The Two Chatbots That Actually Pull Their Weight
I stopped chasing the best and kept the two that do different jobs well. One reasons, one remembers the web.
After a year of trying every flagship, I narrowed to two and stopped the churn. One is the reasoning-heavy model I use for writing code and untangling a messy plan. The other is the one I use when the answer lives on a page from last week and I need it cited, not imagined.
The split isn't brand loyalty, it's task shape. For 'help me think,' I want the model that pushes back and structures. For 'what actually happened,' I want the one that shows its sources. Using one for both means accepting a weakness on at least half my work.
Keeping two sounds wasteful until you price the alternative: a flagship subscription plus the time lost re-verifying a confident hallucination. The second, cheaper, source-grounded bot pays for itself the first time it hands you the link instead of the guess.
If you're subscribed to one and frustrated, the fix might not be a better one. It might be a second one for the job the first quietly does poorly. Two specialists beat one generalist who's mediocre at the thing you need today.