Monthly Community Update - May 2026

Announcements
1

Hi everyone,

This is the May 2026 community update. You can read last month’s update here.

This month has felt especially high velocity. We welcomed Kathleen, our new VP of Engineering, and Merritt, our new VP of Customer Success. They are joining at the right time: With the increases in agentic coding quality we’ve seen in 2026, there are opportunities to completely transform how quickly we can evolve our product. We hired Kathleen to lead the charge on this transformation, and Merritt to ensure that the resulting innovation leads to successful outcomes for businesses built on Bubble.

On that note, you might notice that this update is a little later than usual, because the whole team was together in the Catskills for our annual offsite. This year, the focus of our offsite was a company-wide hackathon, where everyone at Bubble used agentic coding tools to build, thinking big about transformations we could make to our product, new ways for us to add value to our users, and innovations to how we operate internally. We came in with a range of experiences with agentic coding, and a diversity of ideas: We came out passionately excited about what we collectively built, and how fast we were able to build it. Most of the projects were prototypes and proof-of-concepts, but we also built some things that look close to being shippable, so expect some surprise announcements this month — we gave the team permission to go off-roadmap!

April also saw us shipping lots of updates to the Agent, which I’ll get into down below. I’m excited by the progress there. The Bubble AI Agent and app generation features are powered by Anthropic, which recently featured us in a case study! Some stats from the story: We saw a 2X increase in first-week user activation rate and ~30% increase in user satisfaction with editing via the Agent after we switched to Claude back in March. This is concrete, numerical proof of what Bubble and Claude have achieved together so far. If you haven’t tried the Agent in awhile, I recommend giving it a go.

Let’s get into the updates.

What we shipped last month

The Agent got several new capabilities, including a UX overhaul that gives you a “plan state” before it starts edits and an obvious “success state” once it’s done. You can also undo and redo edits directly in the Agent rather than from the nav bar, which makes editing via chat more streamlined.

When prompting the Agent, you can now upload up to five images to help it create the components you have in mind. It currently supports JPG, PNG, GIF and Webp files.This is sometimes more straightforward than trying to explain the vibe of what you’re looking for (no pun intended).

We also improved expression generation, which gives the Agent the ability to support the full range of Bubble operators and data sources, and means that workflow conditionals, privacy rules, and component generation are all more reliable. This work also sets us up for the Agent to provide backend workflow support down the line. You can read more about those updates here.

Additionally, the AI Agent can now make “multi-turn” edits. After completing a change, it will suggest logical next steps (like wiring up workflows after building UI). This work is the precursor to compound editing (making multiple changes at once), which is coming soon.

After its release, we integrated Claude Opus 4.7, the latest model from Anthropic, and have noticed 5-10% better consistency across generated pages. And we deployed improvements to mobile app generation including more consistent style variables (most noticeably in the reset password and update app views) and contextually relevant titles instead of “Current View.”

The Agent now respects view and edit permissions while building, which means it has the same permissions as the user who initiates the chat. If you are an editor, it can make edits on your behalf. If you have view-only permissions, the Agent will only be able to inspect and provide guidance.

Lastly for the Agent, we rolled out access to all Bubble Ambassadors and Gold agency partners, and they’re helping us test the experience using the Agent on complex, pre-existing apps before we roll it out to everyone.

On the mobile front, you can now create a deep link to anywhere in your app via workflow. Read more about that here. We also added a setting for plugin editors so that they can specify whether their plugin is compatible with web apps or native mobile apps. This will make filtering in the marketplace much easier. If you’re a plugin dev, make sure to check that setting is correct next time you publish a new plugin version.

In case you missed it, the opt-in beta for the redesigned property editor launched April 21. Since we launched the first phase of the beta at the end of 2025, we’ve made a large batch of improvements based on your feedback. (Thank you to the readers of this update especially for everything you’ve contributed.) Some highlights: drag-and-drop reordering for conditionals, constraints and more; inline dynamic expressions, workflow fields and custom states; tweaks to scannability and readability via bolder headers and increased contrast; and a searchable color picker.The team also fixed 40 bugs last sprint ahead of this rollout! You can read more on the forum here. And for a more 1:1 comparison of the old property editor and the redesign, you can refer to our migration guide.

Lastly, security scan performance now has much faster median test durations, decreased failure rates, and a reduction in timeouts overall which should improve the first-time experience.

For a full list of smaller bug fixes and improvements, you can check out our release notes: bubble.io/releases.

What we’re working on this month

There’s also lot of work in-progress for the Agent and app generation right now:

  • Agent compound editing: This will let the Agent make multiple types of changes in a single request — UI, workflows, and data all at once — rather than tackling them one at a time. Right now the team is dogfooding it internally and the early results are promising. We’re targeting June for a first release.

