extremely impressed by grok @bot. random musings on it and some of the competition: - native apps far superior to the imessage (e.g. instinct) form factor. imessage is easier to bootstrap for normies and simpler onboarding, but for any reasonable power usage, a dedicated native app is just better. - multi-bot model (e.g. grokbot) is superior to single bot model (e.g. dots). its a tad more daunting, but creating specialist bots with their own identities/rules/routines is more fun, makes me feel incentivized to invest more in each, and is generally less overwhelming than having a dumpster ground of unrelated topics in one thread. the downside of this model is that i often forget which bot i was talking to about something and revert to my main bot as a router. - other than multibot, some key things that make Grok Bot great: multi-device (ios, mac, tesla), super performant (so fast and responsive), exhaustive ecosystem integrations and mcp support. it just works every time. - claude duet is def on to something with the 'tasks' tab. recognizing todos, structuring them, showing status in a familiar mental model like a todolist, and allowing one-click actions feels very natural and magical. excited to play with it. don't have access yet. - openai shipping dots as a section within the main chatgpt app was a mistake imo. it's out of sight and fighting the existing user behavior / muscle memory of how people use chatgpt (which is already conflicted and confused between codex/chatgpt work & chatgpt). dots should have been a separate app and should subsume the o.g. chatgpt use case. what can chatgpt do that dots can't? there should just be Dots and Codex as two separate apps (this is not a statement on what they should each be named). i can imagine the OAI team faced an extremely complicated set of trade-offs when weighing this, and i empathize, but imo there's a much better end state to be had. caveat that i haven't in earnest tried using muse yet, and I haven't been motivated to invest in porting other than occasionally feeling inspired by @alexandr_wang's shitposting.

I’m glad @strawgate's standards for coffee are lower than his standards for code review because many coffees were brewed by yours truly in the creation of this video. I was thrilled to sit down with Bill, who is the head of product at @pydantic, to hear about how he and the team use @Macroscope. It’s an absolute joy to work with Pydantic and to have a small part in supporting their work.