New free working session: The Sponsored Answer Test.

Four phases, Seed to Harvest, about twenty minutes, done entirely by hand. You take one assignment you already grade, write the confident answer that would pass without the work, then add the three requirements that answer cannot supply.

Nothing is saved and nothing is sent. Everything stays in your browser tab, and you copy the rewritten prompt out at the end.

Seed. Put the real prompt on the table.
Sprout. Write the answer you are worried about.
Grow. Add the three things that answer cannot supply.
Harvest. Name what you can now see in the stack.

https://www.harvestkernel.com/the-sponsored-answer-test/

What is the one assignment in your course that would still be satisfied by a confident answer with no reasoning attached?

New free working session, and it is a faculty one.

THE REREAD TAX AUDIT

Take the ten oldest unanswered messages in your inbox. Oldest, not easiest. For each one, name its shape, write one line saying what it is really asking, then send the reply that also answers the question they did not ask yet.

Seed, name the shape. Sprout, write the question in your own words. Grow, answer the unasked one. Harvest, read your own distribution and see what it is telling you about something upstream of the inbox.

About twenty five minutes. Ten emails answered when you finish, and a named pattern you keep.

It runs entirely in your browser. No account, nothing saved, nothing transmitted, so copy your work out before you close the tab.

https://www.harvestkernel.com/the-reread-tax-audit/

Post your shape distribution here when you are done. One line is plenty. Five inboxes side by side is the fastest way to find out whether yours is normal.

New working session: The Ghost Check, a one source provenance audit.

Bring one source. A reading from next week's syllabus, or a citation from a paper you are grading. Fifteen minutes, seven fields, four SeedStacking phases. You finish holding a written verdict you can paste into your notes or your feedback.

Nothing is generated for you. Nothing is saved. Nothing is transmitted. The judgement in it is the part that is yours, which is exactly why it is worth the fifteen minutes.

SEED the source as cited. SPROUT the venue and the person. GROW the timeline and the conversation. HARVEST the artifact and your verdict.

Start here: https://www.harvestkernel.com/the-ghost-check-source-provenance-audit/

And the question it leaves you with: What is the last source you assigned that you have never personally opened the full text of?

New free working session: The Missing Row, a one assignment rubric audit.

This is the Grow phase, and it is a manual method. Nothing is generated for you. Nothing is saved, nothing is transmitted, and there is no account. Seven fields, about twenty minutes, one real assignment.

SEED. Name the assignment and paste the rubric rows in the words they are written in today.

SPROUT. Go row by row. Which rows could a well organized, entirely conventional answer earn full marks on? Then which rows require something only this particular student could have supplied? If the honest answer to the second one is none, that is the finding.

GROW. Name the obvious approach you would rather they reject and argue against, and the assumption their recommendation depends on.

HARVEST. Write the missing row. One sentence, observable from the outside, anchored to something the student has to supply themselves.

You finish holding a criterion in your own language, ready to paste into the rubric you already use.

https://www.harvestkernel.com/the-missing-row-rubric-audit/

The research behind it, from a randomized trial published yesterday: https://www.harvestkernel.com/rubric-rewards-polish-openai-bocconi-study/

One question for the room, and a number is a complete answer. How many rows in the rubric you are actually using this term could a well organized but entirely conventional answer earn full marks on?

New free working session: Draw Your Line, The AI Policy Decoder.

https://www.harvestkernel.com/draw-your-line-ai-policy-decoder/

This is not a checklist about AI policy. It is a working session you do by hand, on one real syllabus, in one sitting. Nothing is saved, nothing is sent, and it needs no account.

Four phases, mapped to SeedStacking.

SEED. Pick the class you are least sure about and paste its AI paragraph in exactly as written. Then name the phrases that actually draw the line, because unauthorized use and substantially generated by are two different rules and you can be fine under one and in trouble under the other.

SPROUT. Sort your real habits, the ones from last week, into three piles. Allowed. Ask first. Never.

GROW. Write your line in one sentence you could say to the professor's face.

HARVEST. Write the email about the one habit you are least sure about, and send it tonight.

You finish holding a written line for that class and an email ready to go.

