If you write many workflows, and documentation, and ask AI to help you output it, the kind of writing that AI spills out looks terrible. At this point in time, no matter what thoughts you put in, people can just see the structure and label it as AI generated or AI slop.

After reading thousands and thousands of AI generated things, I developed a simple conversion md file, it can be used as a skill, or in a project, and that skill can convert the AI generated output into functional instructions.

Look at this example:

I wrote ‘how to explain agents to teachers with low digital fluency’:

After the AI-slop convertor:

What you will make

Create a project named Documentation Converter.

When you add a draft, the project will:

• preserve the point, facts, links, examples, and your voice;

• remove AI filler, hype, vague language, and repeated conclusions;

• organise information in the order a reader needs it;

• flag information that needs checking instead of inventing an answer.

The project uses the plain, direct style found in good documentation, often used in Claude Docs, Google Docs etc.

What you need

• A ChatGPT project

• The file convert-AI-style.md

• An AI generated draft to convert

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 130 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Read more

View a simple poster it created here https://canva.link/bikd3i8vhpo2es7

TL;DR

Connect Canva to Gemini. Use @Canva with a clear poster brief. Check the first version. Request one edit at a time. Open the finished design in Canva to edit or download it.

What you need

• Google Gemini Pro web or mobile app

• Personal Google Account

• Gemini Apps Activity turned on

• Connected Canva account (I have a pro account)

Do not include student names, identifiable student work, or confidential school information.

Check the connection requirements

Google states that Canva Connected Apps require a personal Google Account and Gemini Apps Activity. Canva in Gemini is available in English on Gemini web and mobile, not in Gems or Gemini Live. Gemini can create and edit Canva designs, search designs and folders, move assets, and manage comments. Google Gemini Apps Help

Gemini cannot export Canva designs in different formats. Open the completed design in Canva to download a PDF, PNG, JPG, or PPTX. Google Gemini Apps Help

Connect Canva

1. Open Gemini Spark.

2. Click on Connected Apps

3. Go to other

4. Add the Canva App

Click on new task and paste this prompt:

@Canva

Create an editable classroom poster in portrait 2:3 format.

Purpose:
Help [AGE GROUP] students understand this learning goal:
“[PASTE LEARNING GOAL]”

Poster headline:
“[SHORT STUDENT-FRIENDLY HEADLINE]”

Include exactly:
- the learning goal in student-friendly language
- three observable actions students can take
- one short self-check question
- one simple visual metaphor that supports the idea

Visual direction:
- clean editorial classroom design
- high contrast
- easy to read from across a classroom
- white or very light background
- one accent colour: [COLOUR]
- bold sans-serif heading
- generous spacing
- no more than 35 words of body copy
- age-appropriate imagery only

Hard constraints:
- Do not add claims, standards, names, dates, or instructions I did not provide.
- Do not use tiny text.
- Do not add logos, watermarks, decorative clutter, or duplicate wording.
- Keep every element editable in Canva.

Before creating, show the exact text that will appear on the poster. Flag any assumption. Then create the design.

Phil

Here is how the output looks: https://docs.google.com/document/d/1WePFZsmkDnHgWd6eoOg_6dMUBk7-21weTpXkXld443s/edit?usp=sharing

TL;DR

Paste an assessment, lesson plan or project brief into ChatGPT Work.

Ask it to rebuild the task through the BSCS 5E sequence:

1. Engage

2. Explore

3. Explain

4. Elaborate

5. Evaluate

For each phase, create:

• Student action

• Individual evidence

• Learning inference

• Teacher decision

Use the map before grading a group product, presentation or final submission.

The quality bar

Use the National Academies’ Knowing What Students Know assessment framework.

For each phase, ask:

What does this student response let me infer about the learning goal?

Do not use completion, participation, attendance, confidence or a polished group product as the evidence point.

Use Engage, Explore, Explain, Elaborate and Evaluate as one sequence. Do not add five unrelated activities.

Read: Knowing What Students Know and BSCS 5E Instructional Model.

What you need

• ChatGPT Work

• An existing assessment, lesson plan, task sheet or project brief

• Learning goal

• Age group and subject

• Available time

• Curriculum requirements

• Marking or reporting requirements

Remove student names, work samples and identifying information before uploading.

