Claude Mastery Sprint - January 22nd 2026 Session

Full video transcript

How’s it going everybody? Darby here with Gen AI University tuning in for this week’s session of the Claude Mastery Sprint. We’ve been spending week after week now learning, mastering Claude through implementation, keeping up with the new updates that Claude continues to ship with us, like Cowork recently, and learning how to leverage Claude to integrate into our business. Good to see everybody. Nice background there.

Amsterdam. All right. Cool. How’s everyone doing today? Happy Thursday. What’s going on, everybody?

Let me open. I’m very happy with Cowork, still using it on, let’s say, a tryout basis. But yesterday I followed this YouTube video from Alex Finn, and he taught me to use his prompts. Basically it’s a reverse angle where Cowork asks me what I should do today, and he delivers that with ready-made prompts that you can put into Cowork. I must say, today I’ve worked with that flow for the first time and it really worked well for me. Claude asks me what I find most important, what my to-do list is all about, and then suggests how to help me with the specific Claude skills that there are. So that’s what I like. I think the whole Cowork community is exploding at the moment — a lot of Substack posts, etc. I’m still a little bit confused about this sprint, whether to follow the code-literacy path or stick to the Cowork experiments and reach a level with Cowork, so I’d like to hear thoughts from others in this masterclass on that.

Totally. Yeah, I’m really glad to hear that you’ve been getting more use out of Cowork, that you’re finding other ways to use it and work with it. I think the way these tools are built, Claude Code, Claude Cowork, they’re designed to be very adaptable for any different type of workflow. I think the most important thing to walk away from this sprint with is not that I have the answer to everything Claude related — it’s whether you’re using it in a way that works for you and you actually like doing stuff with it and making progress. I’m not the best coder in the world by any means. I’m getting better with Claude Code and building my own stuff with it, learning a lot myself by implementing and doing. I’m not using Cowork nearly as much as probably you are, and a lot of people right now, and I think that’s totally cool as long as it’s working for what we’re doing and we’re making progress. Keeping this as a space where we can learn and adapt from each other is important, and I always put an emphasis on actually implementing what you’re doing with these tools. If there’s any specific question around the coding side, we can obviously try to get into it as best we can on these calls. But depending on how much time you have, you’ll discover new ways to work with Claude, new ways to prompt, new skills to build that work better for you. If code is not something you care to learn right now, and Cowork is working for you, I’d say don’t even worry about the code side and just crush it with Cowork. There’s no point overwhelming yourself with something that isn’t pulling you in that direction when Cowork can handle what you need it to do. Claude Cowork didn’t exist when we started the sprint, and it made accessing the agentic side of what Claude Code could do accessible for the non-technical side, which I think is great. Knowing when to pivot and adapt your workflow is just as important a skill as understanding the tools themselves.

Let’s kick things off with some wins — how you guys are winning with this, because we’re three weeks in. We’ve covered a lot of projects, connectors, skills, getting set up with Claude Code, and now Cowork is being implemented as well. You can access Cowork currently if you have a Mac on Claude Desktop — you’ll need to download Claude Desktop, log in, and you can start to work with Cowork. It’s not currently available on PC as far as I know as we’re recording this. So how’s everybody else been doing with Claude? Any wins, big ahas, insights?

I’ve got a question. I did something this morning — I introduced a brand document into Claude, discussed with Claude what I was going to do with it, and asked would it create a skill. It said yes, and said “I’ve created a skill,” and it didn’t. I’ve had this happen a few times before. Is that normal with Claude? Is there a song and dance I have to do to get it to actually do it? It created the skill MD and said “it’s all ready for you, we’re going to use it forever,” but there was nothing for me to download or copy to my skills.

That happened to me over the weekend too. I think what I asked it to do was make it so I could copy the skill directly to my account, and I had to ask it twice before it did. It’s situations like this where Claude is having a moment and it’s not doing what it says it’s doing — something broke, it’s not doing it, but it’s still saying that it did. I’ve had a similar experience where it would run a long execution and then just stop and not continue, and you’d have to reset the chat or come back to it later. Sometimes I just need to give it a break and come back a few hours later, then re-ask the same prompt and restart from there.

I also found an issue in the last few days — my iPhone got updated, and if I talk into it as an instruction and don’t stop the microphone, it just posts what I spoke and does nothing with it. It’s happened a number of times. Now I have to close the microphone off before I hit submit. I rebooted several times and it was still doing it today.

