My time as a @cognition / @DevinAI ambassador was very short-lived. Was released from the program a couple of days ago; still no idea why. But while I was there, I was able to put @DevinAI through the wringer with a bunch of projects. Honest review? It's very close to being a truly remarkable platform. I want to start off with "the bad" so I can end on some feel-good. I don't really like to use "the bad"; it's more personal preference, design intuition/experience, and user preference. I would start off with the "Space" sidebar. Organization is great; I do like that projects are in their own spaces, but one issue I would run into is if I run an agent swarm in a prompt, it will spin up new chats (not in the same space) then it's hit or miss if it auto-deletes them after they are done or not. Some prompts will leave a few hanging sessions; again, not a huge downside (doesn't affect quality of code, which is the most important) Also, lack of easy custom /skill creation is a downer; I have a few that I run in Claude, Codex, Cursor that are more easily added from the app. Again, still able to add yourself to the skills.md file, so nothing that breaks code, just a personal preference gripe. Now the good. Holy shit SWE 1.7 coooooooks. I wish I had a model like it on the other platforms; to me it feels like the perfect middle ground between Composer 2.5 and Grok 4.5. For what I do, it has been flawless. Hella fast, accurate, and its ability to audit code, find bugs, and bolster security is pretty robust.....did I also mention it's VERY fast? Fully open system so I was able to create my own bridge to edit and view my local sessions while mobile as well as my cloud (obviously). Fantastic culture; people like @markfenner have been a huge boon for my knowledge growth. @dabit3 @mattbergland and team are growing something really special. Another pro: I feel like I don't have to do ANY nudging. I'll read the prompt text the AI is putting out, and it's asking the right questions, coming to logical conclusions, and it's really dissecting what I mean and want. I have been pleasantly surprised by its ability to just get done what I needed done. Thanks again to the Cognition team for allowing me to keep this extra usage for the next 6 months; looking forward to continue building with DevinAI Ottermaxxing.