Prompting vs delegating: a bot has a job, a memory, a conversation and a screen
'In a chat tool you are clicking things. In a bot, the bot clicks things by itself.'
The chat tool returns a paragraph, synchronously, and forgets the thread unless you recall it. The bot 'gives you a changed thing - a document, a report' and 'can work while you sleep' because it runs on its own cloud computer with a persistent role and memory and drives its own UI. 'Every bot has a job, a memory, a conversation and a screen' - the shared anatomy of the whole category. The 'which is better, GrokBot, Codex or Claude Code?' question gets the standing answer: 'it has got nothing at all to do with the tools... test out your tool thoroughly, then pick.' The peers named - Manus, Genspark, OpenClaw, ChatGPT work, Claude Cowork, Hermes, Perplexity Computer - 'belong to the same family: instead of chat, they do work'; the real pillars are data/Excel, documents and slide-type deliverables, and each has a Slack bot you can tag.
Paul already runs this category (Cowork, Hermes); the value is the clean definition he can give a client of what 'delegating to AI' actually means.
Chat tools work while you watch. GrokBot works while you sleep.
Every bot has a job, a memory, a conversation, and a screen.
Answer from memory first — the recall attempt is what makes it stick. Then reveal.
What is the one property that makes a bot 'delegatable' rather than 'promptable'?
Persistence - its own compute, role and memory survive your closing the laptop, so it can carry a task to completion without you in the loop.
