@georgecollier In build mode, Is it possible to show the number of tokens used and time at the end of a build request?
We will at some point, neither Claude Code nor Codex surface those easily tho which is why it hasnāt been done already
Canāt we relay on the usage object thatās returned by Claude/Codex to show output_tokens?
{ "id": "msg_...", "type": "message", "role": "assistant", "content": [...], "model": "claude-sonnet-4-6", "stop_reason": "end_turn", "usage": { "input_tokens": 25, "output_tokens": 142, "cache_read_input_tokens": 0, "cache_creation_input_tokens": 0 } }
a level of financial security one can only dream of
This is super dope! Man, I wish I had time to get Buildprint rummage through my app. Thereās just so much project red tape and paper work that I am dealing with. Iām only able to work on smaller updates and fixes.
As much as the AI agent is the way forward (and the way Bubble should have implemented theirs) Buildprintās Bubble app observability and troubleshooting tools and capabilities are the real wins for me.
Buildprintās features are true to Bubbleās spirit of small teams working on large projects. Iām still solo deving and I havenāt tried yet but I can see troubleshooting becoming leagues easier. Documentation is also something Iāve been procrastinating but from what Iāve read of what Buildprint can do, Iām actually excited to start working on documentation!
3 Likes
Upgrades to agent and automated tests quality
Today Iāve made a bunch of improvements to how Buildprintās agent can explore apps and build new features. You should notice Buildprint be able to locate things easier in your application, and more accurately build new features that it hasnāt previously seen in your app.
- Buildprint can now see linked elements/workflows (e.g when looking at a workflow, it can also see the element that triggers it, or when looking at an element, it can see all workflows that use it). This lets it efficiently explore logic relationships in your app.
- Comments now supported for API Connector calls and groups
- Fixed issue with diffing history of same version
- You all now have auto-generated project summaries based on your app. You can review and update them from
Project settings. - GLM 5.1 has been made available
- Buildprint will be less likely to omit optional fields that ought to be filled when editing
- Buildprint will be better at implementing actions, expressions, and other logic that it has never seen before
- Tests will now run moderately faster, and agents have improved guidance to deal with common Bubble elements like date picker and searchbox which arenāt super straight forward.
- Buildprint is now better at writing its own tests and docs have been updated with more best practices here.
8 Likes
Killing it as usual
@georgecollier Can we add a feature that converts review comments into a plan then take it to the chat for the agent to build?
2 Likes
Hey @georgecollier, can you give our application access to editing/build mode? Canāt really do much until then.
Editing will be released to everyone in a couple of weeks
Until then, thereās a lot you can use Buildprint for:
- triaging prod issues
- planning new features
- diagnosing bugs
- implementing automated no code tests
- identifying regressions when merging branches
- setting up alerts and monitors for events using the observability tools
4 Likes
Thanks @georgecollier for expanding the Best Practices
You mention āRun reviews on every branch that touches high-risk areas: payments, authentication, user data, or privacy rules.ā
But why not run a review whenever Buildprint modifies anything (not limited to high-risk areas), just to be sure Buildprint didnāt expand the scope from what Iām expecting?
Sure, run reviews on whatever you want
Itās more of an āat leastā run reviews on those branches. But for straightforward changes you might not want a review - ultimately you donāt want to slow down your deployment process with procedure thatās not always going to be necessary
1 Like
is this the first half-decent open model? sounds like a good option for āchange a bunch of stuff exactly according to X planā
Yeah but its same price ballpark as GPT 5.3 Codex which is probably better
hmm surprising, usually open models are dirt cheap and thatās the whole appeal
Itās because GPT 5.3 Codexās cache read price is actually lower than GLM 5.1ās
1 Like
Upgrades to edit speed and Buildprint token efficiency
This update targets speed and token efficiency when Buildprint is completing edits on your application
- Previously, Buildprint would ātestā its edits to see if theyāre valid. It then saw the JSON result. Then if happy, it would apply the real one again and have to write out its edits in full again, and then see the result again. Thats a lot of tokens! Now, we let Buildprint re-use draft edits without having to write them all out again. We also no longer surface the resulting app JSON (as it can already access it if it needs it using other tools). Less write tokens when doing edits, and less read tokens when applying them.
- Buildprint can now create, list, and restore to savepoints itself (the last two are new)
- Buildprint can now clone an entire page + copy UI from anywhere in the app without having to write it all
- Removed support for
/paths when exploring app JSON as it could cause agent confusion
Anyway, who wants first-class support for mobile native apps with Buildprint?
5 Likes
Well, I guess the next step is to just build your own platform 
I would hope Bubbleās AI gets going a little better than it is. I still have confidence it can.
I built my own AI that scans my app and gives me feedbackā¦
probably could be improved, but it works for now.
Havenāt tried Buildprint yet, but with all the positive reviews, I think I should
9 Likes
nothing to lose by trying
2 Likes