MORE OR LESS

@parkerconrad, founder of Zenefits and now CEO of Rippling, joins @lessin and @Jessicalessin. No Morins this week. A year ago VCs were telling him software engineering was months away from mass unemployment. His update: salaries are higher, the market is tighter, and a great engineer is worth more today than ever. He argues the only moat left in AI is seeing the problem before your customer does. That good gross margins can be a sign nobody's actually using your product. And that Rippling isn't a software company at all. @lessin counters with his own setup: a private notes file he hands to four different agents, so they race each other to finish things he never asked them to do. @Jessicalessin brings the mood from New York, where the money people don't think the $8 trillion CapEx bill pencils out, and financing is about to get much harder. Plus Deel fighting a federal grand jury subpoena, Anthropic's IPO date, and Sam's plan to pay people minimum wage with a 5% shot at a million dollars. New episode out now. Topics: 01:14 Parker Conrad joins, no Morins this week 02:13 Everyone watched the election on prediction markets, not TV 04:09 The blue wave, and why the market went up anyway 07:48 Hyperscalers, data centers, and builds that just got harder 09:27 California's wealth tax fails 10:24 Why paper billionaires can't pay a wealth tax 12:17 The financing trick where one investor sets the headline price 13:06 New York's mood: AI financing is drying up 13:45 Sam: we've run out of money in the world 15:09 Parker: good gross margins can be a bad sign 15:30 Anthropic's IPO date and OpenAI's approach to margins 16:44 Rippling's IPO plans and the public market retirement community 18:26 The SaaS apocalypse that didn't happen 19:00 "Rippling is not a software company" 20:29 AI for boring: government forms, faxes and fraudulent audits 25:36 The only moat left is seeing the problem first 28:01 Sam's notes file and the agents that compete over it 29:32 Instinct keeps asking Jess the same question 30:02 Today's agents are shitty assistants 30:49 Permissions architecture as the real advantage 32:21 Where Rippling makes margin and where it makes zero 33:19 Usage pricing and deleting the bottom half of a job 34:36 What you're really selling is a promotion 36:04 Software written at installation time 37:26 Why you shouldn't vibe code permissions 38:07 AI is excellent at using software, humans are not 39:42 Should an AI negotiate your pay? 41:05 Rippling's $17M seed and two years before the starting line 42:07 Why reversibility decides how fast you can move 44:52 The Deel lawsuit and a federal grand jury subpoena 46:06 Sam: your job doesn't look fun to me 48:40 The engineer layoffs that never came 49:25 Random reward payroll, a very bad idea 51:26 Overrated and underrated in Silicon Valley 54:01 Sam's New York Marathon 55:14 Starting a company: the narrow ray of daylight

@parkerconrad on where the advantage in AI will sit. “The most important moat is going to be which companies can have visibility into the fact that there is work to be done for a business.” Right now there’s “this huge gap between the capabilities of these systems and actual uptake.” Then @lessin on his own setup. A private URL to his notes file, handed to all his agents. “They literally every minute or two check it, try to finish things that I haven’t even told them to do and then just deliver me the results back.” On starting a company: “somewhere between completely impossible and very, very, very hard.” “And between those two extremes, there’s this narrow ray of daylight that passes through companies that succeed.” New episode out soon.

@davemorin puts the big AI debate to @bingfish. If AI promises to remove your effort, what part of life gets worse when an agent wins for you? And what do we leave in for the human to do? Bing starts with oysters. “Friction is what makes pearls. If you don’t get sand in the oyster, you just get an oyster.” Sand doesn’t always make a pearl, he says. But it’s necessary to pearl building. So what’s the friction? Constraints. His AI version: ten other people are running the same prompt as you. You’ve got to do something better than them. “What do you do?”