  • Agent access for non-AI apps: Now that the Agent is live for all BAMs and Gold-tier agencies, the next step is expanding access to all non-AI-generated apps. We’re doing this over the next few months to make sure we’re incorporating feedback as we go.

  • Improving the Agent feedback loop: Last month we began integrating the Agent with the issue checker so that it can see issues that were added or removed as a result of its own work. Today it can catch and fix basic issues, but we’re working on significantly improving its capabilities there. We’re also giving the Agent access to screenshots after each edit so it can visually check its builds, not just the underlying code. Additionally, we’re continuing the validation work on the backend to reduce silent failures and help validate events, actions and custom states. This work should be ready in June.

  • Option sets in the Agent: Option sets are a core part of modeling data in Bubble, and soon the Agent will be able to create and edit them for you directly. This first version won’t include deletion, but it rounds out the Agent’s ability to work with your app’s data schema end-to-end. That v1 should be in your hands by late May.

  • Agent editing mobile apps: The next major step in Agent capability: bringing the Agent’s editing abilities from web over to mobile, so it can make changes on your behalf in mobile apps the same way it does for web.

Here’s what else the team is working on:

  • Property editor: Now that all users have access to the property editor redesign, we’re focused on collecting and acting on your feedback. We’re also looking into helping you generate dynamic data and expressions with AI directly from the property editor. That work is still in shaping, but we’re expecting it by late May.

  • Automatic JSON parsing in the API Connector: The API Connector is one of the more challenging parts of the product, even for experienced devs. As more and more people connect to AI models, it’s critical for this to be less confusing. Right now, responses come back as giant blocks of JSON that the editor just treats as text. We’re fixing that with native auto-parsing that will recognize and classify JSON (and allow you to switch it back to text if you prefer), which will make it much easier for you to use the responses from an LLM in your web and mobile apps. That release is scheduled for May.

ICYMI

  • Emmanuel’s next AMA will be May 21 at 10 AM ET. (This is a bit earlier in the day than usual, so if you’ve been unable to attend previously, we hope to see you there!) RSVP here.

  • We hosted a joint webinar with George Collier and Lucas Bennington to showcase Buildprint, agentic development built specifically for Bubble apps. The energy was really exciting.You can watch that replay here.

  • RSVP for the Bubble meetup in Atlanta on May 7!

  • A few members of the Bubble team will be at the Create With conference in the UK June. RSVP here.

  • Emmanuel talked to former Allbirds CEO Joey Zwillinger for the podcast. We released it just after the announcement of the company’s shift to AI. I think this episode offers some interesting insight in the lead up to their big IPO and Joey’s mentality when it comes to building his new startup. Listen here.

  • We also talked to the team at EqualReach, a talent marketplace built on Bubble that’s helping refugees find work. It’s a super inspiring story, and their success so far has been awesome to watch. Listen here.

New hires

In addition to Kathleen and Merritt, we also had several others join the team this month! Big welcome to:

  • Tim, senior software engineer

  • Taymoor, senior software engineer

  • Boni, marketing analytics manager

  • Annie, business operations and strategy partner

That’s all for this month,

— Josh and Emmanuel

3

Greatter things coming to Bubble!!! :tada: :clinking_glasses:

Any available links to an article or guide explaining the “Autofill handling” in detail?

4

Also looking forward on this release note @nick.carroll ! :bullseye:

5

Awesome. I think the final piece of the puzzle for new user retention will be the Plugin Editor Agent. If you go back in time and think about the core “enabler” of Bubble, it’s plugins. The modern version of that is the Agent integration. That way new users basically get all the extensibility and “wow factor” of something like Codex/Claude Code, except the app doesn’t ever fully break down. No matter how much slop is in the outskirts of the app, the base is always protected. That will be the differentiator versus every AI builder: “AI that doesn’t stop working.”

On the topic of power-user retention, what is the current thinking about retaining people who move off Bubble after a few years due to “high costs” or “limitations?” There seems to be thousands of experienced developers who have moved off over the years and now spend $200-$1000/mo on platforms other than Bubble after they hit PMF. They basically use the platform as a stepping stone but you guys never see the majority of the benefit. Have you thought about what’s needed to mitigate this? Is it adjusting WU? Making the database ACID? Loops? Parallel workflows? Enhancing the core “Bubble programming language?” It seems like this might be more important than chasing after $32 subs, but I haven’t seen much discussion on this.

I have always enjoyed the Bubble Boosts, so if this is anything like those, that’s exciting. On the topic of the roadmap, especially for the agent, you can just copy George Buildprint and everyone will be happy. He just released automated tests!

