A children's book made with AI in 72 hours sparked huge backlash

Ammaar Reshi just wanted to make a bedtime story for a friend's daughter. Instead, he ended up in the crosshairs of the global debate over AI art.

Reshi, a product-design manager at a San Francisco fintech company, had zero background in illustration or creative writing. In December, he turned to OpenAI's ChatGPT to write "Alice and Sparkle," a children's story about a young girl learning about tech alongside her robot companion. He then fed prompts into Midjourney to generate the artwork.

Within 72 hours, Reshi had self-published the digital book on Amazon. By the next day, he was holding a physical paperback printed for free via Amazon KDP. His total out-of-pocket cost: zero, beyond his existing $30-a-month Midjourney subscription.

Thrilled with how fast it came together, Reshi shared the experiment on Twitter. The thread blew up with over 2,000 comments and 5,800 retweets. But the honeymoon lasted less than 24 hours.

Waking up to death threats at 4 a.m.

By the next morning, Reshi's phone was buzzing every two minutes with intense backlash, including messages calling him scum.

The outrage came primarily from artists who argue that text-to-image generators scrape copyrighted artwork without consent or compensation. Users frequently input living artists' names to mimic their specific styles. Tension was already running high following a viral update to Lensa AI that let people generate AI portraits, and Reshi's project became the lightning rod.

Midjourney founder David Holz pushed back on the outrage, telling Insider that very few images created on the platform are used commercially, and that training data comes from broad internet scraping where most files are just photos.

The book also took hits on quality. One Amazon reviewer called the writing stiff and complained that the artwork featured bizarre, mangled fingers. The influx of polarized five-star and one-star reviews triggered Amazon's anti-fraud systems, pausing sales from January 6 to January 14 due to suspicious review activity. Reshi had sold 841 copies before the pause.

Changing the plot to fix the AI

Reshi never claimed to be the real author, describing ChatGPT as his ghostwriter and Midjourney as his illustrator. Still, wrangling the AI took work.

He spent hours tweaking prompts to get a consistent look for the main character, Alice. But her robot buddy, Sparkle, was a disaster. The AI simply could not draw the same robot twice. To cover for Midjourney's limitations, Reshi had to add a line to the actual story explaining that Sparkle had the special power to transform into different shapes.

After reading through the feedback, Reshi added to his thread that artists need to be involved in shaping these tools and that their hard work deserves respect. He admitted he would be hesitant to make another illustrated book anytime soon, though he still plans to use ChatGPT for other projects.

Why it matters

Reshi's weekend experiment proved that a solo amateur can now take an idea from scratch to a commercially published physical product in three days for almost nothing. For individuals wanting to make quick, hyper-personalized gifts, the barrier to entry has evaporated.

However, it also exposed the raw friction between Silicon Valley's build-fast culture and working creative professionals. Selling over 800 copies of an AI-generated book trained on scraped artwork touched a massive nerve. Holz's defense that most uses are non-commercial rings hollow to artists watching commercial derivative works hit digital storefronts every day.

We think Reshi's hack of changing the story to mask the AI's rendering inconsistency was clever, but it shows how unready the tech was for truly coherent long-form storytelling. More importantly, it serves as a reality check: releasing AI-generated creative projects into public marketplaces is no longer just a fun tech demo; it is a cultural and legal minefield.