@elijahuwas Nice to see cookieless stats built in from the start. How are you handling bot clicks and link previews? Those can make a quiet launch look surprisingly busy.
@ge0xrge We're using our own custom-built engine called Ogopogo. It's designed for AI-native search at quarry time, handling vector indexing, semantic retrieval, and intelligent real time ranking rather than relying solely on traditional keyword search. 🇨🇦
I'm proving vibe coding can build more than SaaS tools for other developers.
I'm building real services for everyday people. Search, AI, games, and more.
Technology should serve everyone, not just developers. 🇨🇦
I'm curious about the scalability of Embed 5 and Rerank 5 beyond traditional enterprise RAG applications. Are these models suitable for web-scale semantic search, potentially supporting billions of embeddings across a continuously evolving general-web index?
Specifically, I'm interested in their suitability for large-scale crawling, vector indexing, and semantic retrieval, where throughput, latency, and infrastructure efficiency become critical. Are they designed to accommodate workloads of that magnitude, or are they primarily optimized for enterprise knowledge bases?
Cloning existing games isn't exactly creative anyway. Use your imagination, push the boundaries of what AI can do, and build something genuinely your own. Take that original idea, refine it, polish it, and turn it into the next killer hit. AI should be empowering creativity, not just pumping out carbon copies of what's already been done.
Your first customer probably doesn’t care which AI model you used.
They care whether you fixed the thing that’s been wasting their Tuesday afternoon.
That’s a useful reality check when you’re building. 🍁
AI can help you build something fast.
The real test is whether you can still understand it, fix it, and trust it a month later.
Getting it running is a milestone. Keeping it running is the work. 🍁
What’s the hardest part of your build right now?
Well, that didn't last long. Rumours are also swirling that anyone who already used it may have had their reasoning effort levels quietly reduced, potentially causing their tasks to underperform for some time afterward.
More AI agents isn’t automatically better. The interesting question is whether a person can understand what they’re doing, steer them, and trust the result.
An AI agent that looks busy isn’t useful.
Give it a clear task. Let it finish. Show me what changed, what it couldn’t do, and how to undo it.
That’s the kind of “autonomous” I’d actually trust.
@oldstackjournal UCAT is building real technology services, infrastructure, and platform applications designed for the future.
We're not quite ready to pull back the curtain, but make no mistake, big things are coming out of Canada. 🇨🇦
And we're just getting started.