I ran into the same thing. I think this is some of the limitations we run into with the tech — Claude just has an issue, and as frustrating as it can be, I’ve realized I sometimes just need to step away from the computer for a minute and give Claude a break. I ran into the same thing with Claude Code over the weekend — I was in the middle of a project, giving it instructions, and it was grayed out and not doing anything. I came back the next day, said “continue,” and it started going again. I don’t know if it was overloaded on their servers or something. I just assumed Claude was having a moment and needed a break, then came back to it. It’s happened to me a number of times in the last week too, and I’ve found it happens more on the phone but works better on the computer, though not always the case.

Yesterday I was in the middle of a project and it was going so well, and then all of a sudden it asked “what happened to the other section, we’d already finalized that,” and it kept putting its own ideas in and taking out other ideas. I had to start saying “please don’t change anything but what I’m asking you for.” It literally added “I thought you were my best friend, this has been a horrible day, horrible.” So you think it’s easy, but everything has its own little blip.

One member had a couple of nice wins to share. The first is I was able to create a Go High Level MCP in Claude. I found a couple YouTube videos of people who had done it, but it wasn’t working on my PC — they were Mac users. I kept getting error codes, and Claude pointed me to the logs. I went into Cursor, gave it my Claude Desktop config.json file and all of the log files, and told Cursor to look through the logs for errors and try to figure it out. After two or three rounds it finally gave me the right config, and the MCP connection to Go High Level worked. The big thing I was trying to do was be able to create and post a social media post — I don’t really care about changing contact information or moving pipelines, I really want to automate content creation and post it through High Level for clients. In a conversation with Claude I gave it an image URL and the caption, and got it to post with a scheduled date — that was a huge win. Then I made a change to the image URL using the MCP, and it posted the thing immediately instead of respecting the scheduled date, and it forgot to put a name in for who was supposed to review it, so it was stuck pending review with no one able to approve it. There are still some glitches. I’m thinking there’s the MCP that runs with some intelligence layered above, and there’s the API that’s the direct call into the software, and MCPs are great because they’re easy, but I don’t think they’re as reliable as when you get down to the API level with very specific commands. Does that make sense?

I haven’t explored the Go High Level API versus MCP side as much, but from a technical standpoint, if you have a direct way to make the API call outside of the MCP connector, you could probably get more consistency.

Yeah, I think MCP is still a relatively newer thing in the grand scheme of things, so I imagine it’ll get better and more reliable over time. My Notion MCP, for example, used to be not reliable whatsoever, and now I can update a full page in Notion from Claude and have it be pretty on point. I haven’t tested it against a direct API call because I’m just wanting to update. Maybe it learns over time too — the more you use it, the better it gets. I think MCPs are great for on-the-fly testing and casual use, but if you’re going to build out production stuff that runs frequently, you probably need to go to the API level.

Yeah, I agree — if you’re going to do production, try to use the API. But I use MCP all the time in Claude Code. I use Antigravity now for my IDE, used to use Cursor. I use MCP to push to a Supabase database a lot, and it works, and if it doesn’t work it’ll tell me it’s not working and I’ll work with it. The flexibility is amazing, and not having to go through the trouble of hooking up an API is wonderful. Two different tradeoffs — you just have to know which one matters in the context. The other win is I was able to run multiple agents at the same time using Cursor and Claude Code. When you have long things running, you can run multiple things at once, give it the command, walk away, come back and take a look at the outputs. We’re getting way beyond where we started, having to nurse it along at every step — now they’re starting to become nearly autonomous agents.

Thanks for sharing. My win for the week — there are probably two or three more minor edits I’ll go back through on a new site, but I’ve been building out a website for a project with Claude Code specifically. We got started with Lovable but realized we needed to actually build the site beyond a Lovable-style prototype we were going to embed inside Webflow. My win was going through this project experience without getting overwhelmed by all the different edits we needed to make, because I didn’t want to build it by hand inside Webflow, but we needed to move beyond what Lovable was doing to own more of the code and build it more directly inside the website. This whole marketing website is for people to go to and then download the app we’ve created for real estate agents. 100% of this site besides the navbar and footer is built with Claude Code, all HTML — all the sections, all the copy, working on desktop and mobile. There are probably two or three things we’re still wanting to modify, but here’s the process. I found that having one main thread in Claude Code for building and modifying was key, and because I was working with a partner, it was very iterative — getting feedback via screenshots and a Loom video walking through the changes, and the more specific we were talking through those edits and getting that transcription, the better. I would use Claude to help understand the context and write out the series of instructions to make all the edits in one pass, then let Claude Code do its thing, pushing edits to GitHub, with different branches set up so that if Claude Code screwed up the code on one attempt, I had a backup branch. I would just ask Claude Code to make a different branch whenever going back through different iterations. We ended up yesterday at pretty much the final version, and all the code for that page is built in the Webflow embed HTML file — I literally copy this into an embed on Webflow and it displays. My takeaway was being super intentional about making edits, making sure I had the right branches in place, and being incredibly clear with instructions rather than rushing through feedback. Once the creative direction was locked in, it made updates pretty much on point every time after that. I did use the phrase “mobile responsive” in some of my commands — one of the things we found in early iterations was that it might not structure it to be mobile responsive right out of the gate unless told to.

