Jordan Hale: Ryan, okay — my week has been genuinely weird because I keep telling my non-technical friends about this vibe coding thing and they all have the exact same reaction, which is, like, "wait, so I could just... describe what I want and the computer builds it?"
Ryan Castillo: And the answer is yes. That's literally what vibe coding is — you describe the software you want in plain language, the AI generates functional code, no traditional coding skill required. Which sounds fine in the abstract until you realize a CNBC reporter just used it to clone a SaaS company.
Jordan Hale: That's — yeah, that's the thing. Jasmine Wu, CNBC, zero coding experience, and she builds a working Monday.com clone in under an hour using Claude Code — Anthropic's agentic coding tool. Published February 5th, 2026.
Ryan Castillo: For under $15.
Jordan Hale: For under fifteen dollars! And Deidre Bosa is right there running the experiment with her. This wasn't some dev with insider knowledge — it was live, on camera, no prep.
Ryan Castillo: The number that matters here is 630. That's how many people Monday.com cut on July 22nd, 2026 — twenty percent of the whole company, forty-five to fifty-five million in restructuring charges. They called it a pivot to an 'AI Work Platform.' I'd call it a company that watched the Jasmine Wu clip and did the math.
Jordan Hale: You think the demo is what triggered it? Like, directly?
Ryan Castillo: Not the demo specifically — look, the demo revealed something that was already true. And the tell is Monday.com simultaneously projecting nineteen to twenty percent revenue growth while making that cut. Companies don't restructure twenty percent of headcount when they're fine. The code already won. The org is just catching up.
Jordan Hale: But wait — before we land on 'the code won,' I want to poke at the velocity stuff, because that's where I think the story gets genuinely messier. Like, there's this developer, Ken Imoto, who actually measured his output after integrating Claude Code across five sprints. And his shipping speed dropped 40%.
Ryan Castillo: While feeling faster.
Jordan Hale: While feeling faster — that's exactly the part that's wild. His perceived speed and his measured output were moving in opposite directions. Claude Code is less like a faster typist and more like handing your to-do list to a capable intern. Sometimes they nail it in ten minutes. Sometimes they come back with something that takes you longer to fix than to just do yourself.
Ryan Castillo: That's a useful frame. One rigorous data point against a pile of self-reported anecdotes — that's load-bearing.
Jordan Hale: And the anecdotes are everywhere, you know — the '2 to 4 hours of dev work in under 20 minutes' claim is all over the place. But those are self-reported. Nobody's running sprints and measuring actual shipped output the way Imoto did.
Ryan Castillo: So what makes Imoto's case the one to trust? Is it just that he published the raw numbers?
Jordan Hale: Yeah, I mean — it's not that he's the only truth, it's that he's the only one who actually checked. And here's the mechanism I think gets buried: Claude Code isn't suggesting single lines. It's agentic — it runs entire multi-file workflows end-to-end, autonomously, from the moment it became generally available on May 22nd, 2025. That complexity is exactly what can mask a slowdown. You feel the speed on the front end and don't notice the cleanup debt piling up on the back end.
Ryan Castillo: Right — but the CNBC demo is still observable fact. Wu built a working dashboard. That happened.
Jordan Hale: No, totally — the demo proves Claude Code works for some tasks. It does not prove it works faster for all developers across all contexts. Those are just... actually wait, no, those are two completely different claims, and I think we've been treating them like the same one.
Ryan Castillo: And that gap actually vindicates the demo more than it looks like, because the thing Jasmine Wu built — it wasn't a stripped-down mockup. It had calendars. Monday.com's baseline doesn't even ship with calendar integration at that tier. A non-coder with no prep out-featureed the product she was cloning.
Jordan Hale: Wait — she went beyond the feature set?
Ryan Castillo: Beyond it. That's the observable fact. So picture a two-person startup, Sunday night, kitchen table — one founder just prompts Claude Code to build a project tracker. Done before midnight, zero developer hired. The calendar's already in there.
Jordan Hale: Okay and that — I mean, that's the scenario that should terrify Monday.com's sales team, you know? Not a dev shop running sprints. Just... someone who needed a thing and got it.
