Tech, AI, and NYC: Creativity runs this town, with Dan Shipper, Paul Ford, and Rich Ziade

Summary

Recorded as a live event in New York City, this Aboard Podcast episode brings Paul Ford and Rich Ziade together with Dan Shipper, founder of Every.to, to talk about how the latest wave of AI tools is reshaping software, the people who build it, and the city that builds them. Dan opens his laptop on stage to show off a “shadow team” of agents — middle-manager bots with names like Einstein, Hilbert, and Bernoulli — pushing code to production while he stands at the podium. The agents are there because of what he calls a hard turning point in late 2025: a new Claude model and Claude Code crossed what he labels the “infinite vibe code horizon,” the point where you can keep typing into a prompt and get a real, shippable app without ever opening the editor.

The conversation moves from demo to disruption. Dan describes building Proof, an “agent-friendly Google Docs,” in roughly ten days — something he says simply would not have been built a year earlier because the cost was too high. Paul and Rich pull on that thread: if a CEO can ship a real product solo, what happens to designers, PMs, engineering managers, the org chart itself? They land on the idea of a “shadow org chart” where everyone has a personal agent that mirrors their personality, where designers like Lucas can ship the code that engineers like Kieran hated translating from Figma, and where Slack starts to fill up with bots that take pull requests and answer questions. Rich is more skeptical than Paul about how fast the legacy enterprise will move, arguing that big New York institutions — banks, hospitals, museums, food banks — have a metabolism that resists transformation, and that the right move is to “clean up the old mess before you make a lot of new mess.”

The episode circles around three big themes: leadership (“organizations only go as far as their senior leaders go in AI”), culture (“designers are engineering, engineers are designing”), and pace. Dan defends his optimism with the paperclip-maximizer thought experiment as a counter-example — catastrophic AI fears are overblown, but real new disciplines are emerging, including the discipline of deletion (Anthropic, he notes, tracks “how much have we deleted?” as a product metric). Paul lands a maxim that doubles as the show’s punchline: “Never make any major life decisions 30 days after a meditation retreat or after an encounter with a biotech AI product.” The audience Q&A pushes on the cultural identity baked into org charts, the asterisks security teams attach to every new tool, and whether AI will collapse hierarchies or simply make work expand to fill the new capacity. The hosts disagree productively, the agents keep working in the background, and the New York City map on the screen never quite finds Brooklyn.

Highlights

”We passed the infinite vibe code horizon”

“We passed the infinite vibe code horizon where you can just like keep going and going and going… it doesn’t mean that all programming is totally solved, but the horizon of what you can do without looking at the code has totally changed and then the bottlenecks of what makes software engineering hard have changed too.” — Dan Shipper, 12:00

yt-dlp --download-sections "*12:00-13:05" "https://www.youtube.com/watch?v=jJeS4ULBmV0" --force-keyframes-at-cuts --merge-output-format mp4 -o "jJeS4ULBmV0-12m00s.mp4"

”I just wouldn’t have done it”

“This took me about 10 days… I started this, the first commit of this was like 10 days ago. [Go back one year and do it again?] I just wouldn’t have done it. It’s a very important point to start with: it enables you to do net new things that you would not have ordinarily done.” — Dan Shipper, 17:31

yt-dlp --download-sections "*17:31-18:30" "https://www.youtube.com/watch?v=jJeS4ULBmV0" --force-keyframes-at-cuts --merge-output-format mp4 -o "jJeS4ULBmV0-17m31s.mp4"

”The line goes quiet for two weeks”

“Increasingly instead of pitching, I’ll just make an artifact and throw it over because it’s easier than writing the proposal. And then the line goes quiet for two weeks because their organization can’t metabolize or understand what I just sent them. I’ve sent them working software, I sent them a prototype… they really would have preferred bullet points.” — Paul Ford, 27:41

yt-dlp --download-sections "*27:41-28:46" "https://www.youtube.com/watch?v=jJeS4ULBmV0" --force-keyframes-at-cuts --merge-output-format mp4 -o "jJeS4ULBmV0-27m41s.mp4"

”Engineers are designing. Designers are engineering.”

“An org chart is a power structure. It’s a set of instructions on how people should communicate, how far you’re allowed to go. You’re not allowed to stroll up an org chart and cause disruption in a different column in a different department. What’s so unusual about this stuff is people are leaping out of their own disciplines and messing around in other disciplines… engineers are designing. Designers are engineering.” — Rich Ziade, 33:45

yt-dlp --download-sections "*33:45-34:42" "https://www.youtube.com/watch?v=jJeS4ULBmV0" --force-keyframes-at-cuts --merge-output-format mp4 -o "jJeS4ULBmV0-33m45s.mp4"

”Run forward, break things, go back and fix them”

“All of a sudden, I’m like, oh wow, they’re on the data council, they’re already adopting this, they’re doing this, and they’re like, wait, what, how do we get on it? And I’m like, we asked you and you had like a thousand questions of everything that could go wrong… Run forward, break things, go back and fix them when you need to, but you have to progress or you’re never going to move.” — Rich Ziade, 48:00

yt-dlp --download-sections "*48:00-48:56" "https://www.youtube.com/watch?v=jJeS4ULBmV0" --force-keyframes-at-cuts --merge-output-format mp4 -o "jJeS4ULBmV0-48m00s.mp4"

”Clean up the old mess before you make a lot of new mess”

“Don’t give your whole product roadmap or your life to this new technology. You’ve had robots in here before. They scanned all your invoices, they scanned your digital archives… and they’ve made a mess. Let the new robot clean it up, improve their value. Don’t just give them your whole roadmap. Clean up all of your old stuff, clean up your technical debt, and make it better.” — Rich Ziade, 58:03

yt-dlp --download-sections "*58:03-59:00" "https://www.youtube.com/watch?v=jJeS4ULBmV0" --force-keyframes-at-cuts --merge-output-format mp4 -o "jJeS4ULBmV0-58m03s.mp4"

Key Points

• Aboard’s NYC pitch (0:00) - Paul Ford frames Aboard as services + product helping New York organizations metabolize AI rather than hide from it
• Every.to introduction (4:02) - Dan Shipper explains Every’s three pillars: ideas (newsletter), apps (Cora, Sparkle), and training/consulting for big companies
• Living in the future (5:00) - Every’s job is to “live in the future” of AI changes so they can create “magic moments” that take uncertainty out of adoption for clients
• Shadow team of agents (8:14) - Dan demonstrates a live “shadow team” of sub-agents named Einstein, Hilbert, and Bernoulli pushing fixes to production during the talk
• November 2025 inflection (10:54) - A new Claude model plus a new OpenAI model crossed a threshold where you could build a full app without looking at code
• The infinite vibe code horizon (12:00) - Naming the moment when prompt-only development became viable, and the bottlenecks for software shifted to new skills
• Claude Code as a product bet (14:30) - Anthropic stripped out the editor and made the model good enough that you “just type stuff into this black window and press Enter and it goes and does your bidding”
• Proof in 10 days (16:23) - Dan built Proof, an “agent-friendly Google Docs,” in 10 days; a year ago it would have taken a hire and 2-3 months
• AI enables net-new work (17:33) - The bigger story isn’t acceleration of existing roadmaps — it’s that projects no one would have funded now ship in days
• Meetings obliterated (20:46) - Tools obliterate the ceremony of sign-offs and round trips, disrupting the social dynamics of projects
• OpenClaw shadow org chart (21:32) - Dan’s “always-on” agent R2-C2 mirrors his personality and becomes the go-to for Proof bugs; teammates Slack the bot, not the human
• Shopify’s Slack-first dev (24:00) - Shopify is moving development into Slack: people submit requests, the bot does the work, and only engineers watch the channel
• Aboard’s prototype-first sales (27:41) - Paul ships working prototypes instead of decks; clients sit on them for weeks because their orgs can’t metabolize the speed
• Chris’s 4-week production app (28:46) - Aboard’s Chris and team delivered a production-ready app in four weeks; the client is still reviewing the code more than a month later
• Senior leaders must use the tools (30:13) - Rich’s central thesis: “any organization is going to go only as far as their senior leaders go in AI”
• Spend a day on Claude Code (30:52) - Aboard’s training pattern: a day with the exec team on Claude Code, then top-down rollout works
• NYC tech is in service (32:21) - Unlike SF, NYC technology lives inside other industries (real estate, legal, banking); Bloomberg is the rare pure tech company
• The 10% who already love AI (33:02) - Every big company has 10% who love AI; the change job is “not crushing those people” while you spread their work
• Org chart as power structure (33:45) - Rich on org charts as instructions for staying-in-your-lane; AI is making people leap out of disciplines
• One job: solution provider (34:42) - At Aboard, “there’s kind of one job in the company anymore… we’re a solution provider and everyone is a solution engineer”
• Outside actors cut political rigidity (38:31) - Why orgs hire outside teams: to cut through internal politics. The IT services market is in the trillions; transformation is a sliver
• Lucas does the code now (46:41) - Aboard’s designer Lucas now slings code into Kieran’s app; the move from “don’t touch my codebase” to “great, just sling it in there”
• Identify and amplify early adopters (48:38) - Effective change picks the early adopters who exist already and makes them visible so the snowball starts
• The asterisks problem (48:56) - Security/IT pile asterisks on every new tool; the trick is carving safe pieces (e.g. component systems) where speed wins are easy
• Universal Paperclips & gray goo (50:51) - The paperclip maximizer as a frame for “how much is enough”; Dan and Paul both call the doomsday version overblown
• AI as a deletion machine (51:47) - Anthropic tracks “how much have we deleted?” as a product metric; AI’s value isn’t only adding more, it’s pruning
• Lex, Cora, and the long road to Proof (53:32) - Dan started a text editor (Lex) when GPT-3 launched; Proof began as a Mac app in January and was scrapped to become web-first
• Clean the mess first (58:03) - Rich’s pragmatic NYC strategy: don’t hand the new robots your roadmap; let them clean up the old robots’ mess first
• Major life decisions (59:40) - Paul’s maxim: “Never make any major life decisions 30 days after a meditation retreat or after an encounter with a biotech AI product”
• Dragons at the edge of the horizon (1:00:13) - Dan likens encountering frontier AI to medieval mapmakers imagining the edge of the world: your intuitions about it are going to be very wrong
• AI hasn’t made anyone work less (1:01:00) - “I have not found it to make me work less” — the technology gives capacity, but humans use the new capacity to do more