For instructors reading this: this is the artifact to hand a student in week one, and it is the student side of yesterday's piece on instructor disclosure.

How many of your classes this term have told you, in writing, what you ARE allowed to do with AI, rather than only what you are not?

New working session, free, no account, nothing saved.

The Instructor Disclosure Line takes one assignment you are actually teaching this term and gets both halves of its AI policy onto the page in your own words. The student line, prohibited or allowed or required and why in your field's terms. Then your line, where you use AI in producing the course and where you use it in evaluation and feedback.

Six boxes, four SeedStacking phases, about ten minutes. It is entirely manual and it works. You finish holding text you can paste straight into a syllabus.

The last box is the test: write the sentence you would say out loud to a student who asks you in week six. If your half is harder to say than theirs, that is the sentence to fix rather than the one to delete.

https://www.harvestkernel.com/the-instructor-disclosure-line/

Open the syllabus for a course you are teaching right now. Does it say anything at all about where you use AI? One word is a complete answer: yes or no.

New resource: The Two Version Announcement Test.

Seed, Sprout, Grow, Harvest, six fields, about six minutes. You finish holding a rewritten first week announcement you can paste into your LMS today and a test you can run on two sections this week.

Seed. Name the one thing a student has to do next. One action, not five.
Sprout. Cut the middle, not the warm opening. Everything you cut already lives in the syllabus.
Grow. Write the human line in your voice, out loud first.
Harvest. Read it on a phone. If the one thing and the human line are both visible before you scroll, you are done.

Nothing is saved and nothing is transmitted. Copy your work out before you close the tab.

https://www.harvestkernel.com/the-two-version-announcement-test/

Open your welcome announcement on your phone right now. Before you scroll, can you see the one thing a student has to do next? One word answer is fine: yes or no.

New free working session. The Three Column AI Rule.

You bring one real assignment from one real course. You leave holding a paste ready policy paragraph that says something specific and different about generative, assistive and agentic AI on that assignment.

Seed, name the component and what it measures. Sprout, the generative rule. Grow, the assistive rule and the agentic rule. Harvest, the one disclosure line students submit with the work.

About fifteen minutes, done by hand. No account, nothing saved, nothing sent anywhere. Copy your paragraph out before you close the tab.

Open it here: https://www.harvestkernel.com/the-three-column-ai-rule/

Open your syllabus and read your AI paragraph. Which of the three does it actually name: generative, assistive, or agentic?

New resource. The Syllabus Extraction Test.

Seed, Sprout, Grow, Harvest. About twenty minutes, done by hand, on your own syllabus.

You pull every dated item out of your syllabus into one list. Then you list everything that matters and carries no date, the policies and the reasons and the sequence arguments. Then you name the single undated thing that would hurt most if a student never met it, and you attach it to a dated item.

You finish holding a written extraction report with two lists and one decision, plus the share of your syllabus that actually survives a machine reading.

Nothing is saved, nothing is transmitted, no account. Copy your report out before you close the tab.

https://www.harvestkernel.com/the-syllabus-extraction-test/

Open your syllabus. How many things in it matter and carry no date? One number.

New free working session, and this one takes about twenty five minutes.

THE COLLISION MAP
https://www.harvestkernel.com/the-collision-map/

What it does. You pull every dated item out of every syllabus onto one page. It sorts them into the week each one lands in, Monday to Sunday, and marks any week carrying three or more. You finish holding your whole term at a glance and one decision already made.

What it does not do. Generate anything. Save anything. Ask for an account. Everything stays in your browser tab, so copy your map out before you close it.

The four steps run on SeedStacking:

Seed. Name the courses in your own words, not the catalog description.
Sprout. Every dated thing, one per line. A date, a vertical bar, then what it is. Do not filter. The thing you filter out in August is what turns a two into a three.
Grow. Press the button and look. This is the part that changes your face.
Harvest. Pick one item out of your worst week and write the date you start it. A date, not the word early.

Built for students, and it works exactly as well for a faculty member stacking four preps against a grading calendar.

If a syllabus says week seven instead of a date, go to the academic calendar and turn it into a date. A week number you have to translate every time is a date you will never look at again.