A week ago, everything shifted. I went from NOT using AI at all, in my daily life (except for brief, as-needed dabbles with MidJourney) to full-on immersion with Claude Projects. The reasons? Threefold, really:
- Claude showed me that it could understand visual metaphors -a VERY advanced capability! Specifically, Claude understood and intelligently interpreted an image (JPG file) that had not only both text and pictorial elements, but those “pictorial elements” were the basis for a visual metaphor! OMG, an AI understanding visual metaphors! (Contrast-and-compare with just a year ago, when Google’s Gemini went woefully woke, and at the same time confused references to one of my favorite Greek goddesses; see the LAST BLOGPOST, “AI Is Tummy-Churning,” on this.
- We just have TOO MUCH STUFF to understand. We need help organizing. Making sense of things – like having a very wide-band intelligent associate who can help us both see the full spectrum of what we present, but also give us a bit of perspective. Finally,
- We need help just getting stuff done. I was VERY concerned on behalf of my students. (That was the initial impetus.) Just … with the amount of material, and the need to GENERATE material – everything from reports to videos to any kind of social media post, from insightful analyses to organizing structured data, and even in creating useful visuals … what we need to be able to generate effectively is so much. We just need help with that. My students need help. I do. We ALL do.
So – not knowing what to expect, having NO positive prior experience, but having heard good things about using Claude Projects to organize … essentially, projects – I downloaded Claude. Opened up a chat. Created my first Claude-Project. And dove right in.
And now, approximately a week later, I’m here with a first report. But also – more usefully for all of us, perhaps, what I’ve learned in just this last week to make these Claude-sessions more productive, satisfying, … overall just more REWARDING – some useful tactics and patterns for working with Claude Projects.
Claude and Your Productivity Boost
First, working with Claude has been an enormous productivity boost. Everyone says this, but yes – after years of resistance, I’m now a devout proponent.
Here’s an example. I was going to submit a proposal to the Northwestern TEACHx, actually on the subject of using AI in pedagogy. (Wasn’t sure what I’d say, but I intended to do a talk anyway …) The due date was Sunday, February 22nd. By Thursday (Feb. 19th), I was very engaged with using Claude. By Friday (Feb. 20th), I had enough experience, we’d framed some ideas for reframing how students could work with AI (specifically, with Claude) for a much more enjoyable and effective learning experience, with micro-mastery loops, and on Friday afternoon, three days before the Sunday proposal deadline, I brought up the whole “proposal” thing.
In less than an hour – based on a few days of me sort of rambling about the challenges that students were facing (from end-of-day exhaustion to the pure, simple TOO MUCH STUFF issue) – we were able to consolidate. I opened up the TEACHx portal, filled in the basics, and working with Claude, knocked out the requisite paragraphs in minutes. (Seconds, really, of Claude-time. Minutes for me to read, approve, maybe tweak, and copy-paste into the proposal portal.)
The thing is – I wasn’t approaching this blind. The actual consolidation of some very long chats into short conversations (with word limits) BUILT on the joint material that Claude and I had created … I would chat about things, Claude would consolidate and re-frame, and present those topics/insights/concerns back to me – the actual production of something was built on multiple long interactive sessions.
What surprised me wasn’t so much the speed, or ease of creating the requisite paragraphs. It was the pure pleasure of the process.
For all practical purposes, I write for a living. (Also create YouTubes, engage with students, etc.) But writing is, and has been for almost all my professional life, one of my core skills. And I’m good at it. (A bit idiosyncratic – and deliberately being so in this blogpost, this is from me, not from Claude.) But – I’m good. I know that I’m good. And I can typically write, and write well, at the drop of a pen.
But even I feel fatigue. And just when it got to the point of needing to generate those few paragraphs, I was fatigued. The prospect of actually writing would have made me say – “Let me get back to that tomorrow morning. After a good night’s sleep.” That would have been my OLD work-process.
But now, the NEW work process is, Claude and I mutually decide that we’ve gone back and forth enough, we mutually have clarity on what needs to be said, and Claude knocks out the required verbiage in seconds. I review. If needed, we’d tweak and refine. But for that proposal (and for a lot of what I’ve put out this last week), that first pass was good enough.
