts AI for a "museum label" description, then designs the label.
Give students an AI-generated paragraph with one deliberate factual or grammar error. Their job: find it and fix it before the class "publishes" the work. Builds critical AI literacy from day one, aligns to ELA revision TEKS, and teaches kids that AI isn't always right — without a single lecture about it.
Students type their own prompts with scaffolding, iterate on AI outputs, and evaluate results critically. The loop is always: Student → AI → Student. AI drafts; students publish.
Students write a paragraph first, give the same prompt to AI, compare outputs side by side, and revise their own draft — not the AI's version.
Students paste a nonfiction article, ask AI to summarize, then fact-check the summary against the original and mark every error they find.
Take one historical event or science concept and prompt AI to write it as a poem, then a news article, then a diary entry. Students analyze how genre changes meaning.
Students prompt AI to generate 5 problems at grade level, solve them, rate which were too easy/hard/just right, then write a better one themselves.
Students collect real classroom data, enter it, prompt AI to describe the trend, then build their own graph and write their own interpretation.
Students prompt AI to write a script explaining a concept (area vs. perimeter, equivalent fractions), then record themselves — correcting AI errors before filming.
Students describe a Texas ecosystem, ask AI to generate a food web description, then diagram it manually and identify any missing or wrong connections.
Students prompt AI to explain a states-of-matter scenario, then annotate the output — highlighting what's correct, incomplete, and what needs a diagram.
Students invent a fictional animal for a specific Texas biome, prompt AI to elaborate on survival advantages, then create a field guide entry with illustrations.
Students research one Texas history figure, feed notes to AI with "turn these into a 90-second podcast script," then revise for accuracy before recording.
Students describe a simple business scenario and ask AI to model three economic decisions. They evaluate trade-offs and justify their best option in writing.
Students prompt AI to draft a travel brochure for an assigned region, fact-check it against a textbook or vetted source, then design the final version.
Flip the dynamic entirely. Give every student the same AI output and several different prompts that could have produced it. Their job: figure out which prompt generated that response and rank all prompts from weakest to strongest. No coding, no tool login — just analytical thinking about how language shapes AI output. It builds prompt literacy, inference skills, and argument writing simultaneously, in one class period with zero setup.
Students write and run actual AI-generated code, build functional tools, and work with real data. The standard: if it isn't run, it hasn't been tested. One working script they debugged beats ten AI outputs they passively accepted.
Students find two articles covering the same event from different outlets. They prompt AI to analyze both for loaded language, then write their own comparison arguing which source is more reliable — with specific evidence.
Students use AI to generate a counterargument to their own thesis, then write a rebuttal. Forces them to steelman the opposition before finalizing their essay.
Students vibe-code a simple Q&A bot "trained" on a class novel's plot and themes via system prompt. They test it and document where it fails — and why.
Students describe a real-world ratio problem, ask AI to generate a working Python or JavaScript calculator, run it, find edge cases where it breaks, and fix one thing themselves.
Students import a real dataset, prompt AI to generate a visualization script, run it, and present findings with a written interpretation they authored themselves.
Students prompt AI to solve a multi-step equation and explain every step. Then they swap problems with a partner and verify whether the AI's reasoning is correct or flawed.
Students write their own geometric proof, then prompt AI to check their reasoning. They annotate where AI agrees, disagrees, or introduces an error of its own.
Students describe a city where buildings represent organelles. AI generates the narrative. Students code a labeled HTML diagram — then test whether every label is biologically accurate.
Students describe a change to a Texas ecosystem and prompt AI to model ripple effects through a food web. They evaluate the response against a scientific source and flag inaccuracies.
Students describe a genetic cross in plain language, ask AI to generate a Punnett square and predict phenotype ratios, then verify manually. Advanced: prompt AI to generate a Python script that does it automatically.
Students describe two substances combining, ask AI to predict the reaction type, cross-check with a simulation tool, and write a lab-style report on whether AI was right.
Students paste a primary source and prompt AI with historical thinking questions: Who wrote this? Who benefits? What's missing? They write a sourcing analysis using AI output as a starting point only.
Students pick two countries, pull real data (population density, GDP, climate), prompt AI for a comparative analysis, fact-check three claims, then build a corrected infographic.
Students feed AI examples of WWII-era propaganda and modern advertising, ask it to identify persuasion techniques, then write an argument about where persuasion ends and manipulation begins.