Ryan Castillo: Which is why Monday.com's own pricing move in May 2026 is the tell. They shifted to a hybrid model — you pay for seats, and you pay mandatory AI credits. Not optional. You can't opt out. That's not a feature launch, that's a company admitting the commodity threat is real and trying to make AI a revenue line before the floor drops.
Jordan Hale: Mandatory — like, you're paying for AI whether you use it or not?
Ryan Castillo: Exactly. And Anthropic keeps expanding what Claude Code can actually do — dynamic workflows launched May 28th, now orchestrating tens to hundreds of parallel subagents in one session. Auto Mode went GA July 10th. The task types where the CNBC result generalizes just keep widening. Doctolib uses this. Delivery Hero. These aren't toy demos — they're healthcare scheduling and logistics in production.
Jordan Hale: That actually — wait, that changes the stakes considerably. Like if it's working in production at Doctolib, that's not a proof of concept anymore.
Ryan Castillo: Monday.com's counter is the structured substrate argument — boards, permissions, typed workflows, all baked in. Their claim is a prototype can't replicate that architecture. And I'll concede the moat exists. But the pricing pivot tells you they don't think the moat is enough. The part that actually reframes all of this is what it means for organizational speed versus code speed — and that's where Monday.com's own restructuring timeline gets damning.
Jordan Hale: And that timeline is — I mean, that's actually the thing that won't let me go. February 5th, the CNBC demo. July 22nd, the restructuring announcement. That's seven months. A non-developer clones a Monday.com dashboard in under 60 minutes, and Monday.com itself needs seven months just to *announce* a reorganization — and they're still mid-transition right now.
Ryan Castillo: That's the actual asymmetry. Code generation speed is almost irrelevant to the competitive outcome.
Jordan Hale: Right — because the enterprise sales cycle doesn't compress. Regulatory approval doesn't compress. You can't prompt your way through change management.
Ryan Castillo: And that's Monday.com's real moat — not the code, not even the structured substrate on its own. It's that Doctolib and Delivery Hero didn't adopt Claude Code because it was fast. They adopted it because it plugged into existing compliance layers, existing permissions infrastructure. That took years to build. No agentic tool accelerates that.
Jordan Hale: Though — wait, we should be honest, we don't actually have rigorous before-and-after data from Doctolib or Delivery Hero. We're reasoning from signals.
Ryan Castillo: Completely fair. The Ken Imoto case is the only number with actual methodology behind it.
Jordan Hale: Which is — you know, that's kind of where I land. GitHub Copilot started this shift, just suggesting lines in real time. Claude Code is structurally beyond that — full autonomous workflows, parallel subagents. But the leap in tool capability hasn't closed the organizational gap at all. It's maybe made it more visible.
Ryan Castillo: The calibrated claim is: Claude Code is a genuine threat to commodity SaaS features — the CNBC demo proved that — but the competitive moat for incumbents is organizational machinery, and no prompt spins that up in 60 minutes.
Jordan Hale: Yeah — and that's actually a more sobering verdict than either 'AI kills Monday.com' or 'Monday.com is fine.' It's that the disruption is real and slow at the same time, and the companies that survive are the ones who already built the thing no one can vibe-code overnight.
Ryan Castillo: Look, I'll half-concede the Monday.com-is-fine case. The structured substrate is real, the compliance layers are real. But the funniest sentence in enterprise software this year is still: Jasmine Wu built a Monday.com clone in under 60 minutes for under $15, and Monday.com needed seven months and forty-five to fifty-five million dollars to respond. Those two numbers just sit next to each other.
Jordan Hale: That is — yeah, that's actually a devastating juxtaposition when you just say it flat like that. And I think, you know, that mismatch is the real story. Not whether Claude Code kills Monday.com. It's that whoever figures out how to reorganize at tool speed — not just build at tool speed — that's the next pressure. Like, the code got fast. The org didn't.
Ryan Castillo: That's where I land too. Genuinely. Good conversation — thanks for pushing on the Imoto numbers.
Jordan Hale: Yeah, that one data point does a lot of work. It was worth the detour.