Does this include parentheses? For some reason I feel this feature gets overlooked due to its previous beta status. It’s one of the most important parts of the language. Complex nested and/or expressions is a perfect job for the agent and very cumbersome to do manually.

Please keep releasing practical enablers for modern AI apps. The next step should be simplifying the process of sending private files to API providers. This is a simple thing that 99% of AI apps will want to do one day, right? For cases where an extremely high percentage of users will encounter a roadblock, you guys need to preemptively think about and unblock them. They often refer to AI as having “jagged” capabilities, but that’s ironically also how Bubble feels on most days. Extreme capabilities mixed with extreme, often sudden limitations. I’ve been saying for the last year that Bubble needs to be the BEST place to build AI apps, and this is a promising step towards that, so thanks for that, and keep going.

Thanks for everything you do, @josh !

6

@josh What’s the update on the native mobile plugin editor.

7

To be perfectly honest, even without absolute authority to give an opinion, but from what I’ve observed, I see a high risk in the current direction.

I believe Bubble may face difficulties in the near future if most of the time and resources remain focused on building apps with agents, following this crazy vibe code craze. Bubble is already extremely simple to work with—which further reinforces my point: simplicity and control are still the biggest differentiators.

In my view, the strategic focus should be on:

  • Adding new native tools and components
  • Improving existing components, allowing for more customization and flexibility; if this already exists, there’s no need for an agent to build limited pages
  • Making development simpler and more efficient
  • And, most importantly, creating native integration with external databases, for example, Supabase.

This last part is fundamental. A native integration with an external database would allow frontend calls to avoid consuming unnecessary resources (for production apps, Bubble is “expensive” compared to Lovable, for example), in addition to giving users the freedom to choose which database they want to work with.

This would allow users to maintain their applications at a fairer cost, without losing the main advantages of Bubble—especially control over logic and frontend.

I believe that, by following this path, Bubble could build a very strong competitive advantage over other Vibecode tools.

Bubble’s differentiating factor is control, but by the time this agent is ready, it will already be obsolete and very expensive.

Honestly, recent decisions—at least those discussed here in the forum—indicate a path that could bring several problems in the medium and long term.

8

It doesn’t make sense that, even today, Bubble still doesn’t have a truly robust native integration with external databases like Supabase.

For more advanced users who want to build business rules outside of Bubble and reuse that data in other applications, this is essential. Native integration would allow for much more flexibility, better performance, and a more modern architecture.

Today, it feels like Bubble has stagnated in two fundamental areas: front-end and logic.

The components are limited and, in many cases, seem to have received no relevant updates for years. Some clear examples:

  • The calendar component is extremely basic and lacks native integration with Google Calendar.
  • The components, in general, hardly integrate with external services.
  • The icon pack is outdated.
  • The table component has remained in beta for far too long.

Given this, a question arises: why not leverage the strength of the Bubble community to evolve the platform?

Considering the current scenario—where the competition is both to attract new users and to retain existing ones—it would make much more sense to invest in:

  • More complete and modern native components
  • Ready-made integrations with widely used services
  • Real improvements in the development experience

For many users, it is much more productive to have good components and work manually on the logic and customization than to rely on AI generation, which often results in generic pages with limited components—in addition to consuming significant credits.

This current direction worries me a lot, both in relation to the future of Bubble and the sustainability of the applications I have built within the platform.

9

Why would you want it to? The value of Bubble is as a full stack platform.

11

Being a complete platform doesn’t preclude the possibility of having more integrations and freedom for more experienced users. Furthermore, Bubble is originally a no-code platform and now wants to become something integrated with “vibecode,” will you question Josh’s plans? As I mentioned, I’m concerned about Bubble’s chosen direction of abandoning what was already good to focus entirely on “building with AI.” Why would a user prefer to use AI if they already had ready-made, modern, and customizable components??

12

Most all of what you say is true. Personally I’m not worried about database stuff but everything else. I’ve been saying this going on 12 months.

Very worrying to read this and last months updates as only focused on AI and agent. Even more frustrating is rather than having a real boost day they vibe coded some undeliverable stuff we are teased with in this announcement to just wait and see what they’ve done…hopefully it lives up to the hype.

13

this is really discouraging. investing on the most volatile user type, and giving bread crumbs to long run users. what i’m afraid of is that this ai investments not returning expected cash ergo causing bubble to get enshittified to squeeze the needed cash from anywhere else possible.

we’re building a super app on bubble with multi-million investment, trusting the existing state. ai not even once worked in any scenario in our process, because we know what we’re building, and what we want to achieve. ai guessing around and giving half-baked, unoptimized solutions are not for it.

and i believe not many of those new users would want to pay $29/m for a single generic web app, which they can make with any other cheaper tool out there. i hope everything goes well but, waiting for expected platform improvements and additions getting more discouraging each month.

