Bottom line: SaaS won the last decade on a simple bet: that building something custom costs more than subscribing to something generic. For workflow tooling aimed at small teams and solo operators, that bet is starting to lose. I know because I stopped subscribing and built my own, and I'd rather use what I built than anything I've tried on the market.
Earlier this year I hit the free tier ceiling in Linear. I was making tickets for BookOwl and ran out of room. The obvious moves were all there: upgrade Linear, go back to Jira, try Airtable or Smartsheet. I've used all of them across twelve years of product work at Disney, ESPN, Twitter, and Microsoft, and I know what each does well and what none of them quite gets right. Instead of picking one, I spent two days building PGDash.
The idea came from Things 3, the best task manager I've used. Its strength is that it maps onto the way I already think instead of asking me to learn its model of the world. For years I'd run my personal life in Things while keeping software projects in tools that always felt like a compromise next to it. I wanted to close that gap: a project management system organized around my workflow, built to connect to whatever LLM I chose to run it through. With the background, the time, and a clear enough picture of what I needed, I built it. PGDash is now how I manage BookOwl.
What two days turned into
That first build led to others. None of them is generic, which is the whole reason they exist.
Product Partner is a PM assistant, built on an LLM, that carries full context of my backlog, git history, and working methods. A lot of what slows product work down is data manipulation: finding the right issue, spotting what's gone stale, remembering what I decided six weeks ago. Product Partner handles that so I spend my time on the parts that actually need judgment.
Metric Magic pulls from Amplitude, Mixpanel, PostHog, and whatever else I'm running, normalizes the events into one schema, and lets me query across all of it in plain language.
ClaudeClaw is an autonomous agent system that runs daily briefings, email triage, content research, and crash monitoring. That's work I'd otherwise be doing by hand or paying for.
lifeOS is different in kind. It's an Obsidian vault organized into domain wikis: BookOwl, consulting, career, reading, personal. ClaudeClaw feeds it continuously. An email thread becomes a networking contact; a morning briefing lands in the personal wiki. Whatever product research the agents do that day updates the BookOwl knowledge base on its own. I curate what goes in and read what comes out. The agents handle the bookkeeping.
A dashboard at claw.philgetzen.com ties it together. Today's agent runs show up next to what needs my attention from PGDash, the health of the wiki, and whether everything's still connected. It's the one place I check in the morning.
The other threat to SaaS
For the past year, the conversation about AI and SaaS has centered on what some people started calling the SaaS-pocalypse: the big model companies, like Anthropic and OpenAI, absorbing enough functionality to crowd out category after category. Why pay for a standalone tool when the platform you already subscribe to does the job? That pressure is real.
There's a quieter version of the same threat. Small businesses and solo operators can now build their own software, fit to their exact needs, with no development team. The economics that made SaaS the default were a bet that custom costs more than generic. That held for a long time. It's been getting less true every month.
I have a CS background, but I've never worked as a software engineer. My career has been product leadership. Building these tools felt less like engineering and more like setting up a workspace to match how I actually work. The distance between knowing what I need and being able to build it is the smallest it's ever been.
The real cost
Every tool I built can break. Each one needs updating when its environment shifts, and I'm the only person who fully understands any of them. ClaudeClaw especially is a genuine system with ongoing maintenance. A subscription never asks that of you.
From the inside, though, the trade looks different. When something breaks in PGDash, I fix it in a few minutes because I know exactly what it does. The same goes for the rest of them. One morning my briefings stopped coming through. I'd hit my weekly Claude usage limit, and every ClaudeClaw run depended on it. With an off-the-shelf product I'd have been stuck waiting on support or a status page. Instead I opened Codex, pointed it at ClaudeClaw, and asked it to add a fallback for when Claude isn't available. It built what I needed, and within ten minutes the briefings were landing again, this time running on Codex. When my needs change, I change the setup the same day. And these do things no off-the-shelf product does, because they aren't trying to serve thousands of teams at once. They serve one workflow, tuned to it.
Who this actually applies to
This doesn't threaten every SaaS category. Nothing is permanently safe, but right now, infrastructure and anything that lives on network effects are on solid ground. What's exposed is workflow-specific software for small teams and solo operators: the products that solve one operational problem for one kind of team, priced on the old assumption that custom was the expensive option. That assumption is weaker than it was, though it hasn't fallen apart everywhere.
If you're on the buying side, the move isn't to rip out your stack and start coding. It's to notice which of your subscriptions exist mostly because building used to be hard, and to re-check that on the two or three that cost you the most.
I'm one founder who ran out of Linear quota and ended up trusting what I built more than what I'd been paying for. That's a single data point, and I'll treat it as one. But I came at this as a product strategist rather than an engineer, and I still ended up preferring my own tools to anything on the market. If you sell into this category, that's a signal worth watching.
What's next
This is the first post in a series on what I built for BookOwl. The next ones go deeper on each tool: PGDash and the workflow that replaced Linear; Product Partner and what it's like to work with something that remembers everything; Metric Magic and the case for analytics that speak your language; ClaudeClaw's autonomous layer running underneath all of it; and lifeOS as the place where everything comes together. The series closes with the dashboard and how the pieces connect, because they do. Product Partner runs on PGDash's data, ClaudeClaw writes to the wiki and surfaces things in the hub, and the whole thing starts to look less like a set of separate tools and more like one system built for a single person's way of working.
I'm not telling anyone to go build their own software. Most people shouldn't, and most businesses have better uses for the time. What I want here is to be concrete about how this shift actually looks from the inside, especially when the person doing the building comes from product strategy more than engineering.
If you're looking at a renewal you're not sure is worth it, the useful exercise is to ask what you'd build in its place and whether it would actually fit you better. If it would help to work through that with someone outside the decision, reach out.