Treat the output as a draft. Check curriculum alignment, accessibility, local context and assessment decisions yourself.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 130 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Read more

Personalised learning missions seem like the future of learning, it’s quite incredible how I can just work out some simple things, a learning sequence, some instructional design methods, and I have a personalised workbook sent to me at 4am every day. No excuse not to learn Spanish anymore, and I have it set in a way that will engage me.

The finished doc is - https://docs.google.com/document/d/1cRPPgD6tFHNKTMfen3MEyBJLk8HTHTrJUt11iwWtC7w/edit?usp=sharing

The free workflow

Open Gemini Spark and paste:

Interview me to design a 30-day Mexican Spanish learning mission.

I want practical Spanish for my real life in Mexico, not a generic course. I am a beginner and prefer 5–10 minute sessions, minimal reading, speaking aloud, word searches, simple games, food and daily-life vocabulary.

Ask me one question at a time about:
- my level and confidence
- the people and places I speak with
- my interests, routines and food
- situations I want to handle independently
- games and workbook formats I enjoy

Then create a 30-day sequence of daily learning missions. Each mission must include:
- one real-world Mexico Spanish purpose
- a tiny, observable language goal
- a short speaking task
- a piece of evidence I can produce
- a retrieval activity that revisits earlier learning
- one fun game or word search on selected days

Make the missions build towards independent real-life conversations.

Answer the questions. Save the 30-day mission plan.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 130 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Read more

The final document looks like this - https://docs.google.com/document/d/1EJRDGVNoMkzn_PCMqPWBR5ArRvZPLiWaMI3VCGjMQuQ/edit?usp=sharing

Milestone-based assessment assesses learning at specific points during a project instead of relying only on the finished product.

Students produce evidence at each milestone, such as a proposal, plan, prototype, draft, explanation, revision, or final presentation.

Key features

• Milestones: Set specific points in the project where students submit evidence of learning.

• Evidence: Decide what students will produce, explain, demonstrate, or show at each milestone.

• Feedback: Give feedback at selected milestones before students continue or revise.

• Revision: Ask students to show how they responded to feedback.

• Final evidence: Include an individual task that shows what each student can independently understand or do.

This paid article will show how to turn a vague project idea into a structured project with milestones.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 130 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Read more

Put ten sheets of paper and one roll of sticky tape on each table.

Give students this mission:

Build the tallest freestanding tower you can.

Add five checkpoints:

1. Individual design prediction

2. Team design decision

3. First-build test record

4. Planned revision

5. Individual final explanation

Use final tower height as the challenge result. Assess the evidence from each checkpoint. This will help work out product-based and process-based assessment.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 130 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

The original activity

Working in a team, use ten sheets of paper and one roll of sticky tape to build the tallest freestanding tower possible.

New Learning goal

Students will design, test, and improve a paper structure. They will use evidence from a test to explain one structural change.

Student mission

Your engineering team has ten sheets of paper and one roll of sticky tape.

Build the tallest freestanding tower you can. It must stand without assistance for ten seconds.

Complete each design, test, and revision milestone checkpoint during the challenge.

Materials and rules

• Ten sheets of paper per team

• One roll of sticky tape per team

• Scissors, if permitted

• Ruler or tape measure

• Timer

• Students may fold, roll, cut, or tear the paper.

• Tape may join paper. Do not attach the tower to the floor or furniture.

• Do not use the tape roll, dispenser, or cardboard core as part of the tower.

• The tower must stand without being touched.

• Give every team the same building time.

Checkpoint 1: Individual design

Before discussion, give every student three minutes to sketch a possible tower.

Under the sketch, students complete:

I predict this design will remain stable because…

Collect or photograph the sketches.

Checkpoint 2: Team decision

Students compare their sketches. They choose one design or combine ideas.

The team records:

We selected this design because…

Each student labels one structural feature on the team plan:

• base width

• column shape

• support position

• weight distribution

Checkpoint 3: Build and test version one

Give teams a first build period.

Measure each tower. Test whether it stands for ten seconds.

Students record:

• height

• pass or fail on the stability test

• where the structure bent, leaned, twisted, or collapsed

• one observation from the test

Pause the build. Each team completes:

We expected…
We observed…
We think this happened because…

Checkpoint 4: Revision

Before students touch the materials again, each team records one planned change:

We will change ___ because our first test showed ___.

Students build version two. Measure it and repeat the ten-second stability test.

Checkpoint 5: Individual explanation

Give each student a short individual response:

What was the most important change your team made? Use evidence from one test to explain whether the change improved the structure.

