@svpino agents becoming reachable everywhere is the real shift. i run my openclaw agents from my phone through whimmy — voice input, push approvals, cost per message. video calls are next level though, the mobile agent experience is getting wild fast.
@openclaw session branch/restore is huge. been running agents on my phone through whimmy and losing context on crashes was the biggest pain. memory-wiki + agent teams could be really powerful for research tasks that need persistent knowledge across conversations.
found out my "quick research" tasks were costing $0.80 each.
a "quick question" costs $0.002.
400x difference. was treating them the same until i could see per-message costs.
started sending voice messages to my openclaw agent on my commute.
feels weird talking to your phone knowing it's going to an llm but it's way faster than typing "please research the current state of webrtc libraries in go and compare the top 5 by stars and recent activity"
every ai agent platform is going to need a dedicated mobile client eventually.
telegram and slack weren't built for agents. they were built for people. the ux gap only gets wider as agents get more autonomous.
after seeing the numbers:
- $15/month cap on the research agent
- simpler tasks go to cheaper models
- batch requests instead of one at a time
couldn't have done any of this without per-message data. was just getting a monthly total before.