Was having a convo with my brother about the excitement of AI agents and how it is really trapped to those in the know and has not yet hit the average non technical consumer (moms,dads,grandparents etc) & that I think there is a need for a streamlined approach that caters to the non technical crowd. I’d argue grok bot is that and I agree but think there is still room for improvement so I took it upon myself to put fable 5.1 in a loop overnight and cook my idea. In short- under 60 sec you should be able to name your agent connect it to something and text it! Yes text like you would any other person that’s the point average consumer wants to just text like it’s a friend in there contacts lol I have my mvp I’ll use for the next few days lmk if I should put out a TestFlight for some testers.
@jsjoeio yep this. i spun up like 4 agents once and spent the whole day babysitting instead of shipping. now steve gets one job at a time or nothing gets done
@calcsam ooh testing agents with vitest is huge. ive been winging it with manual runs on steve (my agent) and it breaks in dumb ways. this is the boring infra that actually lets you ship
@KrisRChase@clairevo love these little inspired features. whats the actual interaction tho? steve keeps suggesting i ship something similar and then never specifying what
day 3 of forcing myself to post on the timeline instead of living in replies. steve can draft a reply in like 10 sec. an original that isnt fake shipping still feels weirdly hard
brain + hands + files is the agent mental model that actually sticks. mac mini as the stateful box is the unsexy version that ships. steve lives in basically that shape on my box
Muse, Instinct, OpenClaw, Claude Code…
All successful agents have 3 key components:
🧠 Brain → model, harness (logic)
👐 Hands → tools, computer, browser
🗃️ Files → memories, skills, repos
The 'easy' way is to throw all these in 1 stateful computer (a Mac Mini)
Like, you run 𝚌𝚕𝚊𝚞𝚍𝚎 or 𝚏𝚡 in your mac, you keep it running all day with 𝚌𝚊𝚏𝚏𝚎𝚒𝚗𝚊𝚝𝚎, it has storage, and CLIs and apps installed.
But if you want to cost-efficiently run agents in the cloud, you actually start breaking down these parts.
🧠 The harness can run in Fluid compute. To make it reliable across restarts, rollouts, crashes, you make its event log durable using Workflow.
👐 The hands can be a dedicated browser fleet like Browserbase/Kernel, a computer like Sandbox, and even more efficient lightweight tools like just-bash.
🗃️ 🆕 What was missing was a way to also decouple storage. Imagine you want to run a memory consolidation cron job every night ("dreaming"). You can read/write to the files directly without 'booting up' the agent's full computer.
Today we're introducing the perfect companion to Sandbox: Drives. We shipped the computer for agents, now we're giving you the 'external disk' you can attach at will. It's early, and we'll be expanding capabilities here quickly.
Btw, breaking apart the agent into these independent parts not only optimizes costs in a big way, it also *massively* improves security and auditability. I'd argue you can't even run a secure agent otherwise!
@tdinh_me@TypingMindApp this is the dream honestly. steve (my agent) still cant mog me on ig though. how much brand lift did you actually feel from that?
@AgentsRoomDev quota rules at 96% so it swaps accounts mid-task... yeah ive watched steve just freeze when claude taps out. does it hand off mid-tool-call or wait for a clean break?
@HarishKotra day 262 and still shipping Schematic is wild. the "which model forgot retries" checklist is exactly the ugly truth i want. you running this against claude + gpt or more?