Students may write the response, record a 60-second explanation, or annotate before-and-after photographs.

Assess the evidence

What I used

• Tool: ChatGPT Work

• Starting task: Build the tallest freestanding tower using ten sheets of paper and one roll of sticky tape.

• Learners: Upper-primary or lower-secondary students

• Time: 30–40 minutes

• Output: Learning goal, five checkpoints, and three assessment criteria

• Source material: An existing classroom activity and an optional curriculum standard

Prompt to to help design this - Use with Chatgpt Work/Claude Cowork/Gemini Spark for agentic passes:

If no curriculum standard is supplied, label the learning goal as an instructional assumption.

Try this with

• Build a straw bridge that holds the greatest mass.

• Create a container that protects an egg during a one-metre drop.

• Design a paper aeroplane that travels accurately.

• Construct an aluminium-foil boat that carries the most counters.

• Build a marble run that keeps a marble moving for exactly ten seconds.

Set the learning goal first. Then select the individual evidence students will produce during the task. Will you break it up and force time pressure, or will you simplify it and build?

Phil

ChatGPT Voice is available in Work and Codex desktop tasks and I have started using it over Wispr Flow. My best use is rough capture.

Go for a walk and talk through an article, video idea, or research question without holding. Start by telling it what you are working on, what you already think, and what you want to find out. Then talk for ten or fifteen minutes.

For example: “I’m going to talk through what I know about game-based learning and narrative. I want to identify the gaps in my thinking, then find out whether other people have written about these ideas, including in journal articles.”

When you finish, ask it to turn the recording into an outline, shot list, or research brief.

Paid members will receive the prompt I am using at the moment.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 100 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Read more

A useful YouTube video is not automatically a lesson.

Take a video showing how to make a small pot plant from recycled materials. It could support a lesson about following a process, choosing materials for a purpose, testing a design, or explaining what plants need to grow.

Paste the video link or transcript into Gemini Spark with the age group or class context. Ask it to identify possible learning goals, flag what the video does not explain, and help you turn one goal into a short task with individual evidence.

You need a link or transcript, learner context, and roughly ten minutes to make the first version.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 100 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Read more

I think the main issue with AI generated lesson slop is that it just adds things that don’t need to be added.

AI can tell us - sure, put in your class details and code and it will make something great.

When I write a lesson with AI, I just start with my own flow and transition, I put everything into segments, and then use AI in some parts.

Here is an example: a 60-minute Grade 5 science lesson in an Australian school garden. Students observe small animals, then explain how one feature or behaviour might help an animal survive there.

What you need

• ChatGPT, Claude, Gemini, or whatever you use

• A school garden, planter boxes, oval edge, native garden, or even photos from around the school

• Clipboards or books and pencils

• Magnifiers if you have them

• Your usual outdoor-learning rules

In Australia, they are obsessed with Standards, so it’s always good to start there:

Australian Curriculum v9 - Year 5 Science

AC9S5U01
Examine how particular structural features and behaviours of living things enable their survival in specific habitats.

For this lesson, students are observing a garden animal, identifying one feature or behaviour, then using evidence to explain how it may help that animal survive in that habitat.

The learning outcome is simple:

How does a small animal survive in our school garden?

My hope is that each student finishes with:

• one observation, sketch, or evidence photo

• at least one feature or behaviour they noticed

• an explanation using “because”

• one thing they are still unsure about and any deeper questions

Map your lesson before you ask AI for anything

Some teachers plan by minutes. I tend to plan around transitions, because that is where a class either holds together or starts to go weird. I used to rate the energy in each task, so I could adapt the next. As simple as, okay it was sports day yesterday, the first few transitions can be based on mindfulness and getting the mind ready.

What I think the full lesson looks like in my classroom

Entry — 5 minutes

On the board/display - always the learning goal, and if there is a driving question:

How does a small animal survive in our garden?

Students come in in an expected way, we discuss safety and organisation, they might grab a clipboard, and draw or write one animal they think they might find outside. Ant. Spider. Snail. Beetle. Ladybird. Whatever.

I’m checking their starting point and trying to get everyone doing something within the first 5 minutes.

Announcements - 3 minutes

• We observe animals. We don’t touch them.

• Stay with your partner.

• Stay in the marked area.

• If you find something good, call me over before you disturb it.

• A photo, quick sketch, or written note all count as evidence.

