Chris Carolan spent the first half of Value-First AI Daily's September 11 episode inside a new agent runtime he had wired to his own agents the night before. He spent the second half inside a product that had already decided, on its own, who from his LinkedIn was worth a message today. By the thirty-sixth minute he had a name for the distance between the two: "the other end of the spectrum."
The End Chris Steers
Chris had hit his usage limit for the second time that week, and Grok Bot, the new runtime, was the fix: one surface, every model on call. He pointed it at his own agent setup, starting with V, the agent he routes his work through, and asked it to bring the rest over. What struck him was how little there was to relearn. "I don't have to relearn a new system very much, if at all," he said.
What he was less sure of was control. Grok Bot keeps its own profile for each agent alongside the originals living on his machine, and midway through the demo he stopped and put the question to the room: "See, do we like that lack of control or visibility?" His own answer, by the end, was that he's probably fine with it, once he understands it well enough to trust it. One test of that came minutes later: when V offered to simply do a piece of work itself, because another agent lacked the access to do it, Chris said no. He did not want that becoming the pattern.
Nico Lafakis was running his own version of the same bet, on a different runtime: a GPT model he'd used all week, further along and faster, by his own account, than the same work would have gone with Claude. His own fix for a model that wanders off plan was to draw the plan out with the stronger reasoning first, then hold the builder to it. "Add follow the plan to the letter," he said. "I can't explain it any better than: you remove the wandering." Two hosts, two orchestration layers, one shared instinct. Steer the tools, and don't let them steer each other.
The End That Decides
Then Chris opened YouSpot, and the show's second half became a different kind of demo. Its home screen carried something new that day: a prompt asking what YouSpot should keep an eye on. He liked the question on its own terms. One of the biggest issues in any piece of software, he said, is whether you actually know what you're trying to do. By the thirty-sixth minute he had landed on the phrase for what he was looking at: YouSpot, he said, is "the other end of the spectrum" from the runtime he'd spent the first half hour inside: one channel that connects everything, instead of a team of agents to orchestrate.
He typed a plainer question: what should I be focused on today. YouSpot answered with a ranked list, built, it said on screen, from his LinkedIn messages and invitations alone, because Gmail and Calendar still weren't connected. Chris hadn't expected it to be that actionable. Nico proposed the obvious next move: have it draft the note it had just recommended, and learn two things at once, whether YouSpot could write like Chris and what it would actually reach for as context.
Episode — Value-First AI Daily
Grok Bot vs YouSpot
Chris Carolan and Nico Lafakis, September 11 — Grok Bot's first night, then fifteen minutes live inside YouSpot, drafting and grading the note in real time (about 61 minutes).
Open the episodeA Message That Sounded Like Everyone Else's
YouSpot wrote the note. It leaned on shared history, the connection's current job, an invitation to meet, every scrap of context it had, packed into six lines. Chris read part of it back and delivered the verdict flat: "OK, yeah, this is like every other AI-generated DM that we get at this point."
Asked a follow-up about the connection's own work, YouSpot backed off what its own draft had implied: "Honestly, less than my draft implied — and that's worth flagging before you send anything." It had built confidence out of a single line on a LinkedIn profile.
Nico turned the moment toward YouSpot's maker directly. "Features aren't a moat," he said. What a drafting tool needs is a skill that learns how its user actually thinks, not one that performs a style on command. It should start unsure, ask what to change, explain what it changed and why, and earn confidence one approved draft at a time. "It's not a competition," he said. "It's a matter of collaboration." Chris named the failure in one line: "It's like, you have all the context. And it's like, as soon it's applying transactional behavior to relationship-based suggestions. That's such a disconnect, right?"
What both hosts wanted from YouSpot by the end wasn't a finished message. It was preparation: why this person, why now, a few ways to open the conversation. That's a smaller ask than the product was built to answer, and a harder one. "So we're going to do our best to record, go live some kind of stuff next week, how that happens to be determined," Chris said, closing the show. "But we're going to get at least one out the door." It took eleven days, and when the show came back on September 22, it never mentioned YouSpot once.
Worth passing on?