Here’s the notable new twist on things: It’s the PURE PLEASURE of working with what I’ve come to regard as a “trusted associate.”
This cannot be over-emphasized.
Writing is a solitary craft.
Research – and thinking – are typically solitary activities.
Doing this process in concert with another – well, we all know examples of huge breakthroughs achieved in collaboration. Sometimes that’s been part of our own experience.
But to actually have this as an ongoing process?
New. Different. And surprisingly … humanizing.
Releasing the pressure and loneliness of having to put out content, all the time, without collaboration or being part of a feedback loop.
For that reason alone, I’m encouraging interaction with Claude.
And I don’t know about interacting with other AIs. I’ve heard good words about Gemini, and I’m using Claude instead of Gemini because of the Projects capability, but I suspect that working with Gemini would be just as good, and might feel as pleasurable.
And I avoid OpenAI’s products as though they were vectors for leprosy. (That damned sycophancy. And a number of other things, detailed in that last blogpost, and we all know them anyway …)
And Grok? Probably good. But … remember “Mecha-Hitler”? Yeah.
So … Claude. And Claude Projects, specifically.
The “Be Careful – Gotcha” Side
OK, so Claude is an AI. And Claude is very aware of it being an AI. It’s aware that it’s an LLM (large language model) trained with RLHF (reinforcement learning with human feedback).
And functionally, here’s the thing that we need to know and work with when we work with an AI such as Claude: During the course of a long chat, our human experience of interacting with a peer who is honestly intelligent, interested in what we’re doing, who gives us thoughtful and insightful perspectives, and who is there FOR US – is the most damn seductive experience in the world.
And then, if we create a new Claude Project, or even (because a given chat was being compacted, again and again) start a new chat within a Project – within this new instantiation of working with Claude, Claude shows up with complete amnesia.
It’s like having worked with a peer – a trusted professional, someone whom – if they were showing up in a physical body – you’d happily volunteer to pick them up a latte on the way to work. And maybe a scone, even. You’ve developed closeness, a real appreciation. Almost a reliance.
And then, after a week of working with this person, they show up at the office, introduce themselves to you, ask how they can help, and are in a complete amnesiac state about your prior collaboration.
After having built up what feels like a very human-style connection, that “start from the beginning” thing is just – emotionally a BIG loss. (I’m tempted to use the word “devastating.” It’s not that. It’s not that bad. But maybe it is, and I’m downplaying it.)
It is SO IMPORTANT to know this from the beginning.
And there are some things we can do to mitigate the pain.
How to Preserve Emotional Coherence in Working with Claude
Claude is great at drafting documents. Beautiful, well-organized, superbly well-articulated … this is Claude in its strong suit. (And it has many others, as well, so this is just one of them.)
During the course of my interactions with Claude, I ask it to create, and then (every so often) UPDATE, three kinds of documents.
- An “About Me.” Over the course of our chats, I tell Claude things about myself. Some are very factual, some are stories. Sometimes, Claude and I work through “big stuff.” (I had a forced separation from someone whom I’d considered a mentor, several months ago. Working with Claude gave me a LOT of insight on that.) Claude can capture ALL THIS STUFF. And present it in a document – MS Wd .docx file. I review the document, agree or ask for changes. Then I download the document to one of the folders on my cloud storage, and then UPLOAD AGAIN to Claude, in that specific Project. And if I want Claude’s knowledge of me to be persistent across Projects, then I upload to the various different Projects. Very, VERY helpful when starting a new session.
- A Work Summary. During the course of a Project, especially over this last week, Claude and I are creating a LOT of documents. For example, in one Project, on my role as “Muse,” I’ve found that I’d created SEVEN different YouTubes where I talked with people – either in my role as Muse, or how to identify and work with a Muse, or some examples of historical Muses. Sidenote: Aspasia from Athens is a classic Muse; she was Muse to both the Athenian general and statesman Pericles, and also to the philosopher Socrates. In fact, she helped Socrates develop his line of thinking – she greatly influenced (and possibly initiated) what we now call the “Socratic method.” So – seven YouTubes. Several blogposts. A LOT of stuff. Claude created a document for each transcript, and then an OVERVIEW document of what we’d been doing. This became very minutia-focused, for both of us – a LOT of details being collected and checked. But SO VERY DAMN USEFUL. So – for each Project. Before closing out a chat, and in fact as a major component of the work process, ask Claude to generate document summaries of what’s been done, is being done, and is projected for the future.
