Educator Edition · Available now

An AI workspace organized around the way you teach

Choose a task, describe what you need, and review a formatted draft. Every tool below runs on your own computer through the dashboard on the drive.

Create

Create teaching materials

Each workflow asks for the details that shape a useful draft, such as learner level, goals, point totals, or duration.

Lesson plans

Measurable objectives, a timed sequence that fits your class length, guided practice, a check for understanding, and accessible participation options.

Assignments

Purpose, outcomes, student instructions, deliverables, and assessment criteria, matched to the AI-use approach you choose for the assignment.

Rubrics

Analytic rubrics with criteria, point allocations that add up to your total, and observable descriptions for each performance level.

Practice questions

Multiple-choice, short-answer, or mixed questions at your learners’ level, with the answer key and explanations kept separate.

Adapted readings

A clearer version of a passage for a specific audience, with key terms defined and three comprehension questions. The meaning and essential qualifications stay intact.

Messages and general drafting

Announcements, reminders, and professional emails using the dates and facts you supply. The dashboard never sends anything. Ask your AI handles open-ended questions.

See a real draft

The screenshot shows the rubric workflow on the development computer. The local model produced this draft from the request on the left. Like every draft, it is a starting point: an educator still checks the criteria, the point values, and the performance levels before using it.

The Build a rubric request form beside a generated rubric draft with a criteria table.
A rubric drafted by the local model in 9 seconds on the development computer. The sample course and assignment are fictional.

New in update 0.9

Models, pictures, and illustrations

Choose the AI model at the top of the dashboard, add a picture or recording to a request, and make illustrations for your materials, all on your computer.

Three local models

Qwen3.5 4B is the recommended default for writing and reasoning and can read pictures. Gemma 4 E4B also listens to WAV or MP3 recordings up to two minutes. The original Qwen3 4B stays available as a known-good fallback.

Pictures and recordings

In Ask your AI and the writing workflows, attach one picture or recording to a request. It stays in browser memory and goes only to the AI on your computer.

Illustrations and Image Studio

Create an illustration makes classroom images in several styles and shapes. Image Studio, at 127.0.0.1:8081, adds variations, cancel, and a session gallery. Image creation needs an NVIDIA graphics card and at least 12 GB of memory.

Review every image

The first image after startup can take a few minutes; later ones take about 40 seconds on the development computer. Short titles usually render well, but small labels are often misspelled, so check each picture before students see it.

The longer 16,384-token context also lets requests and papers run about four times longer than before.

Image Studio with a picture description, style and shape choices, and a finished illustration.
Image Studio creating an illustration on the development computer. The image engine runs at 127.0.0.1:8081.

Added in update 0.8.1

Design with ARAD 2.0

Does your assignment still measure what it claims to measure when a student uses AI? ARAD 2.0 (AI-Responsive Assignment Design for Human Learning), a framework by Evolve AI Institute founder Tim Mousel, helps you answer that question and redesign the assignment when the answer is no.

A workflow built on the GOALS steps

  1. Goal: name the human capability students must demonstrate.
  2. Openness: choose how much AI the task allows, or ask for a recommendation.
  3. Architecture: build formative and summative phases around one pattern: AI Sandwich, Verification Challenge, Comparative Analysis, or Generative Challenge.
  4. Look for evidence: get a rubric drawn from ARAD’s eight dimensions, plus a Transfer Gap Check.
  5. Sustain: plan a short AI literacy lesson, a pilot, and a faculty pre-test of the assignment.

The Design an assignment workflow offers the same four ARAD modes. As with every draft, you review the design before students see it.

The Design with ARAD 2.0 form beside a generated design that begins with Goal, Openness, and Architecture sections.
An ARAD 2.0 design drafted by the local model in 15 seconds on the development computer (version 0.8.1). The sample course is fictional.

Mode 1: AI-Restricted

AI is excluded because the task measures independent performance, such as a timed writing or an oral presentation.

Mode 2: AI-Limited

Students do the core thinking. AI handles bounded support such as grammar, formatting, brainstorming, or translation.

Mode 3: AI-Collaborative

AI is an open thinking partner, with full transparency, structured verification, and a rubric that rewards the student’s judgment.