Mentions

Companies & Organizations

• Aboard (0:00) - Paul and Rich’s company, services + product for New York organizations adopting AI
• Postlight (0:11) - The hosts’ previous agency, mentioned for “Postlight people in the room”
• makeimpact.org (0:31) - Health org client of Aboard, used as a data-visualization example
• Every.to (1:23) - Dan Shipper’s six-and-a-half-year-old company combining media, products, and consulting
• Hidden Door (2:51) - Hilary Mason’s company, a block away from the venue
• Anthropic (10:58) - Maker of Claude and Claude Code, central to the November 2025 inflection
• OpenAI (10:58) - Released a coinciding model that pushed the same threshold
• Shopify (24:00) - Cited as moving development into Slack with engineer-watched bot channels
• Bloomberg (32:21) - Listed as the rare pure technology company in NYC
• Macy’s (38:31) - Anecdote: spent $130M maintaining their old e-commerce platform
• Accenture (41:58) - “AI innovation lab at Accenture” as the kind of partner clients no longer want
• NY Food Bank (57:27) - Example of a vast NYC institution that also serves as an educational hub for other food banks
• Museum of Natural History (57:27) - NYC institution with a graduate program embedded inside it

Products & Technologies

• Cora (4:30) - Every’s AI email management app
• Sparkle (4:30) - Every’s AI file organization app
• Lex (53:32) - Every’s GPT-3-era AI text editor, the original AI-in-writing experiment
• Proof (16:23) - Dan’s 10-day app: agent-friendly Google Docs; live syncing, sign-in with Every, library, sharable docs for AI agents
• GPT-3 (10:58) - Origin of the public AI conversation
• ChatGPT (11:30) - The moment AI entered public consciousness
• Claude / Claude Code (10:58) - Anthropic’s coding agent that crossed the infinite vibe code horizon
• Codex (30:52) - Mentioned alongside Claude Code as a training option for orgs
• OpenClaw / Open Cla (21:32) - “20% always-on agent” service Dan uses; his is named R2-C2
• R2-C2 (21:32) - Dan’s personal OpenClaw agent that fields Slack pings about Proof
• Slack (22:00) - Every’s collaboration tool, increasingly populated by agents
• Figma (47:00) - Mentioned as the design-to-code translation that engineers used to dread
• Universal Paperclips (51:00) - The browser game / thought experiment Dan invokes for AI maximalism
• Shell Game (podcast) (24:17) - Evan Ratliff’s podcast about running a startup with AI agents, recommended as an entry point

People

• Dan Shipper (1:40) - Founder/CEO of Every.to, on stage as the guest
• Paul Ford (0:00) - Aboard co-founder, host
• Rich Ziade (1:13) - Aboard co-founder, host
• Hilary Mason (2:51) - Hidden Door CEO, present in the room
• Katie Perry (5:35) - Every journalist, cited for writing about Bible study and AI
• Evan Ratliff (24:17) - Host of Shell Game, Dan’s recommended on-ramp to agentic AI
• Kate (9:29) - Every’s editor-in-chief, fielded the chaos of Dan shipping production code
• Willie (22:24) - Every’s head of platform, the human who pinged R2 about Proof outages
• Chris (28:46) - Aboard’s head of solution engineering / delivery
• Tracy (42:51) - Aboard client CEO who mandated AI-driven data transformation across her clinics
• Claire (43:09) - Aboard client team member, briefly missing from the room
• Kieran (47:00) - Every engineer who runs the Cora app
• Lucas (47:00) - Every designer who now ships production code

Surprising Quotes

“Six and a half years old with a mustache.” — Paul Ford, 1:42

“I’ll step away from the laptop. They’ll keep working without me, they don’t need me.” — Dan Shipper, 10:39

“It’s a warning to everyone — all of your technical leadership is doing this and not telling you, and it’s coming for you too. Like this is real.” — Paul Ford, 20:41

“Willie is a human? Willie’s a human. Seriously. And R2 is an agent.” — Rich Ziade & Paul Ford, 22:42

“It’s a great game. It’s true. Just go play Universal Paperclips. … If you want to know why you can’t stand the people who run AI companies, know that this is their favorite thing in the world.” — Dan Shipper, 51:00

“Never make any major life decisions 30 days after a meditation retreat or after an encounter with a biotech AI product.” — Paul Ford, 59:40

“All right, so everybody gets to work all the time forever. That is a really great vision of the future. Thank you.” — Rich Ziade, 1:02:22

Transcript

Paul Ford: 0:00 So, Aboard is a combination of services and product for, which is a theme you’re going to hear a lot when you hear about Every.to. And, we are someone where if you come - we used to be agency people. This is my co-founder, Rich, who will introduce himself in a second. My name is Paul Ford, and we used to run a shop called Postlight in the neighborhood. There’s some Postlight people here, which is really lovely to see. And, yeah. Everything has changed in software in a really complicated and sometimes unsettling way; so we’re here to help New York in particular move along in this new world and kind of get the value. Not hide the value and not pretend that things used to be the - the way things haven’t changed, but really bring this change to people into organizations in an organized way. It’s lovely we have some clients here today who we’ve been able to help. It’s a health org who have been able to help with lots of sort of data visualizations, and I think it was stuff that was out of reach - makeimpact.org, you should get to know them. Makeimpact.org. It’s stuff that was out of reach with traditional development practices, but it’s in hand now. And so we’re here to kind of deliver that. And so if you want to talk about that, there’s lots of Aboard people around the room you can talk to, to me or Rich. But do you think that was a good encapsulation of the offering?

Rich Ziade: 1:13 That’s marketing.

Paul Ford: 1:14 There we go. Well done. Uh, but yeah, we were going to - we have like a video but then we were like, ‘Oh my god, no.’ No.

Rich Ziade: 1:21 Just the 12-minute video.

Paul Ford: 1:23 Yeah, you know, just like ‘la-la-la-la-la’, screens moving by. So we’re not going to do that. But tonight is a special night because there’s a new organization in town. Uh, it’s called Every.to. I guess it’s not that new. How old?

Dan Shipper: 1:40 We’re six and a half years old.

Paul Ford: 1:42 Okay, so there’s a six and a half year old with a mustache. Okay. So I’m going to just - I memorized this introduction. And, so anyway, Rich and I have worked together for a long time. You want to just do a quick intro of yourself?

Rich Ziade: 1:58 Sure. I’m Rich. I’m Paul’s co-founder here at Aboard. It’s nice to see - this is the most attractive group out of the last few events. So a round of applause for yourselves.

Paul Ford: 2:13 Honestly, the healthcare event was strong.

Rich Ziade: 2:16 Paul, come on! Stop it. They look like doctors too.

Paul Ford: 2:20 Yeah, every guy had a sweater. Sorry. You guys look great.

Rich Ziade: 2:24 New York City’s the greatest city in the world. We built the business - I built a couple of businesses here in the past. Uh, and it’s always a little late to the party. So I’m glad to be here, I’m glad to have this conversation, and I want to introduce Dan.

