Ideating with AI can feel like prompt engineering hell.
Today, we shipped Terra v1.0 - our take on a visual, process-based approach to ideation with LLMs.
If you're down to be a super early tester, let's talk: https://t.co/o7cNQswNP7
@help_delhivery my package says it's "cancelled by merchant" and there has been no update for the last 5 days. Your support has no human, cuts the call without any help, and I'm just stuck waiting for an international package that I have no way to track. What am I supposed to do?
@pli_cachete anything with a physical medium (electronics, medical, etc.). the "ship fast" philosophy only applies to pure software companies with low stakes
@archiexzzz Having used GPT-5 in the API, it feels like it thinks of itself as a know-it-all. I found it impossible to steer and it jumps to conclusions even when explicitly asked to "ask the user questions, do not assume" in the system prompt.
@adamsthirdson @WilliamFaix @MaxRovensky Most UI is rendered using efficiency cores on the CPU. GPU is only used for CoreAnimation and accelerations for UIKit. But when every single UI component depends on GPU-accelerated, if not GPU-heavy shaders, it will worsen battery life. More compute = more energy use.
@vitrupo I fully do not understand this take. Existing LLMs already support tool calls, and with CoR, they're more than capable of calling those tools appropriately.
@DylanMcD8 I think the old ones were a lot more charismatic than the new ones. The new ones are boring, but I guess they fit better with the new aesthetic