When Trust Breaks: Our Experience Leaving Harvest and Building a Replacement

We've been Harvest users for years. It's a time-tracking app we reviewed positively on our blog and recommended to others. It did what we needed, and we were happy to pay for it.

Recently, Harvest introduced a significant price increase. We reached out to the company, and they offered us a discount for the next year. It was a reasonable response, but something had changed for us. The trust was broken. Not because of the discount itself, but because we started wondering what next year would look like, and the year after that.

We decided to cancel.

---

Building Our Own

Instead of searching for alternatives, we tried something different. We opened Cursor, described what we needed, and started building. A few hours later, we had a working Mac app that handles everything we use daily—time tracking, basic reporting, the features that matter to us.

This approach has a name now: vibecoding. You describe what you want, the AI generates code, you test it, adjust your description, and iterate. We're not professional developers. The AI handled the technical work. We just needed to be clear about what we wanted.

The result isn't Harvest. It doesn't have all of Harvest's features, integrations, or polish. But it has our features—the ones we actually use—and it works.

---

The Broader Context

This week, software stocks dropped significantly after Anthropic released new AI automation tools. Nearly $1 trillion in market value disappeared over two days as investors reassessed the outlook for subscription software companies.

I'm not suggesting our small experience explains any of that. But it does reflect a shift that's becoming more visible. The barrier to building simple software tools is lower than it used to be. Vibecoding is becoming more mainstream, and for certain use cases, it's genuinely useful.

---

What This Means for Subscription Software

Enterprise software isn't going anywhere soon. Complex systems with deep integrations, compliance requirements, and large-scale support infrastructure serve needs that a weekend project can't replace.

But there's a category of software that exists in a middle space—tools that are useful but not deeply integrated, that serve specific functions without requiring enterprise-level complexity. Time trackers, note apps, simple project management tools, invoice generators.

For these tools, the calculation is changing. When prices rise or trust erodes, some users now have an option they didn't have before: build something that works for them.

Most people won't do this. Most people don't want to spend hours iterating with an AI to build a time tracker. But some will, and that changes the dynamic slightly. The switching cost isn't what it used to be.

---

Our Takeaway

We liked Harvest. We recommended it publicly. We were loyal customers. When the pricing changed and trust broke, we discovered that leaving was easier than expected.

That's our experience. It's a small story, but it reflects something real about where software tools and AI coding are heading.

---

Related Resources

Looking for more insights on business software decisions and AI tools?

• Best Business Software Stack: Essential Tools for Small Businesses
• AI Agent Security for SMBs: 2026 Comprehensive Playbook