Paul Ford: 2:39 Dan, tell us a little bit about yourself and about Every.to.

Dan Shipper: 2:41 I actually wanted you to continue with your introduction of us, because I would love to hear what you think, and then I’ll tell you what it is.

Paul Ford: 2:51 So six and a half years old. Actually, Hillary Mason is here from Hidden Door, just a block away, and we did the same thing on our podcast today where I told you exactly what was… But I thought your company was and you corrected every single… …over. For the first—first crash coming in…

Dan Shipper: 3:07 I really like every—every dot T-O. Every dot to is our domain name. It just—just every. It’s cleaner now.

Paul Ford: 3:13 Okay. So we go Every. Okay. So here’s Dan Shipper from Every. Six and a half years ago, it was a dark night. And— And it was—Dan was wearing—uh, he was like—he’s not just handsome, he’s also intelligent, and he said to himself, ‘I have had a recent entrepreneurial success and I want to do something very new and very special and there’s so much change, so much happening.’ So I’m gonna create a new kind of company that is a combination of media, when people need to understand what’s happening in this changing world of software; product, because we’re gonna ship and create products that people are actually very passionate about and like to talk about; and also, you know what, everybody keeps asking us what to do with all this stuff, so we’re gonna spin up a consulting arm as well. And so that is Every. It’s a relatively small but very kind of impactful org that’s also very transparent and open. Uh, and it’s a good paid subscription product if you’re trying to just wrap your head around like what the hell is new now? What did they do? And so—

Dan Shipper: 3:58 That’s perfect.

Paul Ford: 4:00 All right. Here we go.

Dan Shipper: 4:02 Um, yeah. Uh, Every is the only subscription you need to stay at the edge of AI. We do three things: ideas, apps, and training. On the ideas side, we have a daily newsletter. We write about all the new models when they come out, we get our hands on them early, we do in-depth reviews on whether they’re useful, whether you should use them for all that kind of stuff. We also have a suite of apps that we build that are AI apps, so we have an app called Cora, which helps you manage your email with AI, we have an app called Sparkle, which organizes your files with AI, that kind of stuff. So we build them for ourselves and then we launch them to the audience. And then the last thing is we have training. So we do courses and consulting with like large companies to help them implement AI. I sort of think of it as—or a good model for it is—there is a—extremely important technology change currently happening that a lot of people are very afraid of. And our job is to live in the future of that change, to like actually have our hands in it all the time every day, so that we understand from a first-person perspective how it impacts our work and how it impacts our lives. Like we’re not theorizing it, we’re just living it. And then our job is to make magic moments for people so they sort of get it and they’re like, ‘Ah, this is what it looks like.’ So it takes away a lot of the uncertainty and then bringing them further in once they are AI-pilled.

Paul Ford: 5:35 Yeah. There’s—a good example, like one of my favorite pieces of content, I think Katie Perry writing about bible study?

Dan Shipper: 5:40 She does write about bible study.

Paul Ford: 5:42 And so like, it’s just not what you’d expect. Every—Every has this wonderful journalist Katie Perry who writes about sort of how to really use this stuff. And she wrote about how—and I don’t think she’s a religious person anymore—but she still is interested in historical text analysis and she sort of talked about her whole process and how it— And so it’s a very kind of non-orthogonal way to see this that’s very much in contrast to maybe the Valley narrative, which I like. And it also kind of ties back to this being so New York-y, right? It’s a media company. You’re more a little more worried about the reader than the user. And that’s cool.

Rich Ziade: 6:17 Yes. I like that. That’s very real.

Paul Ford: 6:20 Should hire Paul.

Dan Shipper: 6:22 I hear you’re doing consulting?

Paul Ford: 6:24 Yeah, no I am. Tell me about that. Especially in comms, not tech. Um, okay, so what I wanted us to do, I think Dan, you being younger and more attractive, we’re going to set you up and we’re going to have you do a little… move that slide forward so we can get to… so I’m, oh yay, we’re in New York City everybody! One more, one more. That’s right, well it’s our kind of town. So we’re going to, we could see you do it like it’s got, you know, you can’t, you got to do it. So what I think Rich and I want to quiz you about for a minute is sort of, and to get everybody’s set up, just for like 5-10 minutes talk about what changed and how to kind of articulate that change and what changed in sort of how we’re building code and how we’re building products and platform and design and sort of all of our worlds feel very blurry around the edges including ours. Like I have to say as an AI company I feel that I keep getting smacked across the face by the rate of change. So it’s not, it’s not just sort of like somebody out like a college student who’s been up at an SF. Like this is a confusing moment. So I think, you know, in around November things started to change and I kind of want you to narrate that. I’m going to switch to your screen so you can show us some stuff. And then maybe Rich and I can quiz you about what you’re doing and then after that we’ll talk about what it’s doing to New York and what we want to have it do to New York.

Dan Shipper: 7:45 Please. So, switching to my screen. One sec. Technology, one thing AI has not solved yet is the HDMI… yes, so we’re watching everybody. One more. VTC system. There we go. Look at that! So can you read this?

Paul Ford: 8:11 There’s an ‘E’. Yeah, so the big ‘H’, so maybe…

Dan Shipper: 8:14 Uh, this is, so Every is a pretty small team, we’re about 25 people now. Um, but this is, I also have a little shadow team which is my team of agents that’s currently working on one of our products. And um, I’ve got one agent on this tab over here. This agent is getting a bunch of new code up into production, up into our server while we’re, while we’re chatting. And it has a sub-agent that it’s managing, so we have middle manager agents now. It has a couple of sub-agents, which you can see they have names like Einstein and Hilbert and Bernoulli.

Rich Ziade: 8:51 Dan, what are you doing?

Dan Shipper: 8:53 So like I said, this one is trying to get a bunch of code into production. Um, this one is… The names are actually really bad right now. But this one is… I launched an app last week that I built on the side, and that’s like totally gone crazy that I could even do that, and I think that’s part of what has changed. And I auto-coded it and we launched it, and it was going really well and it just went down, and it was just… it was just crashing. And I was like, this is a very complicated app that I don’t understand at all.

Rich Ziade: 9:23 I think it’s great for everyone else in the company that the CEO can ship and build apps on their own. I think that’s really excellent.

Dan Shipper: 9:29 Yeah, I did get the side-eye of like, ‘Should he really be doing this?’ Kate, our editor-in-chief, is back there and she… yeah, we had a little meeting about the chaos that I caused. But my agents are fixing it.

Paul Ford: 9:41 Did you ever have—

Dan Shipper: 9:44 So this one is… basically, we have a line of work narrowing down all the reasons why it has been crashing over the last couple days, and this one is basically pushing that line of work forward. And each of its sub-agents has a line of, like, a particular class of bugs that it’s working on or reviewing to get to production.

Paul Ford: 10:10 Let me pause you and bring you back to earth, okay? Because… no, seriously, because I’m in this world too.

Rich Ziade: 10:17 Let me jump in real quick.

Paul Ford: 10:21 Not everyone here is a technologist, right? And so when we say ‘what changed’, what the hell are we talking about, right? What we’re really talking about is—

Rich Ziade: 10:30 The app wasn’t doing this before November.

Paul Ford: 10:33 Okay, so again, let’s get into the mind of someone who’s not a coder.

Dan Shipper: 10:36 Yeah, I’m happy to. I was just… I was just proud of my agents. I’ll step back. I’ll step away from the laptop. They’ll keep working without me, they don’t need me.

Paul Ford: 10:42 Are you with us in the room?

Dan Shipper: 10:44 I am. They’re right here, yeah.

Paul Ford: 10:46 You mentioned November 2025.

Dan Shipper: 10:51 Twenty… yeah, 2025.

Paul Ford: 10:54 Okay. Walk through what happened in October, November 2025.

Dan Shipper: 10:58 So I think… obviously we’ve been talking about and thinking about AI for probably three and a half years now since GPT-3 came out, actually really probably for most people since ChatGPT came out has it sort of started to enter the public consciousness and become a thing. And I think that was a moment where everything changed. And then there have been a couple moments since then where everything has changed, and November was a big one. And in November what happened was, or November or December, somewhere around there, a new model came out from Anthropic, a new Claude model, if you’re familiar with Claude. And simultaneously OpenAI also released a model around that same time, and both of them reached this interesting critical point where we started to get to a place where you could build a full app without really looking at a line of code. You can just keep going, you just type into your little— you know prompt a of a chat window over here and what what used to happen before this is if you were doing this and you were not technical you didn’t really understand the code you could actually do, you know, six months ago or a year ago you could actually do like a really like a landing page or like a very basic app, but very quickly it would start to fall over and it you wouldn’t be able to keep going unless you really took the time to understand okay what’s happening and let me let me push the models. You really had to be technical. In November we reached this this point of like we we passed the infinite vibe code horizon where you can just like keep going and going and going and it doesn’t mean obviously like this app is not perfect like it went into production and it started going down. So it doesn’t mean that it all programming is totally solved, but the horizon of what you can do without looking at the code has totally changed and then the bottlenecks of what makes software engineering hard have changed too because there are there are now new bottlenecks and there are now new skills that you need in order to build software applications.

