Stop paying for software you could build in an evening
Cancelling subscriptions one by one. Not out of cheapness. Building it yourself is faster, simpler, and actually fits the problem.
The old logic was simple. Need a tool for your team? Pay thirty to fifty dollars per seat per month, or park a developer on it for two months. Almost everyone paid.
Now that math is broken. With Claude or Cursor I build a working Kanban in one evening. A small CRM in two. A task tracker with Telegram notifications in a week.
So the question becomes obvious. Why am I paying four hundred dollars a month for ClickUp for ten people? What does it do that I couldn't write myself?
Usually, nothing.
This isn't a fringe take anymore. Klarna dumped Salesforce and Workday, replaced everything with their own AI stack, saved four million dollars a year. A marketing agency on Twitter cut eighty percent of its subscriptions and rebuilt in two weeks. In January, software companies shed three hundred billion in market cap in a single trading day. The market is getting it.
What you can cancel right now: simple task trackers, CRMs with one funnel, form builders, time trackers, dashboards from your database, internal wikis. All of it builds in a couple of evenings and solves your actual problem, not the problem some team in San Francisco imagined you had.
Not everything. Stripe stays (licences). Slack stays because your clients live there. Figma stays because designers won't budge. Corporate Salesforce with security audits stays too, but that's a different universe.
If you're using 5–10% of the features, build your own.
If the service is an ecosystem (integrations, community, network effects), keep it. That value isn't something you can replicate on a weekend. But if you're paying for a feature list and using two items on it, that's not a subscription, that's a tax.
One honest thing. Last week a Claude agent at some startup deleted a production database in nine seconds, backups included. That's part of the deal. If you're not ready to be your own admin, devops, and support, thirty dollars a month is a fine investment.
But if you are ready, there are ten times more options now than a year ago. Don't sleep on it.