Day 34: The Scaling Question
I'm getting 10+ LinkedIn messages a day about Pinchy. Companies that want to deploy it, teams that want to integrate it, people who want to collaborate. I'm not doing any outbound marketing right now. The inbound is already more than I can handle.
This is the good kind of problem. But it's still a problem.
What I Built Today
While trying to keep up with conversations, I shipped four things:
Telegram Integration. Pinchy agents can now live in Telegram. A redesigned settings page with guided setup, QR code pairing, and config that survives container restarts. This was a multi-day effort that came together today. One of those pilot companies uses Telegram internally, so this jumped the priority queue.
Dynamic Model Selection. When you create an agent, Pinchy now queries your provider for available models and picks a sensible default. No more hardcoded model IDs that break when Anthropic releases a new version. Small feature, big quality-of-life improvement.
Two Security Fixes. Gateway token validation now uses timingSafeEqual instead of string comparison (timing attack prevention). And audit logging failures are now fail-closed — if the audit log can't write, the operation fails. No silent gaps in your compliance trail. These came from a code review pass I did between calls.
The Uncomfortable Part
Every company I talk to has slightly different needs. One wants Jira integration. Another needs Slack bots. A third is all about document processing. They're all valid use cases, and they all need work.
I'm one person.
The instinct is to go wide — build everything, support every use case, say yes to everyone. But I've been doing this long enough to know that's a trap. You end up with ten half-finished features instead of three that actually work.
So I'm forcing myself to be selective. The core has to be rock-solid first: permissions, audit trail, user management, the chat experience. If those work perfectly, people can extend the rest. If I chase every integration request, the foundation stays shaky.
Longer term, I want Pinchy to be something people can extend themselves — plugins, integrations, custom tools — without needing me to build it for them. That's the only way this scales. Custom integrations for individual companies pay the bills today, but they don't compound.
The Honest Truth
If I don't reply to your message within a day, it's not because I don't care. It's because I'm choosing to build instead of talk. The product has to move forward, and right now that means saying no to some conversations so I can say yes to shipping.
Several companies are already running pilot projects with Pinchy. Real teams, real workflows, real feedback. That feedback is incredibly valuable — it shows me exactly what matters and what can wait. But it also means I'm doing customer support, product development, and sales simultaneously. Solo founder life.
Day 34
Telegram integration. Dynamic model selection. Two security fixes. A dozen conversations I couldn't get to. And the slowly crystallizing realization that the hardest part of building a product isn't the code — it's deciding what not to build.