Rich Ziade: 13:05 I think what was wild about the moment too is that in some ways it wasn’t an exponential change in the model like they released a lot of really good new things in their new, you know, it went from Opus what four to four point five or it was like a jump. But it wasn’t like oh my god it does something radically different. Claude Code is just built on top of it in a really smart way and in a very knowing way kind of as a product and so it got better the the underlying technology got better but the technology built up from it got smarter and you’re watching these sort of like words just popping up behind behind Dan on the screen, right? And that is the agents in these sort of loops where they’re evaluating their own work.

Paul Ford: 13:46 I mean actually describe what is happening at any given moment.

Dan Shipper: 13:52 And I I will say also I think one other simultaneous change which we haven’t talked about is that every three years or every year there’s something new and the the new thing that happened a year ago that started this whole thing is is this thing called Claude Code. And if you don’t know what Claude Code is you can you can think if you if you know, you know, uh when hackers are like in the movies and they have like that black screen open. Uh that’s that’s where Claude Code runs.

Paul Ford: 14:08 Is there an organization that offers like a real beginner’s lesson on these things?

Dan Shipper: 14:11 We have a Claude Code for beginners course starting uh in a couple weeks I think, yeah.

Rich Ziade: 14:25 Those marketing newsletters are working hard.

Dan Shipper: 14:30 Um, so it’s it’s like where it’s where hackers work and what’s what’s really interesting is programming on a computer it used to be sort of like typing into a text editor. So but instead of a word document it was like a file that had a lot of code in it and it was just a programmer looking at that code and then running the code and then finding issues and then fixing the issues. And when we first added AI to programming, the logical thing to do was add a chat on the right side of that screen. So you still have your code editor and then you have like a little chat. Actually, even before that it was like an autocomplete like it completes the line. Then we went to there’s a sidebar with a chat and you can kind of like ask questions but the—the AI’s kind of in the passenger seat. It’s something that you turn to when you need to but it’s not something that you actually are really delegating work to. And what was—was really interesting is that Anthropic figured out how to build this thing Claude Code where you get rid of the editor. There’s no word processor at all. It’s just—you just type stuff into this like black window and press Enter and it goes and does your bidding. Um and there’s a lot of technical reasons why the thing that they did ended up being—being a really interesting way of doing this but like they—they made that bet that you can now just—the models are good enough that you can now just type stuff into it and it’ll figure stuff out on your computer and do work for you. And that is—that is a really important point that then led to this next point which is it can go forever without a lot of intervention.

Rich Ziade: 16:05 So I want to—I want to bring this home before we start talking about New York City and ask you to do this. I want you to tell me a thing that you’ve built since November, and then I want you to tell me how long it took, and then I want you to tell me how long it would have taken you.

Dan Shipper: 16:23 So, um, so this is Proof. This is the thing that I built. Uh, this is the landing page. It’s very beautiful. I’m very proud of it. Um and then this is a Proof document. So you can think of Proof as it’s basically just Google Docs but it’s very agent friendly. So you can have your agent in there and it’s a really good if you’re—if you’re using agents a lot you know that they produce a lot of files and that it’s very hard to share those files with the team and Proof is like a really, really easy way to like just spin up your agent writes a document, it comes into, it’s in your web browser, you can share it with other agents, you can share it with your team. So I built this. I technically started Proof in early January but the—the it was—it started as a Mac app and I ended up throwing out the Mac app. So this whole thing, this whole app which has a lot of like interesting features, it has a sign-in with Every, it has a—it has a live sync. It’s totally broken right now as you can see by the red thing but I think we just pushed some changes I hope. You can add an agent, you can share, it’s got a lot of like really cool things, it’s got a little library feature. Um, this took me about 10 days. Like I started—I started this, the first commit of this was like 10 days ago.

Rich Ziade: 17:29 So go back one year and do it again.

Dan Shipper: 17:31 I just wouldn’t have done it.

Rich Ziade: 17:33 But let’s say… let’s say it’s not… no Dan, you gotta do it, I got the money for you, your team’s gonna do it.

Dan Shipper: 17:37 I see. Um, but I think that that’s a very—it’s a very important point to start with is like it enables you to do net new things that you would not have ordinarily done.

Paul Ford: 17:48 Yep.

Dan Shipper: 17:49 Because I think there’s—there’s all these interesting questions about how does this affect jobs and how does—um, how does it affect like how companies are structured? And you can’t—you can’t tell how it’s going to change things until you factor in that there’s a whole lot of things… bunch of work that people are exposed to that they’re not doing because it would… it wouldn’t be worth the cost.

Paul Ford: 18:06 I mean also every company has a 15-year roadmap, right? Like, they should be picking things off the 15-year roadmap.

Dan Shipper: 18:11 You could just accelerate it. What was… what did you ask me?

Rich Ziade: 18:16 How long it would have taken you back in November?

Dan Shipper: 18:20 Uh, how long it would have taken? I mean, we would have had to… we would have had to hire someone full-time to do it. It would be one person. This is a year ago. It would be one person, and, uh, it would have taken them, uh, a couple of months, three… two or three months.

Paul Ford: 18:41 So they would have still been using a year-ago AI, they would have been still doing the some bad coding. Yeah, so that’s not even like non-accelerated, right? And I think this is real. That’s why November and that’s why this stuff is so disruptive is because it went from like, ‘Hey, we’re speeding stuff up, we’re helping you along, we’re figuring it out, it’s getting better and better’ to ‘No, it took me ten days, he’s the CEO, we’re gonna just get it into production, let’s see what happens.’

Rich Ziade: 19:05 Yeah. I mean let’s talk about the other roles that are typically involved with stuff like this. There’s some of them are in the room. Um, this is, I’m looking on the… well, you changed the screen. No, it’s okay. Um, it’s pretty nicely designed. Uh, you did this in ten days. Did you go to a designer at all?

Dan Shipper: 19:25 We have in-house designers.

Rich Ziade: 19:27 Okay, did they give this a look?

Dan Shipper: 19:28 They give it a… a little polish. I did the… the editor UI, but they did the landing page.

Rich Ziade: 19:36 So I mean the classic process isn’t just pure engineering. There’s design, there’s product management, there’s maybe some technical discussions around what the architecture should look like. Plus front-end and back-end and the back-end and whatnot. So when you say three months, that’s one solo person but you’re going to use some other people that are involved.

Dan Shipper: 19:54 Yeah, they’re going to… they’ll… pull in a freelancer to do their… do the landing page or do the template or whatever.

Rich Ziade: 20:00 Right, which you, you know, which takes time and you’re waiting for that turnaround and you’re going to make some tweaks and you’re going to round-trip. I mean… talk about that. Talk about what happens to all the other roles in this new world. Did you… I mean, you… you did this in 10 days. Did you run other people for… to answer questions as you’re doing it or did you kind of make decisions on your own?

Paul Ford: 20:23 CEO guy.

Rich Ziade: 20:28 Yeah, I figured.

Dan Shipper: 20:31 I mean, I… I like asked people for like a lot of feedback and by the time we launched it, the whole team was using it. Yeah, so I had a… I was pretty confident it was good. Yeah.

Paul Ford: 20:41 And it’s a warning to everyone… all of your technical leadership is doing this and not telling you, and it’s coming for you too. Like this is… this is real. Well, I mean, I think one of the things to highlight uh around this is the… the friction around people coordinating. I mean, the meetings that have been obliterated by these tools, not just the ability of an engineer to trust agents and go, is something…

Dan Shipper: 21:00 The- a lot of the ceremony around these things, sign-offs and round trips and whatnot, it’s very disruptive to the social dynamics that are typically associated with projects, right?

Rich Ziade: 21:15 Yes, it is disruptive and there’s all this-

Paul Ford: 21:19 How’s morale in your company? Give us a- out of ten morale. That’s just a plant.

Dan Shipper: 21:32 Um, well what’s really interesting, so- so I have this- I don’t know if you all know what an open cla is? It’s like a 20% always-on agent. That’s all you have to know. You can talk to them on Slack. Mine is named R2-C2. You can give them fun names.

Rich Ziade: 21:58 Is this your org?

