I'm regular Cursor user. But it's also no longer independent. Some people are leaving, and those people need to go somewhere.
What moves with you (rules, commands, mcp.json), what doesn't (memories, history, hooks), and how to avoid rebuilding the same lock-in one tool over:
https://t.co/zj1hh36CJw
Cognition's SWE-2 reportedly lands within a point of the frontier at 64% lower cost, built on an open 2.8T base. And it runs only inside Devin*. The engines keep commoditising. The layer that knows your business shouldn't be welded to any of them.
https://t.co/wz457cUQFQ
https://t.co/ROfc2sy9Fk
@mntruell Anyone needing to use any model, Oi is on the Cursor and OpenAI marketplaces (or you can install directly) and letโs you keep contexts, workflows, skills etc in the one place for easy switching.
Donโt let drama like this hinder your progress.
Anthropic shipped an exciting update to claude tag this week. It now reads the whole Slack channel and picks one of four moves: reply, open a thread, route to a workstream, or stay silent. Validating Multiplayer AI as the frontier. It's also only half built.
Multiplayer AI isnโt just multiple people, itโs multiple models. Oi provides shared contexts, governed connections, repeatable workflows and guardrails in a layer you own, available to your creative team on GPT 5.6 and your engineering team on fable. See: https://t.co/LntL13iaPm
Perhaps, but this is where we are now and most companies greatly benefit from having their workforce using the same intelligence layer, anywhere. I agree, it will evolve and change shape (to what, we have our theories), but itโs essential to get better and more consistent results out of LLMโs by design.
@danellisona Weโve found very similar results when using the same context across them, Claude can be slower but ChatGPT sometimes needs an additional nudge. But weโve managed to minimise both to rare occurences.
This was one of the primary drivers to build Oi (healthcare and closing the gaps). Medical practitioners can set up a reusable AI system (context, skills, workflows, connectors etc) and just add the patient to it, who can then can continue to use their own local AI apps however they wish, just with extra power, and communicate to whomever needed.