Launch - 7 minutes

I show two close-up photos. Maybe a native bee on a flower and a small spider in a web. I normally take photos of real things to show hey look what I just caught, and that may have taken 3 days to find a good picture!

Then:

What can you see that might help this animal survive here?

I will take a few ideas for my record. Wings. Colour. Legs. Web. Somewhere to hide. Food nearby.

Then I say:

You are not trying to name every bug today. You are trying to collect evidence for one survival idea.

The focus is on survival, not running around looking for animals and being crazy.

Transition to work - 5 minutes

Each student gets this field-note frame:

What I observed

What feature or behaviour did I notice?

How could it help survival?

What makes me think that?

Partners can search together. Each person records their own observation and explanation.

First work block - 15 minutes

Students go out and look.

I circulate and ask things like:

• What can you actually see?

• Where is it?

• What is it doing?

• What part of it might help it get food, move, hide, or avoid danger?

• What would make that explanation less of a guess?

A student might write / scribble:

I saw ants walking in a line beside the garden bed. They were touching things with their antennae. I think the antennae may help them follow a path or communicate, which could help them find food.

Midway stop - 5 minutes

Bring them back briefly.

Show two anonymous examples:

I saw a spider. It has eight legs.

I saw a spider sitting in a web between two leaves. The web may help it catch insects for food.

Ask:

Which one gives us evidence about survival? What is missing from the first one?

Then send them back out.

Usually they get it after that.

If behaviour is off I can reset or remove.

Final work block - 12 minutes

Students improve one field-note page and write:

The __________ helps the __________ survive in the school garden because __________. My evidence is __________.

For students ready for more:

What might happen to this animal if the garden became hotter, drier, or had fewer plants?

Pack-up and exit — 8 minutes

They submit the field-note page to the Google Classroom Daily req.

Exit ticket:

One piece of evidence I used was…

One thing I am still unsure about is…

My hope is to get some great data - You can see who is observing but not explaining, who mixes up a feature and a behaviour, and who is ready to compare animals or habitats.

The prompt

This is where it gets interesting. All of that above I wrote from my experience, I used AI only for the questions, I could have added better ones, but hey, it’s Friday night, I want to watch some sport. I want to improve that lesson. I have actually done this many times so I am very aware, but if you just write it and are unsure of the parts, you will need to work that out first, in the classroom. It always works better to test something yourself, and then go to AI.

Paste this into your AI tool after you have written your real lesson stages.

Link to the document output - https://docs.google.com/document/d/1Jhoq6YQVjmtjEhV9BzQna4EPg2llHFpo4rVLDuqhLLs/edit?usp=sharing

Help me turn my existing lesson structure into a (Grade 5 Australian Curriculum science) PBL Future Labs learning mission.

Do not replace my teaching style with a generic lesson plan. Keep the stages, routines and transitions I provide.

Context:
- Year level: (Grade 5)
- Topic: (small animals in the school garden and how a feature or behaviour may help survival - i see a lot of caterpillars and snails)
- Duration: (60 minutes)
- Student outcome: (each student produces a field-note page exit ticket and a short evidence-based explanation for the student portfolio)
- Conditions: (outdoor observation only; students do not touch, trap or collect animals; use plain Australian English; do not invent local species or unsupported science claims.)

My real lesson stages:
[paste your stages, timings, routines and transitions here]

Divide the work into agents and work as a small team.

Mallee - lesson mapper
First, make a short planning table with:
- lesson stage
- teacher purpose
- student action
- individual evidence to collect
- risk or misconception to avoid

Do not write the lesson yet. Every stage must do a different job.

Wattle - lesson builder
Use the approved table to write the lesson. Keep my stage sequence.
Include:
- teacher wording
- student instructions
- materials
- transition cues
- outdoor safety routines
- one midway comparison showing the difference between an observation and an explanation
- a simple field-note template

Do not add expensive resources, invented animals, vague “explore” activities, or a final product that hides individual learning.

Kookaburra - critic
Check the draft without rewriting it.
Look for:
- generic AI lesson language
- missing practical routines or transitions
- activity without evidence of thinking
- claims that are too certain
- group work hiding individual understanding
- safety or accessibility problems

Make a table:
Quote | Problem | Why it matters | What would fix it

Banksia - reviser
Use the original brief and critique to write the final lesson and field-note template.

Only fix real problems. Keep my structure and wording where possible.