Dan Shipper: 22:00 Yeah, we’ve got a lot going on. I’m just in every channel. We just moved to Slack, so I’m like still figuring it out.

Rich Ziade: 22:16 There is a corporate coach in here right now.

Paul Ford: 22:18 Targeting you.

Dan Shipper: 22:24 So what’s really interesting about- this is where we do the proof stuff. And you can see we’ve got Willie who’s our head of platform who’s not working on proof because he’s working on something more important. He noticed that proof wasn’t working and he just tagged in Slack, ‘Hey R2 like these docs are not working.’

Rich Ziade: 22:42 Willie is a human?

Paul Ford: 22:43 Willie’s a human. Seriously. And R2 is an agent.

Dan Shipper: 22:50 Yeah. And that’s actually really interesting because there’s all these questions about okay, do I have my agent fix the problem because all the agents are capable of doing it. Whose agent gets to fix it? And then the etiquette of like, are you gonna call my agent? Like what can you talk to mine? You just added him, you know? So there’s a lot of interesting etiquette. What we’ve landed on, what’s really interesting is if you are in the org and what ends up happening is everyone had every more or less has one of these and they end up sort of mirroring their personality and what they’re good at because you are using it all the time to be good at that particular type of thing. And what ends up happening then is you create this like shadow org chart where everyone has a thing that’s good at their and an agent that’s good at their kind of thing and then they get used inside of the organization for that. So R2 is very close to Proof because I’m working on Proof all the time and so R2 becomes the go-to source for I have a question about this, I want you to fix this bug.

Rich Ziade: 23:47 So it’s literally a program manager? Like it is the program that is a manager. And R2 is OpenCall?

Dan Shipper: 24:00 Shopify has a whole big thing. A lot of their development is now happening on Slack. They submit requests, the bot goes away and comes back and only engineers are watching that. So there’s definitely some of that in our future. So I’m going to… let’s come back to New York and sort of try to bring sort of all of this back. Cuz I think so for, for those, if you’re looking for like a gateway to this, if this is like a little too technical, the Evan Ratliff podcast Shell Game 2 is him kind of building a business purely with agents and it’s like genuinely hilarious. It’s just him arguing with voice agents all day. And it’s actually just a good way to kind of get the patterns in and then really what everyone probably should spend a couple hours in the Claude room in 2026. It’s very different than the normal chat experience. There’s also this site called Every that you should really check out. Everybody should, everybody should already be subscribed.

Paul Ford: 24:57 That’s true. Oh wait, no we’re back. Rich said it was on the way.

Rich Ziade: 25:00 Sorry, it’s my fault. I just committed it.

Dan Shipper: 25:04 Let’s see. Sometimes also Samsung just likes to update like mid-meeting. You know? All right. It’s okay. I have a laptop too. Okay. So we made a map of New York City and here’s how I kind of… I would love this to go. We can talk, or we can kind of riff on the different industries.

Rich Ziade: 25:28 Where’s Brooklyn?

Paul Ford: 25:30 I don’t see my office.

Dan Shipper: 25:32 Yeah, yeah. No look, it’s the same for all of us. It just doesn’t make it on the map.

Paul Ford: 25:36 Hold on, Dan wants to tell us where Every is on the map. Let’s give him a few minutes.

Dan Shipper: 25:41 You go to Atlantic Avenue…

Rich Ziade: 25:45 My agent is on it.

Dan Shipper: 25:46 And now you have about 15 minutes to get really confused. Um, okay, so…

Paul Ford: 25:52 Let me set this up.

Dan Shipper: 25:54 Well okay, okay but also what I just want to say really quickly like we can talk about all of these things in front of you or we can just get the Q and A started earlier because people are from these industries and we actually, all three of us are loud tech talkers, we’d gladly, we’d rather talk than monologue, not knowing it’s one of us monologues. Okay, go ahead.

Paul Ford: 26:18 So I mean, a good way to approach this is is I think what’s good about what Dan just introduced to us is he fenced off a playground where he does not have the burdens of legacy of the way we handle accounts receivable, he doesn’t have any of that baggage, right? And and the culture is is clearly allowed to run and embrace all these ridiculous tools like their half his Slack is is robots. He’s going to promote a robot ahead of a human at his firm. There’s no doubt about it. There’s a VP title coming.

Dan Shipper: 27:00 And you know as ex-agency guys here, when you go into organizations and you’ve got all this amazing capability and technology even predating AI, you slam into the patterns and muscle memory of those organizations in such a profound way that you find yourself changing your language, changing your approach of how you’re going to interact with them, how do you get them comfortable with change, like dramatic change, potential. And now we’re here and the change is so dramatic that as a tech company, Aboard is dealing with radical change.

Paul Ford: 27:41 I’ll tell you something which is increasingly instead of pitching, I’ll just make an artifact and throw it over because it’s easier than writing the proposal. And then the line goes, and these are to people I know well, I’ve explained what’s happening, the line goes quiet for two weeks because their organization can’t metabolize or understand what I just sent them. I’ve sent them working software, I sent them a prototype, I sent them a tool or something kind of deep and dense. And they really would have preferred a bullet points, but they’re also coming to us to help them with this change. I don’t want to protect them from it, I want to show them where we’re headed. And so I find this really, really complicated because I probably could close more business if I just sent them a PowerPoint deck with a number in it. But I actually don’t want to do that right now. I want to drive this change into the org and see how they respond to it. And that’s not just me, like many of our projects are going this way. We’ll ship something and they will sit on it and we’re used to this from the old days, but they’ll sit on it for like a month because the org just can’t handle that it happened so quickly.

Rich Ziade: 28:46 I mean, Chris landed a, Chris and team landed a project, what was it, four weeks ago? Three, four weeks ago?

Dan Shipper: 28:58 Tell everybody who Chris is.

Rich Ziade: 28:59 Chris is our head of solution engineering, solution delivery at Aboard. A round of applause for Chris. Rockstar! He loves attention, he’s so bad at it. They were banging away at this problem for months. We came in, we delivered a production ready app in four weeks. They are still, it’s in prod, but they’re not, they’re still reviewing the code, I forget. I think you were still having, they’re still processing this change. They’re reviewing all our code and I think testing it very modestly because it’s such a dramatic thing to just parachute in and just use it and trust it. And meanwhile we’re on the hook too, we had to test it. Like our efforts were, and Paul’s, Dan gets roped in, your org’s roped in to consult in big orgs about this change, right?

Dan Shipper: 30:00 For stuff like that too, is they they kind of want to keep you a little bit in a bubble as a change agent, but they don’t necessarily want to give you all the keys just yet, right? Like, how are you finding these engagements? How are you finding people reacting to this much change?

Rich Ziade: 30:13 We’ve done a lot of different things and, you know, we do a lot of work with hedge funds, PE firms, big companies, tech companies, all that kind of stuff. And so it has changed a lot, but our bread and butter is basically like training. We’re going to go and teach you and what has really started to, what what is really clear to me is that any organization is going to go only as far as their senior leaders go in AI. So, it’s not a thing where you can be like, yeah, someone else is going to figure that out. If you really want your company to like live in this world, the senior leaders have to know what they’re talking about and have their hands in it.

Dan Shipper: 30:50 Do you find them delegating AI a lot?

Rich Ziade: 30:52 Yes. Okay. Um, well, not mostly not our clients because what what we do is we will go in and spend a day with the exec team and teach them how to use Claude Code. And once they have a day on it, they’re like, okay, I kind of get a sense for what is now possible and what my expectation should be. And then you can go to the rest of the org and do Claude Code trainings or Codex trainings or whatever you want to do and the, because it’s flowing down, it’s much more likely to work.

Dan Shipper: 31:19 What is the misunderstanding you find yourself clearing up the most? The thing that they expect… no, no, take a second. Take a second.

Rich Ziade: 31:28 That’s a, that’s a good question. I, so I am not as involved day to day in the consulting as I used to be, so I’m not talking to clients all that often.

Dan Shipper: 31:41 They’re making apps on Claude Code! I just left for a week!

Rich Ziade: 31:52 Um, I don’t know. But I do, but I do think, not to, not to like hammer that home too much, but I think that that is the thing is like, when companies are like, yeah, the CEO said we’re going to have an AI project then, but they’re not doing this and they expect the entire organization to just figure it out. That’s the thing that I think needs to flip into if you really want to do this, you have to really understand it from the inside, which means you have to have your hands in it.