Students vibe-code a decision-tree app: users answer questions about rights → app outputs which constitutional amendment applies. Students write the logic; AI generates the code.
Students are assigned to break an AI tool using their content-area knowledge. Their job: find three prompts that get the AI to produce a confident but factually wrong answer on a topic they've studied. They document each failure, explain why the AI was wrong using evidence, and present it to the class. This requires genuine subject-matter knowledge to execute well — expertise beats the tool. No rubric needed. Proving AI wrong is the assessment.
Students build deployable tools, work with real APIs and datasets, and produce portfolio-ready work. The bar: public and defensible. If it isn't deployed, published, or presented to a real audience, it hasn't cleared the standard.
Students write a research paper using AI assistance throughout, then conduct a documented audit of every AI contribution: what was used, changed, rejected, and why. The audit is a second deliverable.
Students identify a genuine problem in their school or community, vibe-code a one-page technical brief, and present it to an actual decision-maker — principal, city council member, or nonprofit director.
Students receive five anonymous text samples — human-written, AI-generated, and hybrid. Using close reading only (no detection tools), they argue in writing which is which and why.
Students select a real dataset (Texas climate, local property taxes, sports stats), write a plain-language spec for a regression model, prompt AI to generate the code, run it, and present findings with appropriate statistical caveats.
Students describe a family of functions, prompt AI to build an interactive Streamlit app with adjustable parameters, and write a mathematical commentary explaining each behavior.
Students vibe-code a compound interest/loan amortization calculator with their own inputs, run three real financial scenarios, and present a recommendation report to a "financial advisor" panel.
Students find three real published statistics used in news or advertising, prompt AI to explain how each could mislead, then write a mathematical rebuttal. Final product: a public-facing fact-check post.
Students load three to five peer-reviewed abstracts into a Claude Project, query it systematically, identify gaps and contradictions across sources, then write a synthesis that goes beyond any single article.
Students pull real NOAA or NASA climate data, prompt AI to generate a Python visualization pipeline, and build a public-facing dashboard showing one specific trend. They write the interpretive narrative themselves.
Students describe a chemical reaction in plain English, prompt AI to generate a stoichiometry calculator, test it with known reactions, and document the chemistry behind each failure it produces.
Students write a spec for a projectile motion or wave interference simulation, prompt AI to generate it as a web app, test it against known physics outcomes, and annotate every discrepancy from reality.
Students receive a fictional patient case, prompt AI for a differential diagnosis, then evaluate it against anatomy/physiology knowledge — identifying where AI reasoning is sound vs. where it would harm a real patient.
Students pick a current Texas legislative issue (TRAIGA, school vouchers, water rights), prompt AI to draft a brief from one stakeholder's perspective, then write a rebuttal from an opposing stakeholder using real legislative language.
Students pick a historical decision point, prompt AI to model three alternate outcomes based on changed variables, then argue in writing which alternate history is most plausible and why.
Students pull real Census or FRED economic data, prompt AI to build a visualization and narrative, fact-check three claims, then publish a corrected data story as a public blog post or infographic.
Students prompt AI to generate an intelligence-style briefing for an assigned world region, cross-check every factual claim against two vetted sources, and publish a corrected, sourced version.
Students write a one-page product spec for a tool solving a real school or community problem, use AI to generate the full codebase, deploy it publicly, and pitch it — including what they changed from the AI's first output and why.
Students attempt to manipulate an AI system into producing outputs it's designed to refuse — using only text. They document every successful exploit, explain the mechanism, and propose one mitigation strategy. Ethical guardrails defined by teacher upfront.
Students collect 50 labeled examples in a content area, prompt AI to generate a Python classification model, run it on new examples, and report accuracy with a confusion matrix.
Students identify a repetitive task a teacher or administrator actually does and build a working automation. Deliverable: working tool + one-page implementation guide written for a non-technical user.
Students are hired (fictionally) as AI consultants by a real local organization — a school district, small business, city department, or nonprofit. Their deliverable is a professional recommendations report: what AI tools the organization should adopt, which tasks AI should not handle, what the risks are, and what staff training would be required. They present to an actual representative of that organization when possible.
This is the only project that integrates every content area simultaneously — research, writing, data, ethics, economics, and communication — and produces something with genuine external value. It also directly mirrors what many of these students will be asked to do in their first professional job within five years.