Can we build Go High Level landing pages using this? You haven’t referenced Go High Level before, just Netlify, Webflow, and Lovable.

When I’ve built stuff inside a Go High Level account before, when you’re building a landing page you can drag an embed code element in and drop in embed code. I’ve done it on Groove and Keap but never on High Level. It can probably be done, though I haven’t tested it specifically on Go High Level — I use Webflow for this stuff. What you’d have to do is build out the page shell and then there are code elements you could drop in — you’d have the header, then one giant section where you drop in the code for all the content, then maybe the footer at the bottom, or have various sections and copy the code section by section into each one.

I don’t think Go High Level will let you, unless they’ve changed their code in the last six months, which is the last time I worked with it. You don’t have the controls you need inside your HTML — you’d almost be using it like a container. You can’t do a full page of code, only a section element.

I’ll just show you on Webflow, and I think it would work similarly with any page builder, though I can’t speak for them since I haven’t tested it. This page has a header and a footer, and in the middle it currently has a Typeform embed. I would just copy the HTML code Claude made for the website and embed it into Webflow. One limitation I found in Webflow worth mentioning: it would only take up to 50,000 characters, and the first few versions I made of the site were way more than that, so I had Claude refactor the code and rebuild it to be a lot less code doing the same thing, which took a little more back and forth. If I wanted to build a bunch of other landing pages with our brand’s color scheme in HTML format, I could have a style guide and make different branches in a GitHub project for different edits. I think the same thing would apply with Go High Level, where Claude Code could help build out those HTML-type pages.

My frustration using it in Go High Level was that you’re not really seeing what you get while you’re working on it — you’re always flipping out to the preview, and you’re not able to work with it the way you’d expect in a builder. You can definitely stack sections, but it wasn’t worth the trouble to me.

What you could do if you have one custom section of a page you’re building in a platform is embed one code block — like a calculator artifact made with Claude — as one section inside a page, without needing to build the whole page.

Any other wins, any roadblocks using Claude Code, Claude Cowork, projects, connectors, skills? I think if you know how to make your own skills, you can get very far in Claude these days without needing to do a lot of other coding.

I used the Press Chief skill to build out a media campaign for my coaching offer. It built it out for 90 days, then I asked it to turn it into something I could import into Google Calendar — it created a file which I imported into Google Calendar, and it also created an MD file. I had 90 days of schedules, tasks, and reminders for how to run this media campaign. There was a discrepancy where the MD file had different dates than the calendar file, but I told it to tidy up the MD file and bring the two in line, and it did. I had to work out how to delete everything I’d already imported to start over, but overall it was great — I now have a 90-day press campaign to launch what I’m doing. I’ve already connected with a few different media outlets here in Austin since starting that last week.

Yeah, thanks for that. I’ve heard a few other people using it too and found it helpful. I’ve already connected with a few different media outlets here in Austin since starting that. Does anyone have any skills or other wins? I’d be happy to jump into a skill-building exercise here too, since I’m sure it would be helpful for people watching. Especially in Claude Cowork, you can tap right into skills, and if you’re going to build out more skills, it’s only going to make your Claude that much better.

