The newsletter that built itself (and what it means for you)
By Nick Basile
Welcome to issue one!
This is the first edition of the Learn AI with Nick newsletter -- a weekly newsletter for small business owners and solo operators who want to use AI to actually get things done. I'm Nick, and I've been building AI workflows for my own business and for clients -- I've seen what works, what doesn't, and I'm excited to share all of it with you.
If you're coming from the Learn AI Houston group, welcome -- same mission, just a little more personal and a lot more room to grow.
Glad you're here. Let's get into it.
The newsletter that built itself (sort of)
Okay, Iet’s get a little meta before we get into the links this week.
This newsletter was assembled by a skill.
Not written -- assembled. I built a workflow that pulls from the newsletter digests I save every day, finds the dominant themes across the week, queries the personal knowledge wikis I've been compiling on topics like AI agents and Claude skills, and hands me a synthesis ready to work with. I reviewed it, shaped it, added my own thinking, and here we are.
That's the story I want to open with, because it's not really about the newsletter.
Here's what it's actually about. You already have information-seeking behaviors. You read newsletters (hi). You save links. You highlight things you mean to come back to. You have a working sense of what's happening in your industry because you're in it every day. The information flows through you constantly. The question is just: what happens to it?
For most people, it evaporates. Inbox graveyard. A read-later list that quietly became a guilt list. Tabs. You absorb some of it. Most of it disappears.
What if it fed something instead?
I've been building a system where my reading habits feed living knowledge repositories. Articles get processed into daily digests. Digests get compiled into topic wikis -- structured, queryable knowledge bases I can actually interact with and refine over time. The newsletter is one output. But it could just as easily be a competitive intelligence brief, a marketing angle, a sales talking point, a quarterly "here's what shifted in our industry" for your team. This isn't just for AI content. It works for any domain where staying current matters. Marketing, operations, whatever your business cares about most.
The information is already there. The system just makes sure it doesn't evaporate.
What’s going on with AI this week
Taking a look at what's happening more broadly with AI this week, there's a growing realization that the AI tools we're using today are heavily subsidized. What you pay for Claude or ChatGPT doesn't come close to what it actually costs to run that infrastructure. These companies are burning capital to compete for market share, betting they can lock in enough users and enterprise contracts to justify the spend later.
That works until it doesn't.
We're already seeing early signs. Access concentrating in higher tiers. Capabilities quietly reserved for enterprise contracts. Subtle model degradation that's easy to miss unless you're paying close attention. Large organizations with dedicated AI budgets will absorb whatever shift comes. Solo operators and small business owners may find themselves working with less capable tools -- with no announcement and no changelog.
I've felt this at the skill level. A workflow I built when a model was performing at its best can produce noticeably worse results a few weeks later, not because anything about the skill changed, but because the model underneath it did. The ground shifted.
This is why human review stays essential. The model can't tell you it's degraded. The skill can't either. You need a person in the loop who's objective enough to notice when yesterday's output was better than today's.
Build your workflows. Automate what makes sense. I'm not saying slow down. I'm saying: keep your eyes open. The foundation can move without telling you.
This week's links
Anthropic took down thousands of GitHub repos after accidentally leaking Claude Code's full source — TechCrunch
Anthropic accidentally shipped 500,000 lines of Claude Code source code inside an npm package. The community found it, mirrored it across 8,100 repos, and the resulting clone became the fastest-growing GitHub repository in history -- 100K stars in a single day. The interesting part isn't the leak itself; it's what the leak confirmed: the real work in AI isn't in the model. It's in the systems built around it -- the workflows, the guardrails, the specialized instructions. That's the part anyone can build, including you. The model is the starting point, not the finish line.
Sarah Friar, OpenAI's CFO, has reportedly raised private doubts about both the Q4 2026 IPO timeline and whether $600B in planned spending pencils out. Internal projections show the company burning $200B+ before hitting positive cash flow. Here's why this matters if you're a small business owner using AI tools: when the economics stop working at the top, companies raise prices, cut access, or both. The monthly subscription you're paying today is likely still subsidized. Plan accordingly.
DualEntry raised $90M at a $415M valuation on roughly $400K in annual revenue. Not a typo. The strategy: flood a startup with capital so early that it looks dominant before it's proven anything. A lot of the AI tools landing in your inbox with massive funding announcements and bold claims haven't earned that position yet -- they've just been handed it. Funding is not the same as working. Evaluate the product, not the press release.
Why people defend the systems that are failing them — The Decision Lab
System justification theory: the people most harmed by a broken system are often its most committed defenders -- not because they're irrational, but because accepting the system is broken forces a more uncomfortable conclusion about their own situation. If you've ever tried to get a team member, a longtime employee, or even yourself to drop a process that clearly isn't working, this explains a lot. The fix isn't to argue the system is broken. It's to show what a better version of the same system looks like.
What actually makes an AI product defensible — Silicon Valley Podcast
Mercedes Bent of Premise VC says we're in the "hubris" phase of the AI cycle -- lots of capital, unclear unit economics, a lot of noise. Her filter for what's actually worth paying attention to: does it feel magical the first time you use it, does it spread without a sales push, and does it get harder to leave the longer you're in it? That last question is the one I'd ask before building any core business process on an AI tool. If walking away feels easy, it probably isn't the foundation you want.
How a recommendation newsletter became a social app — Creator Spotlight
Perfectly Imperfect started in 2020 as a Substack where people shared what they were into -- no algorithm, no optimization, just recommendations in reverse chronological order. It grew into a platform with a genuinely passionate community by doing the opposite of what most platforms do. As AI-generated content floods every channel, the thing that gets harder to fake is genuine taste -- a real person's honest take on what's worth your time. If you have any kind of content presence, a newsletter, a social account, a blog, that authenticity is your actual edge right now.