Meet the First AI-Powered QR Code Assistant

On February 25 I wrote a post promising an AI Assistant "coming very soon." It is now May 12. That is twelve weeks, and that gap is on us.

Today the AI Assistant is live in beta at /en/create. No waitlist. No signup wall. You can use it right now.

Here is what it does, what it does not do yet, and where we are taking it.

---

What shipped today

The beta is an intent parser. You describe what you need in plain language, and the assistant maps your words to a pre-filled QR generator form.

It handles the most common creation requests:

• WiFi — "wifi QR for SSID GuestNet password Hunter2" → opens the WiFi builder with SSID, password, and encryption pre-filled
• URL — "link QR for my Shopify store homepage https://mystore.com" → opens URL type, link pre-filled
• vCard — "contact QR for Sarah Chen, Product Manager at Acme, sarah@acme.com +31 6 12345678" → opens vCard builder with all fields
• Text — "plain text QR that says Table 4 of 12, outdoor terrace" → opens text type, content pre-filled

Type in Dutch, French, German, Spanish, Italian, or Portuguese — the assistant understands all seven languages QR-Verse supports.

If it cannot parse your request, it tells you exactly what to try instead. No silent failures.

---

What this is not (yet)

The February post described multi-turn conversation, automatic quality checks, visual troubleshooting, and natural language analytics. None of that shipped today.

What shipped today is the one thing that resolves the credibility gap: you can now create a QR code by typing a sentence instead of filling a form. That is the core of the original promise, and it is live.

The broader capabilities — conversation memory, quality assurance, analytics in plain language, bulk creation via chat — are on the roadmap. I will not give another "very soon" date until engineering has a build in internal testing.

---

Why it took twelve weeks

Two reasons, and both are honest.

First, we shipped six unrelated production fixes during that window — the Sentry tunnel stall, the pricing page render regression, the Spanish locale typo, the payment failure spike. Each one was the right call at the time, but together they crowded out the AI Assistant build.

Second, the original post over-described what was shipping. Writing "the world's first full AI assistant for QR codes" when we were building an intent parser created a gap that felt too large to ship across. That framing was a mistake, and it is why this post leads with what actually shipped rather than what is planned.

---

The beta is free

No upgrade required. The intent parser is available on the free plan — five requests per day for unauthenticated users, twenty-five per day on Pro, fifty per day on Business.

If you hit the daily limit, the standard form-based generator is always there. The AI input is a shortcut, not a replacement.

---

What comes next

We are measuring two things in beta:

1. Parse accuracy — what percentage of requests does the assistant correctly map to form fields? Target is above 85% within 30 days.
2. Conversion to QR creation — of users who type a request, what share completes the QR generation? If it is meaningfully higher than the form-only baseline, we expand the feature surface.

If parse accuracy stays below 60% after iteration, we will say so publicly. If it hits 90%, we ship the next capability: multi-step creation where you refine the result by typing follow-up corrections.

The GS1 Digital Link and EU Digital Product Passport use cases are next on the creation roadmap — structured regulatory QR codes are exactly where natural language input reduces the most friction. More on that shortly.

---

Try it and tell us what breaks

Go to /create, use the AI input, and if it fails — screenshot it and send to contact or reply to this post. The fastest way to improve parse accuracy is real user input.

We shipped this to stop the credibility gap from widening. Now we need to earn trust by iterating in public.

— Marc