Skip to content
MakerGain
← Back to project

Pieter Levels Says Wayfair Opened 8 Accounts and Used His Interior AI for Several Months

Edited by MakerGain

Pieter Levels claims Wayfair opened 8 accounts on Interior AI and used it for several months; the invoice table he posted contains 10 paid records of 29 or 39 US dollars, and he judges that the company used it for competitive reconnaissance before building its own tool. Interior AI lets users upload interior photos and restyle them from the original image; he says monthly revenue reached 10,000–20,000 US dollars within one week of launch.

Pieter Levels posted an invoice table with ten entries. The customer column is all @wayfair.com, the status is marked Paid, and the amounts are 29 US dollars or 39 US dollars. He wrote that Wayfair had opened 8 different accounts on his interior design tool Interior AI and used them for several months; in the post, he said that in the week he published it, a Wayfair employee asked to cancel a subscription (levels.io[1]).

What Interior AI does, he describes very simply: upload an existing interior photo, use image-to-image to ask the model to change the space into a specified style—his example is maximalist design—and what comes back is an image of the same room altered from the original (Lex Fridman interview[2]).

What puzzled him was this: when he posted, he said Wayfair had already launched its own AI interior design tool the previous year, so why had it come and signed up for his? His own explanation is that the other party used his product for research before making its own version—he calls this competitive reconnaissance. This is the judgment of the person involved; the invoice table he attached does not itemize which plan each record corresponds to or how long a period it covers. The figure of 8 accounts comes from what he says in the body of the post.

According to the invoice table he posted, these records show as paid.

For example, an illustrative task: staging an empty room—upload a photo, specify a style, wait for the model to alter it from the original, and get an image that lets a person viewing the home see “what it would be like to live there.”

Image: Source Interior AI actual image source[3], retrieved 2026-11-10, original interface preserved.

Starting from “getting bored of it”

How the product came about, Pieter himself describes as quite accidental. He first made a website called This House Does Not Exist, continuously generating house images, and he got bored of it—in his words, everything started to look the same—but at the same time he found that this approach worked especially well on interior images, so he switched to interiorai.com. When talking about Interior AI, he also said it and the later Photo AI were “basically the same codebase” (Lex Fridman interview[2]). The title of the publicly released build record is also blunt: built from scratch in five days (levels.io[4]).

He uploaded a photo of his own home, asked it to turn maximalist, and the result was very good, so he judged this was a startup opportunity. Only afterward did he start learning fine-tuning: collecting beautiful interior design images into a gallery, training a model better at interior design, and putting it back into the product (Lex Fridman interview[2]). The order is worth noting: launch first, add the model later. It was not that a specialized model came first and then a use case was found; rather, an unexpectedly useful output came first, and then he went back to add the model.

The product form also grew along this line: upload a photo, describe the desired style, get back an image altered from the original.

The revenue he describes, and the undisclosed costs

In the interview he said that within one week of launch, Interior AI reached 10,000–20,000 US dollars in monthly revenue; at the time of the interview, about 40,000–50,000 US dollars per month. He also said the product had been live for about two years (Lex Fridman interview[2]). This is self-reported monthly revenue, not profit. There is no public data on costs such as model calls, maintenance, refunds, and taxes, so how much this business actually makes remains unclear.

Another path: selling what is in the picture

On the same track, ReimagineHome pushes design one step further toward transactions. On pricing, the Pro tier is 36 US dollars per month and includes 200 credits per month; new users can do 5 designs for free, and the pricing page states that no credit card is required (reimaginehome.ai/pricing[5]).

The difference is in the deliverable. It claims that, without changing walls, windows, or doors, it can design a space using real products nearby and within budget, add the whole set to a shopping cart, and then check out on Amazon or Shopify; products are sold by third-party retailers, payment, delivery, and returns are handled by the store, and it takes a commission from qualifying purchases without adding a markup for users. For real estate professionals, it can also upload up to 50 photos at once, classify them by room, and export a high-resolution gallery (reimaginehome.ai[6]).

What ReimagineHome adds is a transaction layer: a product library connects the design image to a checkout-ready cart, and commission comes from qualifying purchases rather than stopping at the persuasion stage. But this is not mutually exclusive—it also does pure design generation. For a real estate agent or interior designer, the dividing line is which step one wants to solve: to get a client to accept a combination, one image may be enough; to turn that combination into a purchase order, a product library and commission structure are needed.

Looking back at this episode, what is most worth recording is not the drama of “a big company peeking at a small tool.” Interior AI's path is more specific: first there was an output that unexpectedly worked well on interior images, it launched first, and later a specialized model was added. It places the task in a concrete action—upload a photo, specify a style, get back an image altered from the original.

Continue reading

AI companionsAI imagesAI audio
One reply counts as one, a photo as eight: Embraces.AI breaks companionship into messages and sells it

Embraces.AI prices companionship by message: one reply counts as 1 message, and one Picture this image counts as 8. Companions have cross-session memory, send multiple bubbles in a row, and initiate contact; users can edit memories and export chats. Before signing up, there are 40 free messages; the three paid tiers start at $12, and $5 tops up 250 messages. TrustMRR recorded on 2026-11-14 all-time revenue of $48,193 and 426 active subscriptions.

AI imagesAI videoAI audio
59 Million Views but Almost No Revenue: After Publishing the Entire Project, Overseas Signups Increased to 6.5x on the Day

Zohar Vanunu used his self-built Kolbo.AI to make a “biblical phone imagery” AI series, which drew 59 million views and more than 200,000 followers in 30 days, but by his own account it brought in almost no revenue; after publishing the complete production tutorial, overseas signups increased to 6.5x on the day, and in three weeks 3,285 people accounted for 78% of new signups. The product puts more than 100 models, along with chat, image, video, music, and voice, into the same project and charges by credits.

AI publishingAI writingAI imagesAI video
Do Not Start From a Blank Page: CaptureFlow Catches What You Say; the Signature Is Still Yours

CaptureFlow catches expression from voice, links, video, or files and hands it to Content Agent to draft; only after the user revises and explicitly approves does it publish to LinkedIn, and responsibility remains with the person. Chris Koronowski runs the company in Poznań, Poland, charging in three tiers—Pro, Business, and Agency—with the latter two priced by seat; a case study says InStream has 100 paid seats, while TrustMRR separately shows 8 subscriptions, and the two cannot be combined into a customer count.

AI videoAI audio
A pizza maker who signed a ten-thousand-dollar service contract before the landing page was ready

Paris pizza maker Kevin Rousseau taught himself to code during COVID, single-handedly made the dubbing tool VoiceCheap profitable by the account given in VoiceStudio's official introduction, and built on top of it the human-gated localization service VoiceStudio: clients hand over source videos, and the team completes transcription, translation, dubbing, quality check, and delivery, with key milestones confirmed by the client. The official account says the first service contract, for about $10,000, was signed before the landing page was completed.