Show your work

So, I don't endorse this as prompt design. But it does provide my answer to Ezra Klien's question to Ethan, around the difference between "the button" and the blank page.

I honestly was trying to ask Claude about the specific factual question - what are the relevant statues and legal frameworks that I would need to ask, and how should I evaluate the (sure to be) dramatically different positions from my school, my union, and my tech friends.

Instead, knowing that I needed to keep the text AFTER the hypothetical clean, I had a natural filter that pulled chunks from the bottom of the stack and made me reintegrate them into a (I think) clearer description of my big "AI Teacher" concerns.

Because, all love to Tipitat, fuck the VR Second Life content generation train. This is the part of the curve where text is SO MUCH BETTER than the immersive scale visual, and SO MUCH easier for kids to use.

Hey Claude. I'm looking for a knowledgeable friend to talk through a thorny, potential situation. I don't expect you to have all the answers, but I also don't need you to explain the context to me. I am consulting you becuase even if you don't know the specific language of a contract or a GDPR clause, you have enough domain knowledge for intellectual property, privacy and employment law to help me look up the specific information, even in Castellano or Catalan.

As a part of my professional practice, I share student work for a professional development purpose with a curated but public audience branded as a piece of your professional public identity. I am familiar with my understanding of student privacy, as it relates to student work, from NAIS Folk Practice and many international schools, empirically and by 2nd Meno proxy. But at my school, that is seen as dangerous. But we have a school instagam and sports pages, and there's a knot of language around this idea in several different "must sign this" documents.

But I have student work, and I want to talk about it. This is what I did as a math teacher and I got better. It's what I did as a edtech person and launched an acquired startup at a StartupWeekend with that name (neé ThreeRing). I kept bogging, got better and launched Makerspaces.

It's how I grow.

The criteria we used to assess a share was about how specifically it identified the child to a anonymized audience. This was also in a MUCH different data and data privacy environment. The idea of linking between a teacher's public page to their school web page was about as deep as people got.

We used langauge like "the kiddo." Math was easier to share, becuase kid handwriting is a wonderful mid layer of tangible authenticity and functional anonymimity. Or we would be web log diarists and condense and restructure the stories for publication.

The later function is when you are looking to "lead a conversation" about something. That guy is climbing a ladder. All love to ladders! MPershan is an amazing mathematician and his blog demonstrated that!

The former is how craft advances, how we "notice we are confused" and try to apply some practice. And it's what I desperately need, if I'm not going to take 6 months off and just do some full PD-retreat on baja

I've been trying to figure out what the "kiddo" technique could be for what I'm fascinated by, which is kid/LLM conversations about test questions. I am legit trying to build a conversational test delivery LLM bot. Something that can run through ... like, shit the system local IRC. Hahah

HHHAHAhahahahah

oh, shit. Ok. So.... possibly villanous. Need to check with Nathan on TWO counts.

Here's what I'm excited about.

I have students writing, in the form of a transcript between their ChatCraft front end to my APIKey running to (school chosen service¹ LLM frontend.

The bot's system prompt (availiable on request) pushees Socratic reflective questioning, providing function defs with docstrings, a fixed comment style and a HARD LIMIT on how much code to produce as a function of how much code is input.

It's not perfect, but I am getting "show your work" from kids in a way that makes SO MUCH SENSE to me. And the ones who struggle are NOT asking the same kinds of questions.

I feel like I've discovered a new kind of math homework. As someone who read dy/dan when we were both teaching in the Santa Cruz COE, and who had Kernodle and Molly as a cube buddies for a year, and who is a personal and enduring (and I hope in some way endearing) source of agitation for Cynthia Solomon and Slyvia Martinez,

I have discovered a LOT of great new math homework over the last 20 years, even without all of MakerEd. Like Tom Scott, I know the feeling of this curvature.

I want to post and share and discuss "kiddo" anonymized versions of this material. I want to talk about if and how these chats can help us udnerstand students on the PERSONAL and INDIVIDUAL scale, not as a statstical model. (Although, yes, the other shoe that would drop from this is SOMEONE is making a statistical model out of this. We have to choose who and how). I want to expand a community of craft around understanding how we talk to each other, specifically by looking at how MY KIDS, who's internal model I am modelling, and trying ti update as directly as possible.

So while I am entering into this looking for a way to protect kid identity, I don't want to summarize, diarize and condense. I want to look at langauge, directly, as a proxy for understanding. Becuase when I work with kids face to face, we look at language. "What did you mean when you asked this?"

So I consider this a classic Test Setting case, and I want to be able to talk to SOMEONE about what that would look like.

So consider this hypothetical.

At the begining of each assignment, I put a data field where students have to AFFERMITIVLEY, for EACH ASSIGNMENT,

provide the following

(I am ok with Mr Carle sharing my work with teacher nerds and robots.

I [real name] would like to be called [psuedonm] with [pronouns].}

If they fill it out, I use it performing programatic swaps betwen their identifiers andthe psuedonym between publication. That lets me keep my personal, meaningful PD task tied to THESE KIDS and their world models, but provides the specificity to start a public conversation.

How would I even know if that was possible? Claude, tell me where I should look to find out how to find out!

I've decided to try Linked-In. I don't understand the vernacular, so I'm letting various LLMs chunk and rewrite them until it's something I can stomach. I'm trying to figure out what #visionary looks like now, and if they are all bullshit. So... uh, watch out for the next two weeks as this ...thing... gets turned into "provocative questions facing passionate educators embracing the frontier of AI scholarship. #edtech #blessed #loveToAudreyWattersImstuckinacontentfarm"