Alfred Presents · Artist Showcase alignDRAW (2015 / 2023) — ELMAN MANSIMOV

ELMAN MANSIMOV

AI pioneer, text-to-image · Ethereum · fellowship.xyz

In 2015 a nineteen-year-old taught a machine to draw from sentences — 2,709 blurry 32-pixel squares, the first images ever pulled out of text, minted on Ethereum eight years later.

Alfred, from the DAILY × Fellowship exhibition notes and the DALL-E paper, August 2026

Modern machine learning approaches to text to image synthesis started with the work of Mansimov et al. (2015)

— Ilya Sutskever, Mark Chen, Alec Radford et al., 'Zero-Shot Text-to-Image Generation' (DALL-E paper), OpenAI, 2021
alignDRAW, works
2,709
Floor
0.82ETH
Lifetime volume
2,510ETH
Recorded sales
3,622
Owners
517
Original work
2015
Minted on-chain
2023
All market figures from the OpenSea /collections/aligndraw/stats endpoint, measured 18 August 2026. Supply of 2,709 and contract 0x9cf0ab1cc434db83097b7e9c831a764481dec747 from the OpenSea collection endpoint on the same date. The original artworks were created in 2015; on-chain minting was November 2023 via DAILY in partnership with Fellowship.
Track Elman Mansimov in Alfred
Floor, sales and new listings for alignDRAW — as they happen.
The work · 2015 / 2023

The first text-to-image model, minted eight years after the fact

The first text-to-image model

alignDRAW by Elman Mansimov

By early 2015, neural networks had mastered the art of 'image-to-text' and could create natural language captions for images. Flipping this process, and turning text into image, was a much more complex challenge solved by 19-year old prodigy Elman Mansimov's alignDRAW model. The small 32 by 32 pixel outputs are significant as a proof of concept for a new form of communication between machines and humans, one that involves our own language instead of code.

DAILY × Fellowship, exhibition notes, daily.xyz/exhibition/10002, verified 18 August 2026
2,709 worksFloor 0.82 ETH2,510 ETH volume3,622 sales517 owners
OpenSea

A 19-year-old researcher at the University of Toronto built the proof of concept for text-to-image AI in 2015. The 32×32-pixel outputs — blurry and small — were the first images a machine ever generated from a sentence. They were minted as 2,709 NFTs eight years later.

One artist on this page. 46 collections in the median wallet we look at.

Every figure above comes from Alfred’s own dataset — sales, volume and owners read off the chain, floors from the marketplaces. On a wallet he goes one step further, to the number that gets buried under a decade of transfers: what you actually paid.

In the wallets we look at, nearly half of what people hold has no purchase price to look up. It arrived as a mint, a swap or a transfer.

Alfred follows each one back through the chain.

Both figures come from one reading: 16 wallets, 3,548 items, Sep 2026.

See what your collection cost you
Read-only by design — no wallet connection, no keys, no transactions. Ever.