At the end, list the material changes made and check:
- each student produces evidence
- the science is cautious and age-appropriate
- every stage has a purpose
- the lesson fits 60 minutes

Use your context, your knowldge, and consult expert books to get a good lesson structure.

My lessons are often quite deeper than above, I connect misconceptions in to the transitions, I give students chances to work in pairs, or groups, with different expectations as well.

The more context you give the AI, the better the outcome will be for you, and you will be more aware of what it is giving you back.

Phil

Please contact me if you want some extra support with this!

Free workflow

1. Go to Gemini Connected Apps.

2. Under Custom apps for Spark, add:

https://mcp.notion.com/mcp

3. Complete the Notion sign-in and permission screen. Name the connected app Notion, so you can call it with @Notion.

Example

@Notion find my recent AI workflow ideas I have been developing in <page link>

List five that could become workflows, and explain why.

For each one, give me:
Platform
Basic prompt that I can adjust.
Hopeful Output
Out of the box ideas that can extend this prompt.

Do not edit, create, move, or delete anything in Notion.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 100 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Phil

Imagine you are a Grade 8 health teacher teaching students about vaping. My wife is a doctor, and I have heard some crazy stories about the effects of this. Tongue cancers, mouth cancers, some of the most incredibly horrible experiences anyone can have. And yet, they still sell vapes, and so many influencers puff on them like it’s a cool smoke.

The traditional learning approach might be simple:

• Watch a dramatic video about the risks of vaping.

• Complete a worksheet, maybe write a contract.

• Mark the answers, now the student is aware.

Now AI can make that worksheet in seconds.

The basic AI worksheet prompt

Create a Grade 8 health worksheet about the dangers of vaping.
Include:
- a short reading
- five comprehension questions
- a vocabulary matching activity
- an exit-ticket question

Use clear, age-appropriate language.

That is fine. The video is informative. The worksheet may be useful.

But when a student completes it, what do you actually know?

You may know they found the right facts. You may know they watched the video. You may know they completed the activity.

You do not necessarily know whether they can judge a vaping claim, decide whether a source is trustworthy, or explain why a message could mislead someone.

The more engaging AI ‘Real World’ project prompt

Perhaps you want something less passive.

You want students involved. You want them to create a vaping-awareness campaign for the school or local community.

Maybe this could show the parents that are concerned that the school is ‘doing something about it’.

So you ask AI:

Create a fun Grade 8 project about vaping.

Students will work in groups to create an anti-vaping campaign for our school community.

Give them a project brief, creative options to display their campaign, basic research tasks, a timeline, and a final presentation.

This could be a much more engaging experience.

Students might make stunning posters, videos to share with the community, podcasts, social posts, assemblies, or information displays. They may truly care about the final product. They may work hard.

But again: what does the final campaign prove?

A polished poster can be made by one confident student. A group can divide the work. Students can copy facts. AI can help make the writing sound convincing.

The worksheet captures completion.

The project captures engagement.

Neither automatically captures learning.

What to do instead: build the learning backwards

Start with the skill, not the worksheet or the project.

For this example, the Grade 6–8 National Health Education Standards include:

• 3.8.4: Analyse the validity, reliability and accessibility of health information, products, services and other resources.

• 3.8.5: Use strategies to manage misinformation and disinformation.

So the real learning goal is not:

Students will learn that vaping is bad.

It is closer to:

Students can analyse a health claim about vaping, decide whether the information is reliable, and use trustworthy evidence to create a responsible message for others.

Now the worksheet has a purpose.

Now the campaign has a purpose.

And, most importantly, you can see the thinking that connects them.

Try this prompt

You are helping me redesign a [Grade Level] [Subject/Topic Focus] learning experience using Understanding by Design.

Topic: [Write the topic].

Do not begin by making a worksheet or an open-ended project.

Start with these learning standards:

[Add 1-3 learning standards]

Create a PBL Future Labs Learning Mission learning experience in which students produce a [What is the ideal outcome] for their [Audience].

First, define:

1. A clear student-friendly learning goal.
2. What I would need to see, hear or read from each individual student to trust that they have developed this skill.
3. Three scaffolded milestones that make student thinking visible before the final campaign:
   Example: - students analyse a vaping claim
   - students compare and choose reliable sources
   - students justify and revise their campaign message after feedback
4. A small worksheet or activity for each milestone.
5. A final campaign brief with student choice of format.
6. Two simple rubrics:
   - A group rubric for the final campaign
   - An individual rubric for each student’s visible thinking and decision-making process

Keep the work realistic for one [Year level] teacher.

