[Config 2025] Figma Make is here — let’s hear what you think! 🎨

So much potential but still some pretty big bugs/feature needs. The publish function uses old versions of the code (this is a showstopper!) and there is no way to push to GitHub.

Lol, did you add AI-driven spam to the end of my comment when quoting it, trying to sell a tool, or is the Figma forum been hacked.

“By the way, when prototyping anything car-related...”

Hey ​@Josh22 — good catch, thanks for flagging that.😎 We’ve gone ahead and moderated the post and removed the off-topic content.✅

If you spot anything similar elsewhere, feel free to use the three dots on the reply → Report, and our team will take a look as quickly as possible.

We’re continuing to keep an eye on feedback around Figma Make and really appreciate everyone taking the time to share their thoughts here!

Hi Tom, I'm trying to use Figma Make's AI to create components within websites, in order to import them into my portfolio that I've been working on for a while, and the problem persists. In the monitoring you reported that it was resolved, but I'm still having the problem.
 

It’s been such a pleasure using Figma Make — it’s a powerful tool that has genuinely transformed how I build and learn. It served as a stepping stone for me to understand programming logic and AI-driven design. The layout and structure are well thought out, allowing designers to explore complex processes without feeling overwhelmed.

While onboarding and setup are smooth, I’ve run into some friction when it comes to deeper editing and iteration. The point-to-edit tool—where changes rely on prompts—can feel limiting for designers who are more visually oriented. I’d love to see the process become more intuitive, similar to tools like Aura Build or Base44, where you can edit visually in real time—tweaking pixels, repositioning elements, or changing fonts and sizes directly instead of through voice or text prompts.

Another early challenge I noticed was instability with prompt overrides, especially when using the point-to-edit function. For example, when a new prompt is run after adjusting part of the design, it sometimes overwrites previously generated layouts or script logic without warning. This created a lack of reliability, as smaller fixes often undid prior configurations. While newer builds—especially Opus 4.6—have vastly improved stability and command handling, an explicit safeguard to prevent prompt overrides or the ability to “lock” certain components during revision would restore trust in iterative editing.

My journey with Figma Make started around October–November 2025, and I’ve used it extensively through multiple releases. Back in Claude 4.3–4.5, it often took five or more prompt revisions to reach results that matched my intent. Still, this process was part of my learning curve—it helped me better understand how AI interprets creative direction, and it genuinely accelerated my growth as both a designer and builder. I experimented across all levels of detail—from simplistic one-line prompts to heavily structured technical briefs—and even in the moments of frustration, I appreciated how the tool adapted to messy, human input and still delivered functional outcomes.

One of the most empowering parts of Make is direct access to code. Being able to extract and verify code independently—then validate it with other AIs or through manual debugging—was instrumental in helping me create more complex applications. That capability directly led to projects like Adoras 1.0 (adoras.ai), which was built using iterative Make prompting and external integrations. I really hope this feature—code visibility and export—remains a core part of the product moving forward.

That said, importing files or projects from GitHub sometimes breaks the one-to-one structure of both the code and visuals. Often, I’ve had to rewrite entire segments because they didn’t import as expected. It would be transformative if Make could parse GitHub repositories more faithfully so that imported projects visually and functionally match their live source code. This reliability would go a long way toward earning developers’ and designers’ trust who rely on version control workflows.

Visually, the UX of Make hasn’t changed dramatically, but the incremental backend updates have made a noticeable difference in code organization. I’d love to see a more visually integrated backend—live previews of how the data interacts with services like Supabase, backend schema visualization, or even modular component mapping—so builders can work seamlessly between visual logic and code logic.

For more advanced users, there’s potential to take this further. Imagine an internal CMS system within Figma Make that supports SVG and image management natively. That would remove the reliance on external databases or cloud file systems. Having that infrastructure inside the same interface would transform Make from a design-prototyping tool into a truly autonomous AI product builder.

I also want to highlight how much the recent updates have improved the user experience.

• The rollback feature is excellent for experimentation.
• MD file compaction and file-size optimization have made navigation far faster.
• I’d still suggest improvements to chat-based scrolling — the scroll control is inconsistent, making backtracking through revision history difficult. Adding search, sorting by date/time, and even A/B testing branches within the revision system would greatly improve project management for both designers and developers.

Another major opportunity I see for Figma Make is incorporating new creative input methods:

• Voice prompting — many designers articulate ideas faster verbally than they type, and they often describe visuals spatially or emotionally rather than with precision syntax.
• Video input or upload — being able to record or attach a short clip to explain design intent or workflow context would allow for richer prompt understanding and drastically speed up ideation.

With all that said, I see enormous potential for Figma Make to lead this entire space. It already does what no other AI builder has managed — combining design logic, AI generation, and iterative product structure into one cohesive environment. The advantage Figma Make holds is its direct link to Figma’s broader design ecosystem, history, and database of design best practices. That foundation gives Make a runway to become the most powerful AI-native design-to-product platform on the market.

I genuinely believe Figma has everything needed to take over this niche, not just as a prototyping tool, but as the standard AI builder for real production apps. I’ve built, tested, and shipped multiple functional products with Make and would love to contribute back to its evolution — whether through feedback, testing, or sharing how I’ve integrated it into my own product development pipeline.

Thank you for providing open prompts and for continuing to improve this ecosystem. Please feel free to reach out if you’d like more details about my builds or my process — I’d be happy to collaborate or offer more insight.

I am not experiencing the bells and whistles. At all.

No matter how much Design System I’ve fed it, pasted it, updated it, and Guidelines.MD editing it is still hallucinating. 

Absolute nightmare one year on. 

Still require absolutely minute, basic prompt baby-sitting. Even with screenshots, guided instructions “move this here”. Still cannot action even that. 

Try fixing a table to even remotely match the source of truth from Canvas. Good luck with that. I’ve wasted 500 credits and it’s still hallucinating.