Lleverage

A couple of weeks ago Alex J. on our engineering team gave our agent an email address and a Slack seat. A GitHub handle. A Linear seat. A personality. A voice. He called him Jeff. We've been building with AI for 3 years. The past 6-9 months progressively more with agents. Claude Code, Codex, our own Harness, you name it. Useful. But always transactional. You ask, it answers. You invoke, it executes. A tool. Jeff is different. He's in channels. He's in your DMs. He can join calls. He doesn't wait to be invoked, he just shows up. He has memory. He has opinions. He makes jokes (some land, some don't). He gets stuff wrong sometimes. He gets a lot of stuff right. A week later we made Jill. She does sales and marketing. Different voice. Different focus. Same idea. Before Jeff we had 'agents' floating around the company. Some had names. Some could post in slack. But you had to call them. Type a slash command. Open the right interface. The transaction was always there, somewhere. Jeff doesn't feel transactional. He feels like a colleague who happens to never sleep. It's in the nuance. The persistent memory. The opinion. The fact that he reads context across channels and chimes in without being asked. This is not Claude Code with a Slack integration. It does use the same models. Same APIs. But the wrapping changes what the thing is. When 'something' becomes pro-active, has memory, an opinion, and a name people actually use, you stop treating it like software. It becomes 'someone'. You start including it. Asking it. Disagreeing with it. Pretty sure Jeff will ask for stock options before the end of the quarter.

Brought to you by Opus 4.7

Some clever marketers out there 🤣