@brit asks @bingfish whether he gamifies his board meetings. “Absolutely.” The rule: the CEO or chairman needs one actionable idea they commit to, per hour. Then a note out within 24 hours to every participant covering what was valuable, who’s taking it on, and when they report back. His measure of whether it worked isn’t the minutes. It’s energy. “A management team goes back to work. They don’t want to go home yet. And the board kind of wants to hang around.” And the best ones share something specific: a risky decision gets made, and you look back later and realize it was the tipping point. He sits on Take-Two and Duolingo’s boards and spent 15 years on Amazon’s, so this is a lot of rooms.

@brit asks @bingfish if he scores his board meetings. Literally grades them and measures each one against the last. He does. His sense is that a founder who loses the board two meetings in a row gets replaced. "If you're wasting the board's time, you're probably wasting your employees' time." Part of the relationship with investors is a confidence game. When a CEO starts mailing in board meetings, it usually correlates to something else. His comparison is the factory floor. Bad board meetings are to CEOs what messy, junk-strewn floors are to manufacturing. Let people keep messy floors, and you'll let them get away with putting a bolt in wrong.

@bingfish on what a wow moment actually is, drawn from forty years of building them: World of Warcraft: the first time you ride a griffon and realize the world is contiguous, that you can see all of it from the air. He compares it to a first balloon ride. John Madden Football had two. The first was parallax scrolling, watching the field move beneath a kick. Nobody had seen that. Doom and Castle Wolfenstein: the speed you move through the world, which was a tech trick. And early Facebook, where the wow was social rather than visual. “Look at these people paying attention to me.” Being tagged. Then photos. Before any of it, the signal was people changing their profile pictures and getting noticed for it, and the team realizing they had to build toward that.

@brit asks @bingfish about the eight steps to gamification in his book. The first one is getting to wow, and his method for finding it is a stopwatch. He walks into a product as the novice. “When I’m talking to developers, I can channel the everyman. I can be the dumb user.” Then he times it. Every time he feels something, he stops the clock, writes down what was on screen and what the feeling was. The feelings sort themselves into a scale. At the top, wow, I’ve got to tell somebody about this. Below that, good, nice, leaning in, feeling warm. At the bottom, the physical tells. Sighing. Brow knitting. Arms crossing. His benchmark from running this on Hall of Fame games: a wow in the first session, and ten goods for every not-good.

@bingfish’s method for backing an idea he thinks will fail: “I don’t think it’s going to work. I’ll back you. You just have to write your letter of resignation and I can pull it out.” The part that makes it work is what happened next. Every single time somebody took that bet and the idea failed, they got promoted. “You’re more valuable after you take a big risk and try stuff, as long as you pay attention to the details.” @brit asks whether anyone actually got fired. One way only. “The only way you got fired is if you blame someone else for the failure.” He puts it to @davemorin directly, as a former CEO. As a founder the whole world is malleable to you, so it is your fault, the way a quarterback owns the defense.

Asked which products today have the Farmville problem, @bingfish offers a test he's run on himself for years. Look at a hundred things in a feed. Count how many you were glad you saw. When it drops under 30%, unsubscribe. What pushes the number down, in his read, is advertising built to look like it isn't advertising. "Putting in not organic ads, but putting ads that kind of seem native. I think it's a risk." @brit takes that somewhere uncomfortable for two companies at once. The usual answer to ads is paying, except her Prime subscription still doesn't remove sponsored products from Amazon search. She's paying and getting served anyway. Then she points at the newest version of it. ChatGPT ends every answer with another question. Here's your answer, now would you like to do this? Now would you like to do this? Same mechanic. New surface.

@davemorin’s question for @bingfish: games make effort meaningful, and AI promises to remove effort. So what part of life gets worse if an agent wins for you, and what do you deliberately leave in for the human to do? “Friction is what makes pearls. If you don’t get a sand in the oyster, you just get an oyster.” He adds the qualifier most people skip: sand doesn’t always produce a pearl. It’s just necessary to one. @brit asks what friction means here. Constraints. And AI generates its own. “Ten other people are doing the same prompt as you. You’ve got to do something better than them.” Which raises the bar continuously. “What we now think is great, three years from now we’re going to think of as slop.”