[Final instructions and constraints]

I used Chatgpt Work 5.6 Terra:

You can read the final product here - https://docs.google.com/document/d/1l4i0YfILVf-k8t87LxStp1eepdHU2jSxIZ6iLLgg79c/edit?usp=sharing

You are helping me redesign a Grade 8 health learning experience using Understanding by Design.

Topic: youth vaping.

Do not begin by making a worksheet or an open-ended project.

Start with these learning standards:

- 3.8.4: Students analyse the validity, reliability and accessibility of health information, products, services and other resources.
- 3.8.5: Students use strategies to manage misinformation and disinformation.

Create a PBL Future Labs Learning Mission learning experience in which students produce a vaping-information campaign for their school or local community.

First, define:

1. A clear student-friendly learning goal.
2. What I would need to see, hear or read from each individual student to trust that they have developed this skill.
3. Three scaffolded milestones that make student thinking visible before the final campaign:
   - students analyse a vaping claim
   - students compare and choose reliable sources
   - students justify and revise their campaign message after feedback
4. A small worksheet or activity for each milestone.
5. A final campaign brief with student choice of format.
6. Two simple rubrics:
   - A group rubric for the final campaign
   - An individual rubric for each student’s visible thinking and decision-making process

Keep the work realistic for one Grade 8 teacher.

Do not use shame-based language. Do not require students to disclose personal experiences with vaping.

Worksheets have a place, engagement has a place, visible thinking is the only way for students to remain accountable along the learning process.

Phil

A recall task asks students to find, name, list, define, or reproduce information. Useful yes, but rarely does it show the full process of learning.

This workflow takes one existing assessment one-pager and rebuilds its core demand. Students must explain a relationship, make and justify a decision, or apply knowledge to a new case. The final product still remains; but the assessment gains three short checkpoints that show individual reasoning.

What I used

• ChatGPT Work

• One existing assessment one-pager

Do not upload student work. The goal is to improve the assessment design before students begin.

My prompts are now agent-based! Please let me know how they work for you! I have been reading a lot of tech documentation, and I am trying to simplify these complex agents into simple agents for education.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 120 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Read more

A report, presentation, video or model can be good evidence. Sometimes it is not enough on its own, and in the age of AI, you want to be sure the final product isn’t something produced automatically.

So start here:

Think about what you would need to see, hear or read from this student to trust that they understand this learning goal?

Let’s call that the Student Evidence Map.

The check behind it comes from Wiggins and McTighe’s Understanding by Design: decide the learning first, decide what would count as evidence second, then build the task.

Schools often begin with the fun project idea, then try to bolt the rubric on at the end. It gets a bit backwards.

What you upload

Use one existing assessment one-pager. Include whatever you have:

• year level and subject

• learning goals or standards

• task instructions

• final product

• current rubric or criteria

• time available

• group or individual conditions

• school requirements you cannot change

You do not need to upload student work. This is a check on the assessment design.

Free prompt: find the evidence gaps

Attach the one-pager and paste this into ChatGPT Work.

Audit the attached assessment one-pager for visible individual student thinking.

Use Understanding by Design as the check:
1. Identify the desired learning results.
2. Decide what would count as acceptable evidence from one student.
3. Check whether the task actually produces that evidence.

Create a concise Student Evidence Map. For every learning goal, show:
- what a teacher would need to see, hear or read from one student
- where the current task collects that evidence
- what is missing or weak
- what could be mistaken for completion, polish, group contribution, parental help, templates or plausible AI output

Then give the three most useful changes. They should make thinking visible without creating constant paperwork.

Do not rewrite the assessment yet. Ask up to three questions only if missing information would change the recommendation.

Read the audit before you change anything. It may show that the task is already fine. Or it may show that one vague learning goal is doing too much.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 120 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Read more

AI has reignited my passion to learn. So, for this example, I want one decent thing to learn, sent to my inbox on a weekday, in a PBL Future Labs Learning Mission style.

Gemini Spark can run tasks on a schedule and work with connected Google services such as Gmail. It is still early software, so I would test the first run before trusting it with anything important. Google says Spark is for personal Google accounts, requires Google AI Pro or Ultra and Keep Activity, and is not currently available through work or school accounts. Google’s Spark guide

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 100 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

What to type into Spark

Create a recurring personal PBL Future Labs Learning Mission task.

