Come on Josh. You’re technically right, but you’re missing why people keep saying this.
The Buildprint point deserves a direct response. V1 and v2 are not the same product. V1 was about turning Bubble’s JSON into human-readable documentation and making it queryable. V2 is a new tool entirely. An autonomous agent that diagnoses and fixes real builder issues. It was built in a few weeks by someone with a full-time job whose livelihood does not depend on this. He built it because he doesn’t want to be forced off Bubble. Bubble has had a funded team working on this problem for two years and hasn’t shipped anything close.
That’s where the meme comes from. Not “Bubble ships nothing.” It’s that Bubble ships, but not the things we’ve been asking for. And when it does ship something in that direction, it’s a watered-down, hype-first version that doesn’t deliver the actual utility we need. The AI Agent is the clearest example of that. Years of anticipation, and what we got doesn’t work on our existing apps, hallucinates completed changes, and has broken bindings in its own announcement video.
Then there’s the other category: things that haven’t shipped at all, in any form, after years of asking.
- Front-end to back-end communication that feels native, not duct tape
- Actual looping primitives for workflows (not “schedule API workflow on a list” as the answer to everything)
- Parallelism and fan-out/fan-in patterns that don’t require hacks and fragile orchestration
- Native UI elements so we stop rebuilding basic platform controls from scratch
And it’s not just Buildprint. I built the first version of BUBS, my Bubble AI agent, in a single day. It wires front-end UI elements, reads the Issues Drawer, fixes them, and works across all Bubble apps including existing production apps. Improvements since then have come from actually using it. That’s the bar. Bubble’s AI Agent launched in October for AI-generated apps only, and five months later it still isn’t usable as a daily driver for apps we actually have in production.
Here’s what Bubble shows you the agent can do:
Here’s what actually happens:
“Changes approved and applied.” Twice. 28 errors before, 28 errors after. Then when you push further, it tells you it can’t read the issue drawer and can’t accept a screenshot and can’t fix the issues when pasted as text..
And it’s not just user sessions. In a new AI Agent video, Bubble’s own produced content has fields showing [not found: destination_text] and [not found: trip_notes_text]. Broken data bindings baked into the marketing material.
So yeah, maybe the company is shipping a lot across surfaces. But when someone with a full-time job and zero financial stake can ship a leap like Buildprint v2 in a few weeks while the core builder asks sit for years, “we ship a lot” does not land as an answer.
The question isn’t whether Bubble is shipping. It’s why the things we’ve been asking for the longest are either still untouched, or show up as a press release before they’re ready to actually use.