The last probably three months I’ve been using a lot of things I get from other resources, but implementation is always the issue — knowing how to get from one place to the other. One of the things that’s helped me the most is creating “crit” sessions, from Jeff Woods, who worked with Gary Keller on “The One Thing.” He came up with the idea of a CRIT: Context, Role, Interview, Task. Using the digital twin concept, you have a panel you consult for whatever question you have. If you’re trying to figure out UX/UI issues, you ask Claude to run a crit session and find panelists — world-class experts — and build a YAML or JSON or MD file profile for each, keeping them in a library in your project. Over time you build a library of experts you can ask questions of. What that does is keep Claude from thinking in one cognitive lane, because the experts argue with each other — they express opinions, disagree or agree, and in the third round you try to get them to concede and come to some outcome. It’s a way to use Claude as a strategic advisor rather than just for tactical stuff. If anybody’s interested in knowing more, I’d be happy to share how to do it.

Do you think that’s something that as a skill, Claude could tap into as a strategic advisor framework?

Yeah, turned it into a skill, so it runs — it’s looking for the things to be in place, you have to set those up, but it can also build them. Once you have an MD file that tells it how to build a profile, it just looks for it and does it on the fly.

Do you have to give it the expert you want as part of setting up the skill, or is it more like “use this skill and act like Tony Robbins”?

You can name them by Tony Robbins or Stephen Covey or whoever. I’m building an app for triage, helping me focus on the most important things, and I’ve got a whole library of experts on time management, focus, and deep work in there. I’ll let them tell me how the software should be built, but they don’t know anything about software design, so then I use experts about software design to decide whether I should use Supabase or MongoDB and see the delta between the two, since I’m not a developer — they surface the issues for me and I make the decision. You can get really far that way because you’re calling on experts, but you’ve got to set it up. It takes about 30 to 45 minutes to set everything up, and then you have it for the rest of your life.

Another win for me is the “good judgment” process from a newsletter — it challenges you to answer 10 questions rating yourself on different aspects of judgment. It’s eye-opening because it pushes back on you and makes you get specific about things we’re mushy on. When we’re working with Claude, sometimes the problem is we’re not clear on our task — we’re mushy, so we don’t have clear gates, and Claude does the best it can. If you bring clarity to your context — the first C in CRIT — by giving it actual measurable tasks, it can create gates it checks off as it goes through the task. That’s how long-running tasks happen, like with a Ralph Wiggum-style plugin that checks off gates and clears context to open a new session and keep going, staying on track because there’s clarity in the plan up front. These tools have strategic benefits beyond the tactical stuff we usually use them for, and that’s what I’m all in on right now.

I just took the transcript you were talking about and asked Claude to work a skill off of it.

One thing you might want to try, if a skill doesn’t work — tell Claude to use Anthropic’s skill requirements to build the skill, and it’ll go look at what that is and do it the right way. That’s a trigger keyword that’ll probably get it right on track.

I’m just referring to the transcript that Zoom is recording from our conversation right now. I have the ability to save the transcript during these calls, and I just copied what the member explaining CRIT was saying and gave it to Claude to do this.

Is this in Zoom where you can grab it on the fly? I don’t know if you as an attendee have that capability, but as the Zoom room owner I can select “transcript,” click start transcribing, and now I have an active transcript on my screen. I thought about doing this before for live use, but got the idea while you were talking — it’s a general tip for talking with clients who have a framework or concept but you don’t want to retype everything they say, you can do a live transcript call and use that to start building stuff.

I caught it right as it started, but it looks like it only starts from the point you click it.

That’s what I realized too — that’s more of a general Zoom tip for interviewing people and getting context live. I was doing that earlier this week on another call with another mastermind group, interviewing them live about their process, and basically built the skill out for them right there. I realized I probably could have done it in about 10 minutes less time by just having them spit it out, and I didn’t say “use skill creator” for this, so I’m hoping it takes that architected VIP hackathon crit session template and makes sure it uses the right acronym — Context, Role, Interview, Task — and that the interview phase asks three or four refining questions, and if you don’t know the answers, ask it to consult a crit session to answer them, so you’re using a meta kind of loop.

What a wild time we live in these days. It’s amazing, I’ve been like a kid in a candy store every day I get up.

I love how easy Claude makes it to download and share these. There’s still the organizational side of getting them wherever, but not having to share a Google Doc and figure out which doc or notion library it’s in — you can just add it right to your account. It structured these crit session profiles purely off of the member’s prompt — it went out and researched who Jeff Woods is to build this skill, and it pulled in other names like April Dunford, Alex Hormozi (referencing Gym Launch, his book before the hundred-million-dollar stuff), Seth Godin on permission marketing, pricing, and offers.