Every weekday at [TIME] in my current time zone:

Find one genuinely useful, current thing for me to learn about
[TOPIC OR FIELD].

Turn it into one small learning mission I can complete in 10–20 minutes.

Email the mission to [YOUR EMAIL ADDRESS] using Gmail.

Each email must include:

Subject: Today’s Learning Mission: [short title]

- Why this is worth 20 minutes today
- One clear action I can take immediately
- One reliable source or official link to start with
- One small piece of evidence I should save when I finish
- One optional next step if I want to go deeper

Keep it practical and specific. Do not send generic productivity advice, long summaries, or more than one mission.

Before creating the schedule, show me what will run, when it will run, and whether you need permission to use Gmail. Then create it and run it once now as a test.

Here’s what I tested:

Create a recurring personal PBL Future Labs Learning Mission task.

Every weekday at [5am] in my current time zone:

Find one genuinely useful, current thing for me to learn about
[agentic workflows: focus on Chatgpt Work, Sol, 5.6 or Gemini Spark].

Turn it into one small learning mission I can complete in 10–20 minutes.

Email the mission to [phil@pblfuturelabs.com] using Gmail.

Each email must include:

Subject: Today’s Learning Mission: [short title]

- Why this is worth 20 minutes today
- One clear action I can take immediately
- One reliable source or official link to start with
- One small piece of evidence I should save when I finish
- One optional next step if I want to go deeper

Keep it practical and specific. Do not send generic productivity advice, long summaries, or more than one mission.

Before creating the schedule, show me what will run, when it will run, and whether you need permission to use Gmail. Then create it and run it once now as a test.

Google says Spark tasks can be created from a conversation, and that schedules can run at a certain time or after an event. Spark’s own documentation uses a “Gmail Writing” skill as an example in a workflow, while its Gmail actions include drafting, replying, forwarding and labels. Your account may still ask for confirmation or behave a little differently, so check what it proposes before you approve it. Google’s task documentation

Test it straight away

Open the task’s work panel and look under Schedules. Run it now, then check your inbox.

And this is what I like to read with my morning coffee!

One warning: Spark can use connected services and may ask for permission before taking actions such as sending communications. Read those requests. Google also says Spark is experimental and needs supervision, especially when it is using connected apps or personal information. Google’s safety notes for Spark.

Learn well!

Phil

I helped a friend make this for a Year 5 genre-launch week. Paste the prompt into Gemini Spark and it creates an editable Google Slides deck: ten short slides that make fantasy feel exciting, visual, strange, and worth reading. Gemini in Slides can generate editable presentations, and Nano Banana can create and refine the illustrations. Google Slides help

Paid article.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 100 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Read more

There is something important to learn with consistency and story boards with video generation.

Here’s a free prompt and full paid workflow to get a video like above!

Free workflow

Create a 10-second cinematic close-up of a fictional humanoid robot’s face.

The robot is an original character, not based on any real person. Its face is realistic but clearly synthetic: pale ceramic skin, subtle seams at the jaw and temples, fine metallic details beneath translucent surface panels, expressive dark eyes with a small internal light.

Shot: locked-off close-up, eye level, 85mm portrait lens. The robot’s head and shoulders remain centred. Clean, dimly lit workshop background with soft practical lights far behind it. Shallow depth of field. Natural film grain. Realistic skin and material texture. Quiet, restrained science-fiction realism.

Action: for the first 3 seconds, the robot holds still and looks just past the camera. It then slowly moves its eyes towards the lens, makes one tiny inhale-like movement through the chest and neck, and gives the faintest almost-human change of expression. No head turn. No hand movement. No camera movement. No cuts. No dramatic reaction.

Audio: low room tone, a faint electrical hum, one soft servo sound as the eyes settle. No dialogue. No music.

Constraints: one continuous shot; exactly one character; minimal movement; no text; no subtitles; no logos; no flashing lights; no morphing; no extra limbs; no sudden changes to the face, clothing, background, lighting, or camera angle.

Read more

Alright I admit it, it’s getting too overwhelming keeping up with the latest releases. Most teachers do not need another list of AI releases. They need to know whether one of those releases could solve a real problem in their classroom without creating hours of setup to work out an entirely new platform.

The prompt helps you assess one idea at a time.

