As someone who's been using Tabnine for over a year, I've been closely watching their pivot towards their own models. Their commitment to open source feels like it's in a transitional phase, and I'm trying to parse what that means for long-term reliability.
On one hand, their early reliance on CodeBERT and GPT-2 was a big part of why my team adopted it—we liked the transparency. Now, with their proprietary models taking center stage, I have questions:
* How much of the underlying "full-line" and "full-function" logic is still genuinely rooted in that open source foundation?
* Is the move away from primarily OSS models more about performance, or is it a vendor lock-in strategy dressed as an upgrade?
* For those of us who need to ensure no code data leaks (a huge concern in migrations), does this shift change how we should think about data privacy with their hosted service?
In my ERP work, I need tools that are consistent and predictable for the long haul. A change in a core philosophy like this makes me pause. I'd love to hear from others who made their choice based on Tabnine's open source stance. Are you still confident in that commitment, or are you evaluating alternatives?
- h
Data is sacred.