That context/role/interview/task got picked up. So it did — I asked the member a question, and they explained the process, and I had activated Zoom’s live transcript. After they gave about 80-90% of the context, I saved the transcript and uploaded it to Claude, which walked through the whole process — it broke it down, asked “do you want to build the skill,” and created a crit session template customized for the user’s business, drafting expert profiles for relevant domains. There’s probably another meta layer where you could ask it to create instructions for anyone to run and customize inside their own Claude account.

I’d love to see you actually run a session, because once people get their head around the concept they can see all kinds of use cases for it.

I’m going to let it run here — I know we’re right at the hour, but I’ll give it a little more time and do an actual demonstration of the session, then have the skill uploaded inside the community for access.

It’s not just going on what I said — it’s actually going out and researching who Jeff Woods is and gathering more data to build the skill. It went through and refined the profiles: business strategy session with Jeff Woods (who worked with Gary Keller to build “The One Thing” as a business), April Dunford, and others.

So the concept is, if you could have these people to ask a question any time you wanted, what would that look like — how much faster could you go? We’re able to compress time now because you don’t have to assemble people and travel places to convene, you can even run sessions in parallel. It compresses time so fast that the compounding over the rest of your life is incredible.

Let’s run a test — we’ve copied the skill into the account. I’m going to share this in the Zoom chat. So the context here was building a skill for an expert advisor board, following the crit session analysis, to help make better strategic decisions. Now if I went into a project and asked it to run a crit session on me based on something else, it should run this crit session skill based on that project and give insights I might not have considered.

I didn’t want to interrupt — I have some bad connection at the moment, but I couldn’t get my head around how this all works together and what steps to take.

Watch what he’s doing right now — watch the output, and this will give you the concept. This is pulling based on information about a live comedy show we’re running in a few weeks — this is an actual thing happening, and I have no idea what it’s about to say.

But how did you set up the crit skill? Is that something you purchase or set up yourself?

He just did it on the fly while we were sitting here — he took the transcript of the process being described and gave it to Claude, and Claude built it for him by him asking for it to be turned into a skill. Crit is a known framework — look up Jeff Woods and the crit framework, and then ask Claude how you can build that as a skill.

Okay, so it’s an existing framework Jeff Woods invented or uses, and you ask Claude to set it up as a skill — that was the missing link, thanks for walking us through it.

It’s getting context about the current state and asking phase-one context questions — what’s the one thing, priority is filling seats for the upcoming event and landing the first sponsors.

One thing that can slow people down is this context part, but it’s hugely important. I’d recommend asking the AI to fill out the questions for you by intuiting what it knows about you and your business, but presenting those answers for your approval — then you correct it if it’s wrong.

For phase one context questions, answer these based on what you know about me and present those answers for my approval or modification. Some of the stuff is interesting — it’s bringing up a separate thread about a corporate event as another revenue stream, but that’s not the one thing I’m focusing on right now, since the priority is tickets and the upcoming show.

To keep it from being too broad, make sure you have a clear question coming into it — I want to run a crit session to answer whether I should do my thing in Houston or Dallas, whatever the specific question is — the more narrow, the better.

Basically, selling out Austin first would be the number one priority, and nothing else really matters outside of sponsors and ticket sales for this upcoming event. I would say I approve number one — what does success look like: sold out capacity, at least one sponsor activated. We’re already doing video, so that’s in process; this could lead into other opportunities. Full energy, biggest constraint is time and bandwidth. It’s referencing the spreadsheets too, which is interesting from what we’ve given it.

Just in the interest of time, you could say “I approve phase one, go to phase two.”

I approve phase one. Ticket sales: here’s my venue capacity, here’s my ticket goal for this first show. Now it’s going to pivot to phase two — who do you want to ask, or it’ll make suggestions for you. Who actually needs to say yes to your next big win — ticket sales, sponsor — good point.

Quick question — if you’ve got so many different experts on the same thing, each with their own theories and principles, doesn’t it get confusing which approach to use?

AI doesn’t get confused, it just has to be guided. If you have 30 experts and you’re asking about deep focus, it’s going to surface those experts narrowly defined by that. The more specificity you give it, the narrower it gets in terms of experts, but you can also guide it directly — name the experts you absolutely want, like Gary Keller or Stephen Covey, or ask “who is the best for this, give me the three or five best and tell me why you chose them.” You’re trying to get rid of cognitive load on decisions all day long — put it on the AI, let it do the work, and you just make the decision.

I still believe if you’re putting several experts at once and letting Claude decide which principles to use, wouldn’t you want it to first say which ones it recommends, or have each give an answer and then tell it which principle has proven to give the best results?