Paul Ford: 32:21 I mean, I think this is a weird thing about this city, right? So we tend to see technology as a service inside of all these other industries: real estate, legal. And if you don’t have any questions, we’re going to start talking about legal, so you better get your questions in. Um, but they, you know, so we’re, like, even at the big banks, right, there’s huge tech orgs as big as could be, but we don’t have a lot of pure technology companies. Like Bloomberg is actually a pure technology company, we don’t even think of it as one. So, but we’re in service. And I think what’s different is like then you have to go ask leadership to like take this very seriously and actually think like technologists, or otherwise, um, this is going to, this is going to pass them by.

Rich Ziade: 33:00 They’re going to… they’re going to struggle with it.

Dan Shipper: 33:02 I think that there’s also… there’s this dynamic in big companies where because of all of the rules and all the processes, any big company has like 10% of the people at the big company loves AI and is like wants to experiment all the time and probably has Claude on their phone and is allowed to use it. And so part of the whole process of change is actually just not crushing those people and allowing them to figure out how we use this inside of our organization and then spread that as much as you can. And it always runs into, okay, how can we do this with IT and what’s with safety and privacy and all that kind of stuff, but actually just unlocking the resources you already have and the excitement that you already have is a big thing.

Rich Ziade: 33:45 I think… I think when you look at an org chart, an org chart is a power structure. It’s a set of instructions on how people should communicate, how far you’re allowed to go. You’re not allowed to stroll up an org chart and cause disruption in a different column in a different department. That’s “stay in your lane,” right? That’s what an org chart is, right? And what’s so unusual about this stuff is people are leaping out of their own disciplines and messing around in other disciplines. People who should not be drafting a contract are like, “You know what? I can’t wait for legal anymore. I think I can put something together.” And that’s great! I mean, engineers are designing. Take that in for a second. Designers are engineering.

Dan Shipper: 34:40 That’s a big thing for us.

Paul Ford: 34:42 Wait, we actually… there’s kind of one job in the company anymore for us. We’re solution provider and everyone is a solution engineer. Including me, including Rich. Like, you’re just kind of on the hook to deliver.

Rich Ziade: 34:52 It’s a generic sort of all-encompassing representation of how everyone is empowered now. That’s cool for us to do as a young company that doesn’t have baggage, doesn’t have history. But in an org where literally entire columns of hundreds of people in the org chart exist to work a certain way, and the other columns are sort of tired of waiting, or they want to mess around, they have ideas of their own, are starting to now infringe upon that structure… it’s the org… what it will tend to do is revert back to stability. And that is—we think the whole world’s going to change in the next 12 months because of these tools, but the resistance of the entire organization is incredibly strong.

Paul Ford: 35:39 So one, we disagree on this. I think the whole world is going to change. But two, I have amazing news. There’s a question. Gentleman’s back there.

Rich Ziade: 35:44 I wasn’t done with my profound—

Paul Ford: 35:46 No, you get an award for questions, but there’s going to be questions. Go ahead here. All right, go.

Rich Ziade: 35:51 Go ahead, go ahead, go, go, go. All right, this better be good.

Paul Ford: 35:56 Well, Rich, I’m glad you said that because this actually tees directly up what you were talking about…

Dan Shipper: 36:00 …what you were just saying, which is I worked in one of these really large organizations, where the interest in someone from one column touching anything in the other column was zero. Like, zilch. ChatGPT-3 was out and I had, I was comfortable enough in JavaScript, I was a product manager, and I had, I landed a pull request that was a one-line, I need my tracking, and it was like weeks of meetings about how I’d fucked up. And so I think my question for you all, especially like across tech services, is like now that everyone can just get in everyone’s org charts, like I actually for product orgs specifically, eng, design, product, like that’s the dream for me, is like everyone’s on everyone’s toes.

Rich Ziade: 36:43 That’s how you get better product.

Dan Shipper: 36:44 Yeah. But like for a 25,000 person financial media company, that’s harder. And so how do you all think about how you help coach through change where there’s so much identity wrapped up in, no, I do this, this is my CI/CD.

Rich Ziade: 37:01 It’s a great question. Um, if you look at the technology services space, right? And we ran these, I forget the exact numbers, but the IT services industry’s trillions of dollars, right? There’s giant mega companies like Accenture and these large organizations that, you know, uh, service companies. And then the, the transformation space, where like, I need to transform my business, either on its competitive, was like 20 billion. Like a sliver of it, right?

Paul Ford: 38:29 What’s a garbage number, billions?

Rich Ziade: 38:31 It’s a garbage number, billions. It’s a tiny number. Why? Because the appetite to completely reset things… we had, we had someone come in from Macy’s at the old agency and they’re like, ‘You know, we spent $130 million maintaining our old shitty e-commerce platform that, like, runs all the old beige boxes.’ And we were like, ‘Do you want help with that?’ and they’re like, ‘No!’ No! Most don’t want it, right? What’s unusual now is the tools have sort of introduced massive transformational, transformational opportunity, right? And so does it happen from within? I think if you want to spin up an AI agency now around bringing these tools for transformational change into companies, boy, it’s a time to do it because Dan literally did it right now. Dan’s doing it. Why? Because… Wait, I’m gonna I’m gonna say it’s even simpler than that. When you bring an outside actor into that, it cuts through the political rigidity that’s within that org, right? And that’s, that’s part of why you go outside. Why did you go outside? Well, because it was resisting, the org was resisting, right?

Paul Ford: 39:00 I’ll answer in two seconds and then go to questions because I’m tired of my own thoughts. But I just think we’re in an era of chaos and suck and I don’t know why something else would cause… no, like why wouldn’t something else cause? Why would our shit industry suddenly be have like the perfect infrastructure when we can’t… our government isn’t working. So why would this keep working?

Rich Ziade: 39:19 Um, now, is something else going to happen? Paul differs on this, so I’m curious to hear what he thinks. He thinks people just like, it’ll, the org chart will collapse. I don’t think it will. I think it’ll take a very long time. But I’m curious to hear Paul Ford’s thoughts. Why would what keep working?

Paul Ford: 39:21 The bureaucracy that keeps people from doing great… like I think it’s going to be like ‘this is really cheap and really fast’ and the boss is going to go like ‘beh, I don’t know what to do with that one’. People are going to be pissed off.

Rich Ziade: 39:32 I think, I think when the competitor across the street makes it way easier to, I don’t know, book flights because they’ve modernized everything more aggressively, I think competition will do it. I think economic downturns can do it, like cost pressures can do it.

Paul Ford: 39:50 Then I have good news. Alright, there’s a question over there. Eat all the food everyone. Um, I guess my- my- like two questions. One was a bit of a follow on to the culture question. I’ve also worked in a- you know, a large organization which- one of the core values were ‘not hierarchical’ but it was incredibly hierarchical and- and just fundamentally not a tech company. And so culturally regardless that there were so many talented people, you know, global management consultancy that’s here to help big companies go through these transitions but cannot get its own stuff in order. Um, which is kind of fascinating. And so I guess like, yeah, it’s like identity, it’s org charts, but like culturally like how really do you grapple with that? Or- or do you? Like do you just say like ‘okay, well you’re not our client because just culturally you’re not, we just see that you don’t have the ingredients to really carry this through’? And I guess the second part to that is like do you think that for these orgs that are just already not like product mindset, they’re still in project kind of domain, you know, like not tech-enabled, um, could you just like go… is the gap going to become greater? Like…

Rich Ziade: 41:11 I kind of want to throw this at Dan because I think you’re seeing them when they’re little babies and they’re going like ‘what do I do?’ And like, so why do they call you at… no, I mean versus the other set.

Dan Shipper: 41:18 Yeah, I think… I think basically like they read our stuff and there’s a… there is a distinct feeling right now that things are changing rapidly, the stock market is down for SaaS stocks, they’re like ‘maybe I’m next, I need an AI strategy’, which we’ve been hearing for many years now already. But I think we’ve been through several cycles of ‘we have an AI working group’ and like ‘we’re going to do some stuff’ and… and people feeling like ‘ah, that stuff didn’t actually necessarily turn into what I thought it would turn into’.

Rich Ziade: 41:51 I think also people don’t want like the AI innovation lab at Accenture to come in, right? They want to talk to people who are doing it in this very active way.

Dan Shipper: 41:58 I do, I do think that that is… One of the differences that we care a lot about and that is effective is we’re not just reading about it and then prognosticating. Like we actually just run our organization this way and then we take what we learn through the lab or the playground of the internals at Every and then we try to bring it in a practical way to other companies. And we don’t have like a system necessarily where it’s like one, two, three, four, this is how you do it because I think every organization and every industry has its… it’s going to look a little bit different. But what we try to do is have a lot of different things that might work for you, both internal to us and from other people who we know that we think are really good, and then all of the… any one of those things might be the thing that you can start with and then grow from there. I think I mean I gave one path which is go outside and sort of cause disruption within your… I think another way is Tracy still here?