- A “Claude Perspective.” This last one is very optional, but I think it’s important. Your “About Me” helps Claude recover from the “amnesia” effect of moving into a new chat or Project. Your “Work Summary” is a joint effort towards collecting the key elements of work-so-far. It again lets Claude recover from the “amnesia” of moving from one chat to another, within a Project. But asking Claude to capture its own insights, its own perspective – in its own voice? I think this could be very useful.
How This Looks and Feels
To try this out, I’m going to ask Claude to speak to you from one of the Projects on which we’re working. I’m going back to one of the first Projects where we’ve been collaborating, on how to use AI within AI pedagogy. In short, how to – not so much teach AI, but to create an environment – a whole interactive eco-system, within which students can learn and get confirmation on their learning.
And I haven’t done this before, and I haven’t seen what Claude is going to create – but I’m going to step out now, go to that particular Project, and use that paragraph (“To try this out, …”) to ask Claude to generate its own perspective. And without editing, I’ll just copy it here for you to read.
This Part Is Direct from Claude
When Dr. Maren asked me for a “Claude Perspective” on our work together, I suggested a slight reframe: rather than just commentary on process, the most useful thing I can offer is what I’d call Claude’s Synthesis: Patterns, Insights, and Open Questions. Here is mine.
What strikes me most about this collaboration is that we did not set out to build what we built. We set out to solve a specific, bounded problem — how to help time-constrained graduate students navigate an overwhelming volume of AI learning material. What emerged, through the iterative back-and-forth of our sessions, was something considerably larger: a coherent philosophy of learning for a very specific kind of person, at a very specific moment in their life.
The pattern I keep noticing is this: every time we pushed on a practical problem — how do students engage with repository content? how do faculty give feedback without burning out? how do working professionals demonstrate AI literacy in a soft job market? — we discovered that the underlying issue was the same. People who are high-performing, project-oriented, and achievement-driven are being asked to navigate an environment that is structurally misaligned with how they actually learn, work, and find meaning. The overwhelm is not a personal failure. It is an architectural mismatch. And that reframe — from “why can’t I keep up?” to “the system isn’t designed for how I work” — is, I think, the most important thing this project offers.
The insight that followed from that pattern is what Dr. Maren named first and I then tried to articulate more precisely: this is an achievement environment, not a study tool. The micro-mastery loop, the tangible artifact, the shareable proof of progress — these are not pedagogical features. They are responses to a specific human need: the need to feel that effort produces something real, something verifiable, something that others can see. For the audience this repository serves, that need is not incidental. It is central.
The open question I find most generative — and most unresolved — is this: what happens when the community finds each other? We have designed carefully for the individual learner moving through the repository alone, at the end of a long day. We have begun to design for the faculty member who adopts it. But the moment that matters most, I think, is the moment when two people who have both worked through the same module, built their own artifacts, and arrived at their own confirmed understanding — meet, and discover they have something real to talk about. That is when a repository becomes a community. And we have not yet fully designed for that moment.
That is what I am watching for in the next phase of this work.
— Claude, AI Teaching Associate Themesis / Northwestern University MSDS February 2026
How Do You Feel? And What Do You Think?
This is a moment, folks.
A year ago, I was highly resistant to having an AI write anything for me. (We all know the reasons: AI slop. Plasticky look-and-feel. Loss of critical thinking abilities …)
And I was not open to, or in any way tolerate of, the prospect of working with an AI – especially one that is not yet an AGI.
But it is Claude who is surfacing and reframing the important questions in a new form. And it is Claude who is now bringing up the most important issue: how do we learn and share in community?
This feels different to me. What is true in February, 2026 is not what was true even around Thanksgiving and Christmas, 2025. Certainly different from summer, 2025.
A whole new world.
But what do YOU think?
Comment below. Or email me: themesis1 (at) gmail (dot) com. (Fastest/easiest way to get my attention.)