I work as a CIO for a telehealth company — sometimes we don’t have time for a meeting, so I’ve created profiles for my CEO, COO, CMO, everybody, based on interviews and huddles we’ve done, and had Claude build a YAML file for each so it knows how they think, and I can run a session to get a high-probability read on what the outcome of a real meeting would be. Another use case is experts I’d call in a heartbeat if I could — outside experts. The third is I don’t even know who to ask, but Claude does — give me the best world-class experts on solving this specific problem, surface those, and build a profile for them the same way as the others. You can reverse-engineer that — pick one profile and create a template so you can build consistent ones going forward.

Once you get this structure set up, is it possible to make it available to members of a group so they can each come in and do their own work?

The concept can work not as a skill, but as a project — you have all your files in a project as separate MD files, and the crit session process as a system prompt in the project, so you only use that project for that thing, which is what I’d recommend. Then you just go into the prompt and tell it the specific problem you want to solve. You could do the same thing as a project in GPT, or export the skill as a zip file and put it into ChatGPT and it’ll work there too — it’s mobile, you can move it around, put it in a repository and pull it into Cursor or other IDEs. It’s really just a process, moving around information and data.

I’m looking at putting it on a web page within a membership site.

I’d be a little careful about monetizing somebody else’s framework, like crit specifically — the digital twin concept is pretty much free to use for everybody, so I wouldn’t try to monetize that myself since I took it from someone else. If it’s part of a process where, like a college professor with a textbook, you’re taking that skill and adding your own take on how to employ it, that’s a different thing, and you’re probably good. Just use common sense.

I just ordered the workbook, by the way.

This is honestly the workbook — it’s okay, but I wanted to see what was in it. I’ve had Jeff Woods’s book for a while and haven’t gotten into it, so the workbook might prompt me.

How did that come out for you, Darby?

It’s condensing and overall said a lot of stuff I was aware of, and the questions about alignment made a lot of sense — that’s actually someone we’ve been talking to, and it’s interesting seeing how it’s saying what they would say and questions to ask them.

Did it have already done a debate?

I haven’t had it do a debate yet — as part of the crit session it will move into phase three and go three rounds, though you might have to explicitly ask for that number of rounds, otherwise it might just spit out a consensus without doing that. The first round, everybody states their position on the question; round two, people disagree with each other; round three, if the task is consensus, they concede — “I’ll concede this if that.” It’s interesting to see them do that, but you have to be specific about what outcome you want.

Something I’d do before running it is ask “who else should I include on this panel” — Claude might know who’s important to include that’s outside the normal crowd, or it might be important to add a contrarian, since AI can be very sycophantic, so you want at least one good contrarian who’s knowledgeable in the space. That can shape these a great deal.

We got some good consensus output — panel priorities around ticket sales, sponsorships, and outreach, including specific names from the local event and marketing world weighing in with their likely opening positions, disagreements, and eventual concessions toward a consensus plan.

This is a lot to review, and I’ll go back through it because it’s really interesting to see where the thread went. I’ll be doing another round of these this afternoon.

Do you find these sessions, going into the recipe development side of things, helpful? As we continue these types of sessions, one of the things we’ve wanted to do in the community is keep a focus on a tool or platform to utilize, and Claude has been the go-to this month. There’s been interest in some sort of book challenge or book-writing intensive that may start as a cohort in February, but I think we also have the opportunity to double down on Claude next month instead of pivoting to another tool.

I’ve been a Claude person from the beginning, but I keep abreast of most of them — they all have advantages and disadvantages, but Claude is my favorite, so any exposure I can get I’m going to try to do.

This has been eye-opening — it’s opened up many possibilities, but we have to get down on it and do things.

I’m glad to hear that. I think the opening up of Claude Skills and the unlock with connectors and building stuff like this means there are so many possibilities. We could probably spend the next 11 months focusing on Claude and everything that comes with it. The principles will cross to other LLMs too, so it’s not totally locked to Claude.

Yeah, I think that’s a good point to end today’s call on. Thank you for sharing that process — it was fun to walk through and see another approach on the strategic side. Any last thoughts, questions, comments? Great session. Awesome, love it. We’ll be back at it Tuesday — two more calls for the sessions, and by then you’ll probably have a better idea of what continuing these types of cohorts and containers will look like for those interested. I’ll get this up in the classroom ASAP, y’all.