Paul Ford: 43:00 She’s in the bathroom.

Dan Shipper: 43:02 Okay. Great, Tracy, who’s in the bathroom, but she’s a client too, so you just…

Rich Ziade: 43:06 She’s a client. She’s the CEO in fact.

Paul Ford: 43:09 Is Claire here?

Dan Shipper: 43:11 Did she leave?

Rich Ziade: 43:12 She’s in the bathroom.

Paul Ford: 43:13 Also in the bathroom.

Dan Shipper: 43:14 It’s work…

Rich Ziade: 43:15 Search party for Claire.

Paul Ford: 43:17 They were a client.

Rich Ziade: 43:19 They were a client, I guess.

Paul Ford: 43:21 We talked to them for a couple of months, they visited this office…

Dan Shipper: 43:32 This opportunity would have never come together this smoothly without leadership saying, instead of saying “go learn about AI, let’s do some R&D around AI,” she, like the authority with which she said “we are going to transform how we look at data across our clinics with AI and I want this done.”

Rich Ziade: 43:54 I mean the synthesis is like the people who are saying drive this value into our org now. Or else. I think most of the sentiment is like we should be doing something about AI, go learn about AI.

Paul Ford: 44:08 Dan, you got something? But she was always leading.

Dan Shipper: 44:13 We’re talking about you, Tracy.

Paul Ford: 44:15 Welcome back. She was doing AI before anybody else was, that’s the point. She was leading by doing, by using Claude before anybody else was. She gave me my first lesson on Claude.

Dan Shipper: 44:31 Yeah. I mean we’re switching now to this. I mean we’re signing up so we’re obviously going to transform. No, but we’re talking about how you sort of, you decided to drive the value of AI and of this change into your org, which is a very large set of, it’s a consortia of all kinds of different children’s behavioral health centers and data products and service, it’s complex, and you said like, we’re going to get this in the org. And we’re part of that, right? Like that’s, and we’re…

Paul Ford: 45:00 Actually talking about inspired leadership here, your client is is going great.

Dan Shipper: 45:03 I mean, thank you. Yeah, I mean it’s a great partnership though for— I mean, you can have a vision and think about what you’re going to do and then, I guess I could say this with people that know this, I mean when you start talking about it, a portion of the people that work for you think you’re crazy and that you’re overreaching and that you’re like super optimistic in a way that’s unrealistic, but you think it’s not and then you meet a partner who says, ‘We can do what you’re saying.’ And I’m like, ‘They said it, see?’ Like these are people that do this. So I think it’s you have to have that trust in the partnership and that’s where I think dragging people along becomes difficult because you’re asking people to share data or asking you to trust AI, not not just throw your hands up and believe anything they say, but this has been a relationship that I think helps move our team forward in a way they need to.

Rich Ziade: 45:59 Good. And I think clearing the way, you have to really, it’s a force, I mean Tracy’s a force, I mean I’ll say that out loud, I mean the vision was bought in and she she asserted that this has to happen. And why I don’t mean to single you out, but that’s very different from an exec saying I’m going to put a little budget aside for some AI stuff, right? It was like, this is going to be part of our how we think about things, how we advance going forward, the world is changing, and you’re going to find some some organizations are going to have advocates that are going to be very assertive about that change. And they know they’re going to break some things along the way to make the change happen.

Dan Shipper: 46:41 I really do think like leading from the front and having a mandate is effective. And if you’re depending on how you do it, you can also encourage people to use tools before they’re ready or, you know, there’s a lot of there’s a lot of things to add to that. So for for me, there is you have to have a little bit of the stick maybe but also the carrot is really cool where where you’re transitioning from oh my God I have to use this to like I get to use this cool thing, you know, for a senior engineer who’s like don’t touch my pull request, don’t touch my codebase, this is my CI. They they have things like one of our engineers who runs Cora, which is our email app, like Kieran Kieran is a beautifully designed app and Kieran hates translating Figma into code. And now Lucas, who’s our designer, just does it. He can just make the code. And um so that’s a thing where where it might flip from no this is my codebase to like great like just just sling it in there, let’s go. Um and I think the way to create those moments is to get people to put their hands in it in a way that is maybe outside of the bounds of your day-to-day job and is more in a kind of collaborative hackathon workshop type environment where you’re like oh my God I can do all of these things. And that I think also breaks down some of some of those barriers too.

Rich Ziade: 47:59 One thing you need that… And then I’ll add one thing I would say today. So in the organization you’re talking about we have a coalition of 715 members in 24 states, right? So there’s a lot of diversity in what you can and can’t do. But I also find the best way to get people to want to do something is to tell them somebody else is doing it, right? So all of a sudden, it’s true, all of a sudden I’m like, oh wow, they’re on the data council, they’re already adopting this, they’re doing this, and they’re like, wait, what, how do we get on it? And I’m like, we asked you and you had like a thousand questions of everything that could go wrong and I’m full force ahead. Run forward, break things, go back and fix them when you need to, but you have to progress or you’re never going to move.

Dan Shipper: 48:38 Totally, and that’s why it’s really effective to identify the early adopters that already exist in your organization and make sure that they’re really visible because once they see someone else doing it and oh, they got kudos or they got a promotion or whatever, it becomes a snowball that starts to roll.

Rich Ziade: 48:56 One of the terms that everybody’s looking for is the asterisk in works. And when I say the asterisk I mean, so you just described a very good use case, which is designer can now ship component system that’s ready to go for engineer. And that is a very low risk thing because the component system is it doesn’t have a deep security profile, there’s not a lot that can go wrong except that a component can kind of look bad or break a little bit. And so that is a net win in terms of speed and flexibility in the organization. But what’s happening is you, I think these things are coming monolithically. We’re going to use co-pilot to build a new thing, everyone’s like, it’s not going to be secure, we have to use our thing. And so these asterisks they start to pile up. But the reality is you can do a lot and you can do a lot very safely if you carve pieces out. And so now you have this designer who’s fully empowered to do things that used to take a lot of engineering that engineers hated. And so now the org can run a lot faster which is pretty much a net good and there just aren’t a lot of asterisks to add. So I think culturally we’ll see that change, but just know like every time I go into a room there’s a million asterisks and it’s just we’re sort of crossing them off one by one, some are legit.

Paul Ford: 50:04 Yeah, I… Wait, there’s last questions. Go. Go. Go. Go.

Dan Shipper: 50:12 So you just talk, you said it, the org runs faster, pretty much a net good. AI generally speaking a machine for more. Yes. Pretty much a net good is not a net good. At what point do you declare a cutoff, right? At what point do you say we’ve gone far enough then when you build, when you build proof, at what point do you say this is, we’ve gotten there, we can call the agents off and because this is the concern right, is that end up, we end up in a world of gray goo or fucking paper clips, right? I totally, I don’t think that we’ll end up in the paper clip world, but I think that’s a really real concern which is…

Paul Ford: 50:58 We have to just pause for a sec. The paper clip…

Dan Shipper: 51:00 It’s a great game. It’s true. Just go play Universal Paperclips. But the idea is a thought experiment that you make a little machine and its job is to make paperclips, and it’s got AI and it’s like, ‘Oh, well, you know, I need to make more paperclips.’ It’s really smart and it captures all the resources, controls the economy, and it turns the entire universe into paperclips. It’s called the paperclip maximizer problem, and if you want to know why you can’t stand the people who run AI companies, know that this is their favorite thing in the world, and this is just what they live and breathe.

Paul Ford: 51:30 And to be clear, I don’t think that’s an actual problem, but I do think you can stop way short of that and still stay a morale…

Rich Ziade: 51:35 Yeah, for me, you’ve got a paperclip factory that’s like the entire economy of a major state or something.

Dan Shipper: 51:39 That’s right. Yeah.

Paul Ford: 51:41 I mean, this is the delta between, like, should and could, and, yeah, I’m just curious, like, your clearly…

Dan Shipper: 51:47 I think it’s a totally new thing. So, uh, so one of the really interesting things for Claude Code is they have a metric that is ‘how much have we deleted?’. Because it is not just a machine for more, it is a machine to delete stuff too; you can find the right things to delete. And I think what that says to me is there are actually new problems to solve and new conventions that we need to have for what good products are because, yeah, when AI… I would say probably five or six people have committed code to Proof in our org that are not like me, that are not looking at it all the time. And they’re not agents; they’re humans. And so when you have that velocity of things coming in, you risk losing the coherence of the product. And what’s even more interesting is we live in this prototype world—you were talking about that earlier—and one of the risks of prototypes is it’s just new and shiny, so you’re like, ‘Whoa, this is amazing!’ and then you get it in your product and you’re like, ‘Well, no one’s using this and this, like, makes the product worse.’ So, so I think those are all things to be aware of, and it’s part of the new skill of building products or building software to be aware of them and to mitigate them. So for example, one of the practices that Anthropic does, like I said, is they have a metric for how much have we deleted from the product? Uh, and there’s lots of other ways to do it.

