If you opened Base44 recently and your app behaved slightly differently than you remember — a redesign suggestion popped up somewhere new, or a prompt that used to work landed a little off — you’re not imagining it. Base44 just changed its brain.
On June 29, 2026, Base44 launched Base1, its own proprietary AI model, trained on what the company describes as “tens of millions of real user interactions.” Up to now, like most AI app builders, Base44 was routing your prompts through third-party foundation models (the usual mix of large general-purpose LLMs). Base1 is Base44’s attempt to own that layer instead of renting it.
This came about a month after Base44 shipped Design Pack, a new visual editing surface with a collaborative canvas, and about a month after Base44 crossed $150M in annual recurring revenue — up from $100M just two months earlier. None of these are small moves. Here’s what each one actually means for someone with an app already built on the platform.
What Base1 actually is, in builder terms
Strip away the model-training language and the practical claim is this: instead of sending your prompt to a general-purpose model that also handles poetry, legal summaries, and customer support tickets, Base44 is now running at least some of your requests through a model trained specifically on how people talk to app builders — the actual back-and-forth of “fix this,” “make it blue,” “no I meant the other button” that Base44 has presumably logged across its user base.
The pitch, and the reason TechCrunch covered this as a competitive story rather than just a feature release, is defensibility. Any builder can call out to GPT- or Claude-class models under the hood; that’s a commodity layer everyone has access to. A model trained on your own product’s interaction history is much harder for a competitor to replicate, because they don’t have your interaction history. Following Base44’s acquisition by Wix, this reads as a vertical-integration play — control the model, not just the interface on top of it.
What that means concretely: prompt handling, code generation quirks, and how the agent interprets ambiguous instructions may now differ from what you’re used to, because a different model (or model mix) may be doing some of that work.
What Design Pack changed about the editing workflow
Design Pack, which shipped around May 25, 2026, is a separate change but worth understanding alongside Base1 because it changes where you work, not just what’s answering you. It added:
- A Canvas — an infinite board with live multi-cursor support, so editing (and presumably collaborating on) a Base44 app looks less like a chat thread and more like a shared design file.
- AI redesign prompts — a way to ask for a visual overhaul of a screen or section rather than describing changes element by element.
- A Theme panel — centralized control over the visual system (colors, type, spacing) instead of prompting each instance separately.
- In-editor video generation — producing video content without leaving the builder.
If you’ve been living entirely in the chat/prompt interface, Canvas is a new surface you haven’t had to learn yet, and it changes the default unit of work from “one component, one prompt” toward “one screen, one visual pass.”
What this means practically for your existing app
Two changes, same underlying theme: the tool you’re using has shifted, and some of your existing habits may not transfer cleanly.
- Prompts that worked before may not behave identically now. If you have a set of prompt phrasings you’ve learned work well for a specific kind of edit, treat them as a hypothesis again, not a guarantee. Re-verify on a low-stakes change before you rely on a familiar phrasing for something that matters.
- New surfaces mean new places to look when something goes wrong. If an edit doesn’t land the way you expect, check whether it was applied through Canvas/Design Pack versus the standard chat flow — the two may produce different diffs for what looks like the same request.
- Theme-panel changes are global by design. If you adopt the Theme panel for colors or type, understand that it’s meant to apply site-wide. That’s the point, but it also means a theme change is a bigger blast radius than a single-component edit — check more than one screen after using it.
- This is a good moment to save a known-good checkpoint. Whenever the underlying model or editing surface changes, it’s worth confirming your current app still builds and behaves as expected before you keep layering new prompts on top, so you have a clean rollback point if something behaves differently than before.
Honest unknowns
A few things TechCrunch’s coverage and Base44’s own announcement don’t spell out, and we’re not going to guess at:
- Whether Base1 is used for all prompts now, or selectively alongside third-party models for certain request types.
- Whether existing apps get any different treatment than new ones, or whether the change is purely forward-looking for new generations and edits.
- How this positions Base44 against competitors also moving fast right now — Lovable, for context, crossed $500M ARR in June 2026 and raised a $330M Series B at a $6B+ valuation. Base44’s move looks like a response to that pace, but neither company has said so directly.
Takeaway
Base1 and Design Pack are two different bets stacked close together: one changes what’s answering your prompts, the other changes where you work when you’re not prompting at all. Neither is a reason to panic about your existing app, but both are a reason to re-test rather than assume — especially for prompt patterns and workflows you’ve come to rely on. Treat the next few sessions in Base44 as “verify, don’t assume,” and you’ll catch anything that shifted before it costs you a real edit.