14

There is an idea called using a proxy when system is dynamic or the data set is hard to get, so if you need data set X, you use data set Y which is easier to get, accepting the margin of errors. So, here is the proxies for what happens if you don’t care about customer base,

Jaguars bet
Jaguars fall
Gilettes fall
I like the Disneys fall most, but I don’t know how you guys will feel about that, I am not from your culture and don’t want to offend anyone, but basically they don’t respect their customer base and fall in stock prices reflects hat 50% reduction in one of the greatest stock market bull runs.

Anyways, my base case is, regarding how updates on LLM’s are going, gpt5 didn’t met expectations, claude 4.7 didn’t met expectations, now Im sure mythos will never meet expectations, and seeing in U.S these LLM’s are cited for ad purposes at the very top level leadership and there is a war going on in the Hurmuz and soon to cause serious energy problems.

I think all in on AI is a mistake and will fail, but 100% sure leadership in Bubble won’t listen.

So, what I would do if I were Bubble is to have some sort of buffers, like if you go in for new users only, make the platform beginner friendly too to don’t churn that much, push for SSR or other performance metrics for newly introduced cheaper subscription models, make the plugin ecosystem more efficient, make the native better, make the product actually global.

15

Honestly, I believe Bubble should focus more on optimizing the web app performance rather than heavily focusing on a competitive AI app generation. Web app performance is important as Bubble is only serve their apps from US datacenters. This is just my opinion.

16

100% true and agree,

17

Completely right about database!!!

That’s the weakest thing in bubble! The path of paying for ultra expensive workflow units makes no sense. At the end, only limits bubble apps until you don’t wanna pay more workflow units.

Currently using 1.5M/month. Was not a consideration in early day, but now it’s a nightmare.

For a little database modification that with regular coding would be super easy with MQTT bubble charges hundreds every month.

18

Yeah, it’s like, “nice UI for me, but not for thee.” I’m not sure what the thinking is around creating a good UI for developers who don’t care about UI, and then leaving the developers themselves with scraps when it comes to making a nice UI for their users (who actually care about a nice UI).

Look at the native tooltip, for example. It’s a weird 1995 black box with sharp corners. Why does it matter that the property editor looks nice when the tooltip still looks like that? The UI is only as strong as its weakest component. It’s kind of like going to a really expensive 3 star Michelin restaurant but the door handle is broken or the plates are dirty. The quality of the food doesn’t matter when you get those fundamental things wrong. I think new users will end up feeling tricked or misled and churn when they see stuff like this.

Screenshot 2026-05-06 010540
Screenshot 2026-05-06 010540645×154 4.95 KB

Other examples of this include:

  • Why does the slider element have weird rectangular handles that can barely be customized?
  • Why does the dropdown element’s triangle sit on the edge of the box and require custom css for it to look normal?
  • Why is there no “toggle” element? A simple toggle which 99.99% of apps will use has to be made from scratch.
  • Why is the dropdown not searchable and why is the searchbox itself so limited? You can’t even customize the results of it, it has to display the field that it’s searching.
  • Why no native html datepicker?

They need to have a dedicated person on the “Bubble design system.” Just like how there is shadcn or other design frameworks, Bubble has to maintain its own system of components.

Historically the value of Bubble has been around the Disneylandification (abstraction) of various components (backend, frontend, auth, etc.) during a time in web development where there weren’t things like Next.js, Supabase, Convex, Auth0, React, etc. That’s why it was OK that none of the individual components were best-of-class. Even legit programmers and web developers would use Bubble for their own projects because it was so fast to iterate on. I don’t see that class of user around much anymore. Now that every aspect of web dev has been abstracted to a high degree and every other company has created their own version of lego blocks, the value of Bubble is in “convenience.” But I wonder whether that is enough anymore.

Typically generalists are less successful than specialists. Being a “jack-of-all-trades but master-of-none” doesn’t work out too well for organisms, people, companies, etc. The natural state of things drives towards maximizing efficiency in one area, that’s how society is structured. There are a few one in a billion exceptions (Amazon comes to mind) but I don’t think it’s a realistic path for Bubble. If they want to be a jack-of-all-trades in this current meta, they have to also be a master-of-all.

The goal should not be to make the best possible integration with Supabase as one user suggested, but rather to eliminate the need or desire to use Supabase to begin with.

19

100% serious, please make a separate post about this. Would love to get more examples.

20

Links

21

Yes. It’s the “low hanging fruit” like this that could be resolved so fast and yet have such a positive impact overall.

Imagine if the tooltip (for example) had a couple of basic styling options… Corner radius, background colour, font size, hover delay. Surely something that a bubble intern could knock up on their lunch break!