whats-coming-next

---

title: What we're exploring next excerpt: Honest roadmap, not a sales pitch. Some things we're investigating, some things we explicitly aren't. tags: [roadmap, future]

I'm going to skip the standard roadmap theater. No quarterly plan, no fake confidence about ship dates. Here's what I'm actively investigating, what I'm passively watching, and what I've decided we're not building. If something is real and shipping, it lives at the studio already.

Investigating: AI lock chime and Boombox sound generator

Tesla supports custom LockChime.wav files on a USB drive. The constraints are tight: WAV format, 5 seconds max, under 1MB, filename case-sensitive. The Boombox folder accepts WAV or MP3 for the external pedestrian sound. Most owners I've talked to don't know this feature exists, which is the same observation that led to FrunkLab in the first place.

I spent a couple of weeks looking at the AI audio space. The most promising option is ElevenLabs Sound Effects V2 at around $0.008 per generation, 48kHz WAV output, with a commercial license on paid plans. Replicate plus Meta's AudioGen is cheaper at the model level but ships under a non-commercial license, which kills it for a paid product. Stability AI's Stable Audio 2.5 is interesting but pricing is opaque. Self-hosted AudioCraft is the cheapest at scale and the most expensive in engineer-hours.

The technical work is small (text prompt in, validated WAV out, enforce the 1MB and 5-second caps server-side). The product question is the harder one: do owners actually want a custom lock chime, or is this just a feature I personally want? I haven't decided. If we ship it, it'll fit inside the existing AI usage budget on each tier; the per-generation cost is small enough that even a Pro user generating 20 sounds a month costs us pennies. See pricing for the current tier limits.

Status: researched, not committed. If you'd use this, tell me.

Investigating: more vehicles, as the source material catches up

Right now the 3D preview works for Model 3 (Pre-2024 and Highland), Model Y in several variants, and Cybertruck. Model S and Model X are 2D only because the upstream OBJ models with usable wrap UVs don't exist for them yet. I'm not going to fake a 3D preview on the wrong body shell just to fill the grid.

When Tesla expands wrap support to more vehicles, or when the open source modeling community ships better Model S/X assets, we'll add them. This is mostly a waiting game, not an engineering one.

Investigating: a "preset packs" idea

This one is half-formed. The thought is curated AI prompt packs by theme (cyberpunk, vintage racing, minimal monochrome) so a user can apply a coherent visual style without writing prompts from scratch. The studio AI already does this in single shots; the open question is whether bundling them adds value or just adds clutter. I'm watching how people use the studio and which prompts get re-run most.

Passively watching: mobile app

Web works fine on phones (after a lot of pain getting there). A native iOS or Android app would mean app store overhead, separate auth flows, and a second codebase, in exchange for marginal UX gains. The case for a native app gets stronger if we add features that benefit from camera or AR. As of today, it doesn't.

Status: not on the list. If web stops being good enough, this changes.

What we're not building

This list matters more than the investigating list, because the temptation to add features is constant and most of them are wrong.

• Engagement email upsells. No "you haven't designed in 3 days" emails. No "your friend liked a wrap" digests. Transactional email only (approvals, password resets, sub changes). My inbox has enough garbage and so does yours.
• NFT or blockchain anything. No, thanks.
• Generic AR features. Pointing your phone at your car to "see" the wrap in AR sounds great in a marketing video and works badly in real life. The 3D preview already shows you the wrap on a clean studio render of your exact model. AR adds lighting noise, tracking jitter, and a permission prompt.
• Social-network-style feeds. The gallery is a gallery, not a feed. No follows, no algorithmic ranking, no streaks, no notifications. You can like a design and fork it. That's the whole social model and I'd like to keep it that way.
• Cosmetic tier gates. A Pro subscription gets you more AI generations and a few quality-of-life features. It does not wall off cosmetic-only features that are arbitrarily restricted from free users. If a feature is shipped, the core version works for everyone.
• AI-generated brand logos or copyrighted IP. The AI generator declines obvious brand and IP prompts. Not a roadmap item, a permanent product decision.

What "exploring" actually means

When I say "investigating" I mean I've done research, I've talked to owners, and there's a non-zero chance we ship it. When I say "watching" I mean I've thought about it and decided to wait. When I say "not building" I mean it's a no, and changing my mind would require a real reason, not just user request volume.

There's no public ship date for anything in this post. If something ships, it'll show up in the studio. If it doesn't, you won't hear about it again, which is also fine.

If you want to push something onto the list, the best way is to use the studio and tell me what's missing.