Paul Ford: 53:08 Let’s do… related… So, related, so like you said that you built this Proof in ten days.

Dan Shipper: 53:13 It’s called Proof, yeah.

Paul Ford: 53:14 Thank you. With two O’s. That’s all I’m checking. Okay. In ten days, but like is that fully encompassing, like, the creation of the idea or the insight that it’s a need that you had? Like what’s the timeline for that, and how did you decide, how long did it take you, you know, like…?

Dan Shipper: 53:32 Yeah. Well, it’s really hard to find the exact starting point of an idea. And we actually launched three and a half years ago, when GPT-3 first came out, even before ChatGPT, we launched a text editor called Lex that had AI. So we did that then, and then since then, Kate will tell you, I have every three or four months come in and be like, ‘I think I can automate your job!’ Um, and have come up with—in a nice way, in a nice way… hey! You know, you can do other more interesting things or your job.

Paul Ford: 54:03 Yeah, your job.

Dan Shipper: 54:06 Um, let’s say, I think I could automate the parts of your job that are not worth your time, let’s say. Um, and I have been mostly incorrect. It’s actually some of these problems, which for example, one thing that we’re trying to do is we spend a lot of time copyediting. We spend a lot of time like figuring out like where does a period go and like when is a semicolon appropriate and are there spaces around the em dashes and all that kind of stuff.

Paul Ford: 54:30 And you didn’t know?

Rich Ziade: 54:33 You can’t put spaces around em dashes.

Paul Ford: 54:35 Right! Are there spaces? Do we delete them? Next to em dashes?

Dan Shipper: 54:43 Um, and it turns out it’s a really hard problem. And every single time I’ve thought it was solved, I’ve run into there’s - the models are not ready yet. And this was another instance of I think the models might be ready. Like as soon as I saw, okay, we have the infinite five-quoting machine now, my wheels started turning. And so I built the first version of Proof in early January. But it was a Mac app and it was a very different product. And that took me, you know, a week or two to get a thing that I could use. And then I spent a month or so trying to adopt it myself and get other people in the org to adopt it. And what I realized eventually is the Mac app part of it was not that interesting. That people were actually just using this ‘I get to share my plan now that I made with the agent, and now I get to share it with other agents on my team.’ And so I was like, I’m going to scrap that and I just started from scratch basically. So you can pick the starting point, but yeah.

Rich Ziade: 55:46 Alright, we’re going to do one question to close out because we’re at the hour. And, Molly, let’s go. I know, I know. Well, no, we kind of got there-ish because really what the point is of these industries is that they’re big and they have a lot of friction, and it’s going to take a while to convince them to bring these new tools in. And we could go, there are kind of apps for every part of it. I mean, let me throw it to you, you were thinking about this too, Paul.

Paul Ford: 56:12 Yeah, I mean, I talked about it before. I mean, pick an industry and there’re probably 30 AI startups that are going to disrupt it, right? There’s all these tools to accelerate things and make your work smarter and whatnot. And I think, I think it’s going to go slower than people think. I’m going to stand by that. Um, because of the rigidity of how people work and the legacy of how people work. I think on desks, people will do stuff and then have an afternoon because the thing that used to take four hours takes an hour. Um, but I think resetting those industries, I think it’s just a very, very, very stubborn thing. We still have calls to this day, like, ‘How can we move…’

Dan Shipper: 57:00 Help me. Meanwhile, the whole business is on a spreadsheet, right? And it’s a business in the nine figures of revenue, and it’s on a spreadsheet. It’s like, well, how about we hang back a second here on AI, just get you off spreadsheets.

Paul Ford: 57:11 And the truth is a huge part of the economy is still running that way. And I think when we jump ahead and think we can sort of parachute these magical things into these orgs, I think that’s— I think we’re getting ahead of ourselves.

Rich Ziade: 57:27 So I can give us a concrete strategy here. So first of all, New York tends to be run the world or organizations are bigger, they have more bureaucracy, and they have more power. We just talked the other day to the New York Food Bank. And I mean, it’s like— it’s a combination of like— it’s also like an educational institution that teaches other food banks. It’s just— it’s vast. The Museum of Natural History has a graduate program inside. So I think part of it is that things that the West Coast thinks it can solve with yet another tool just don’t get solved here in the same way, and that’s never been the case for thirty, forty years. It just doesn’t get a lock. So I think if you’re really going to go meta— and then we could go industry by industry, we don’t really have time, I sort of wish we did, I mean it’s just— but the— I think the right way in is to actually come in and be like, look, don’t give your whole product roadmap or your life to this new technology. You’ve had robots in here before. They scanned all your invoices, they scanned your digital archives, they organized all your folders, you have a digital asset management system if you’re advertising, you have electronic health records, and they’ve made a mess. You have archives of stuff everywhere that the old computer systems messed up. Let the new robot clean it up, improve their value. Don’t just give them your whole roadmap. Clean up all of your old stuff, clean up your technical debt, and make it better. And then once that’s acculturated— and nobody wants to do that work. It sucks. The teams aren’t there. So maybe we can do that cheaper for you. In some level, that’s a reason why we exist, right? Like we’re going to be people who go and do that. And so now, I cleaned it up, it works better, it’s nice. Maybe we could fix some bugs. Maybe we could work with you on some other stuff. Maybe we could help you accelerate your product roadmap, maybe we could do some things along those lines. But step by step. And I think if you really think about New York, as fast moving as we are, the velocity of this, the city is too big to really turn around. So there’s going to be a lot of steps like that. Clean up the old mess before you make a lot of new mess. And I think if you do that and you talk to orgs about that, I think they’ll say like, ‘Okay, I would like to try some of that.’ But I do think here we have a resistance to that much change because we have so much power with the institutions that we have. They’re so big and they’re so complex, so it’s going to be a lot of steps like that. Clean up the old mess.

Paul Ford: 59:40 I totally agree. One of my maxims is never make any major life decisions 30 days after a meditation retreat or after an encounter with a biotech AI product.

Dan Shipper: 59:51 It’s a really good maxim.

Paul Ford: 59:53 It works. And the reason for that is—

Rich Ziade: 60:00 When we were outside in like Peru?

Dan Shipper: 60:02 Panama.

Rich Ziade: 60:03 Okay, so there’s this… I’m getting an Ayahuasca vibe.

Dan Shipper: 60:06 There was no Ayahuasca that I know of, it’s just, it’s just agents. Sure. Just checking. So, you have this reaction to it that is… it’s… the thing I like to think about is in the Middle Ages when they imagined what it was like when you got to the end of the horizon, you know, what happened at the end of the horizon. It was like it falls off into nothingness, or there are dragons, or whatever. That’s the same thing that you feel when you encounter frontier AI, because your intuitions about what it might mean are going to be very wrong. And one of the intuitions that happens a lot that I think you’re pointing to that is, I think very important, is maybe this time it’s different. I know that in the history of the world and in the history of technology, technology has tended to not necessarily give us less time with work. We actually like probably work more. I know… you can sort of go down this… I know that it takes a really long time for companies to adopt this kind of thing, and I know that when I go to, you know, my dry cleaner, like they still don’t even take credit cards. And that there are all these different organizations and all these different little niches that have adopted new technology to some degree or another, but it’s a very, very wide range and it’s a constantly moving target. And when it comes to AI you’re like, ‘Oh my god, maybe all that is no longer true and all jobs are going away and we’re all going to be replaced by robots immediately.’ And I just don’t think that that’s true. And the evidence for me is just my own experience with this stuff, either working with big companies and watching how hard it is and watching how long it takes, or internally for us, like I’m fucking working right now. I’ve got my agents working, I’m like looking at it. So I have not found it to make me work less actually. And that’s not about the technology. That is about me and just like humans in general, what happens when we get more ability to do things.

Rich Ziade: 62:22 All right, so everybody gets to work all the time forever. That is a really great vision of the future. Thank you.

Paul Ford: 62:29 Yeah, let’s one more round of applause for these two people. All right, so that was a lovely… it actually… I usually try to keep it for 40 minutes, this kept going. I know there were more questions, just ask them to us, we’ll… I can yell them out. But we’re not going anywhere, we’ll be here for about another hour. Make use of the space, ask any questions, eat lots of food, and make yourselves at home and then we’ll be out. Shoot you out. I’m tired.

Dan Shipper: 63:02 Thanks everyone.

Rich Ziade: 63:03 Um, okay. Thanks everyone.