Mode 4: AI-Required

Skilled human-AI collaboration is the learning target, with authentic scenarios, verification, error identification, and an oral defense.

Packets

Build coordinated teaching packets

A packet builds several related documents from one request, one component at a time. Later components receive the earlier ones so the materials stay consistent with each other.

  • Assignment packet: student instructions, a shared grading rubric, a student submission checklist, and an instructor implementation guide.
  • Assessment packet: student questions, an instructor answer key with explanations, and an instructor review of both.
  • Lesson packet: an instructor lesson plan, a student activity, a student reference handout, a student exit ticket, and an instructor exit-ticket key.

Each packet exports as a student bundle and an instructor bundle. The dedicated bundle export controls leave instructor-only components, such as answer keys, out of the student bundle.

Documents

Work with your documents

  • Import TXT, DOCX, and PDF files. Reading happens locally with parsers bundled on the drive.
  • Turn on optional local English OCR (optical character recognition) for scanned PDF pages.
  • Search the words in your imported documents.
  • Select relevant excerpts, PDF pages, or a passage to include in an AI request. Only the text you select goes to the local model.
  • Review source excerpts beside the draft. References such as [S1] open the exact excerpt that was supplied.

The session library holds up to 12 documents. Each file can be up to 20 MB, and OCR handles up to 10 scanned pages per file.

Courses

Organize course work

Course profiles let the AI know your context once, so you can stop retyping it.

Course profiles

Store a course description, learning outcomes, student level, class format, schedule, and your preferences. They accompany future requests for that course, and you can switch them off for any request.

Kept drafts and request templates

Keep a finished draft or packet under its course, and save a request as a template to reuse its fields later.

Optional workspace saving

The dashboard starts in temporary mode and saves nothing automatically. When you choose, it saves your courses to a workspace file in a folder you pick, such as one on the drive.

Optional encrypted backups

Protect saved workspace files and downloaded backups with a passphrase. There is no passphrase recovery, so keep a secure record of it.

Export

Improve and export

  • Formatted output by default. Drafts appear with headings, lists, and tables, ready to read.
  • Editable drafts in the supported creation workflows, with Simplify, Add examples, Shorten, and custom revision requests. Undo keeps up to five earlier versions in the session.
  • Word export produces genuine .docx files with editable headings, lists, and tables.
  • Print / Save PDF opens a clean document in your browser’s print dialog.
  • Layout options for document style, text size, line spacing, course and author lines, and page numbers.

Assignment and rubric grading

Proposed rubric scores, with you in charge of every grade

Upload an assignment, your rubric, and up to 30 student papers. The local model proposes a score, feedback, a suggested improvement, and supporting quotations for each criterion. You review each paper, adjust anything you disagree with, and approve the grade yourself.

  • Each request contains the assignment, one confirmed criterion, and one paper. Other papers stay out of it.
  • A proposed score is accepted only when its quoted evidence appears in the paper.
  • Only grades you explicitly approve go into the CSV export.
See the seven-step grading workflow
Instructor review form for paper P001 with editable scores, feedback, and an approval checkbox.
Step 6: every proposed score, comment, and improvement stays editable until you approve the paper.

Included at no extra charge

Everyday Assistant

Not every question is a teaching task. Choose Everyday Assistant at the bottom of the dashboard sidebar and a general-purpose chat opens in its own tab, using the model that is already loaded. Your Educator work stays where you left it.

  • Multi-turn conversations with Brief, Balanced, or Detailed replies
  • Stop, retry, continue a long reply, or edit your last question
  • A chat memory meter, so you know when to start a fresh conversation
  • Word, text, and PDF export, plus saved chat backups you can reopen

Chats live only in that browser tab and go only to the AI on your computer. Everyday Assistant is included with every Evolve AI edition.

Everyday Assistant open in a browser tab, with a two-message conversation and a formatted reply.
Everyday Assistant, included with the Educator Edition, answering on the development computer. The conversation is fictional.

Try the Educator Edition in your own teaching

The Educator Edition is available now on three drives, starting at the Pocket option. Choose your drive, or check that your Windows computer can run it first.