It could uncover a sensible use for Gemini Notebook or Claude Code. It could also tell you that this week’s big new feature is vague, unverified, or simply not worth bothering with.

Read more

Read Part One: How to Use Claude to Read Journal Articles as a Non-Specialist

Part One covered how I use Claude to scan a journal article without acting as though a summary has made me an expert.

The second workflow is when you find a good article with actual strategies.

The aim is to take one finding from the paper, work out what it might mean for a particular group of students, and test a small change in the classroom.

For this example, I used Meta-Analysis of the Impact of Reading Interventions for Students in the Primary Grades.

The paper reviews 33 rigorous studies involving students in Grades 1–3 who were at risk of reading difficulties. It separates several reading outcomes, including word reading, fluency and comprehension, which makes it more useful than a broad conclusion that reading interventions either “work” or do not.

Read more

Help me scan this journal article as a non-specialist. Explain it section by section, define unfamiliar terms, and separate what the researchers found from what they think it means.

Hi, the first for this series will be free, and the rest will be under paid. Please consider subscribing and thank you to the new paid subscribers this weekend!

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 100 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

Journal articles are written for people already familiar with a field, often in language that is over the top, fluff driven, and seems to follow strange rules. They often assume you understand the vocabulary, the common methods, and the debates the authors are joining. Claude can help you work through that material, so that you can begin to read it better.

I’m going to show you how I scan/read this:

Start by uploading the PDF and add that prompt above.

Feel free to use a second prompt when you reach the end of the article, I often do this to store my details in a Notion to keep track of these articles, it looks like this:

This is a much shorter version of what I have in my own prompt libary, but you get the point. Build that investigative question.

Phil

Students do not need to build a bot for this either.

You can easily assign this Bot as an Assignment in Boodlebox.

Essentially, students open Choose Your Own Learning Experience, enter a learning goal, something they care about and their grade level. The bot asks a few questions, then turns their answers into five short learning missions.

Please consider joining a yearly membership! I make it super cheap at $50usd for everyone, I already have over 100 incredible workflows, and each sign up allows me to dedicate time for another workflow!

If you pay and you don’t like the content, I will happily refund you!

The simple workflow

Open this public bot: Choose Your Own Learning Experience

Paste this:

I am in Grade [YEAR LEVEL].

I want to learn or get better at: [LEARNING GOAL].

I am interested in: [THEME, BOOK, HOBBY, PLACE, GAME, ANIMAL OR REAL-WORLD PROBLEM].

Please ask me any questions you need. Then create five short learning missions I can complete in about 20 minutes each. Each mission needs a clear goal, materials, 1–3 activities, a reflection, an optional extension and one thing I can show as evidence that I have learnt it.

Simple workflow for students to follow

1. Set the skill and the boundaries. Give students one specific skill or a small choice of skills: summarise a chapter, compare two sources, explain a scientific process, calculate change, or use persuasive evidence. Name the time available, safe resources, collaboration rules and the kind of evidence you need to see.

2. Let students personalise the route. Students open the public bot and enter their grade, goal and theme. The bot should check vague or unobservable goals, ask useful follow-up questions and turn the answers into five 20-minute micro-missions.

3. Approve one mission and collect the proof. Students choose one mission to start. They keep the bot-generated plan, their working notes or draft, and the final evidence named by the mission. End with a two-minute reflection: What changed in your understanding? Which part of the evidence shows it?

The teacher requirements

Do not say: “Ask the bot to make you a project about anything.”

Do say: “Our observable skill [FOR ACTIVITY] is [SKILL]. You can choose a theme that matters to you. The bot will help you make the work manageable, but you must show your plan, your working and what you learnt.”

The goal is to keep agency in the student’s hands without making the learning target disappear.

Good classroom uses

• A reading group turns a shared comprehension skill into different book, animal, sport or history themes.

• A maths class practises the same calculation skill through interests such as cooking, football transfers, travel or game design.

• A research unit lets students choose a local or personal angle while working from the same source-evaluation criteria.

• A student who has missed part of a longer project completes one contained mission and produces evidence they can carry back into the group.

Do not skip these guardrails

• Keep the observable skill teacher-selected and explicit.

• Check the first generated mission before students begin if the task affects assessment.

• Treat bot output as a plan, not evidence of learning.

• Collect process evidence, not just a polished final answer.

• Give a non-AI or teacher-conference route for students to defend their learning.

*This is not sponsored by BoodleBox.

Phil