
An AI image generation service.
Midjourney shows up mainly as a model others copied or defined themselves against. Suno copied its describe-it-and-it-appears bot format, launched rough on Discord, and drew heavy use despite poor early audio. Midjourney also tested its free tier, once offering slow free generations and now requiring payment. FLORA users treating it as a Midjourney substitute use it 100 to 200x less. Lesson: ship rough, and track usage depth.
From Suno's Mikey Shulman on Midjourney
Midjourney showed four fast drafts because about one in four was good. Instead of making users wait minutes for a single image that would be bad three times out of four, Midjourney returned four quick low-res images at once, since roughly one in four came out well.
Gustav Söderström uses it as the model case for measuring how often your ML is right and choosing a UI to match: at a one-in-five hit rate, show about five options and give users an easy escape hatch to something familiar. A single big play button only works at near zero prediction error. "If you don't understand the performance of your machine learning, you can't design for it."
Midjourney's Discord bot became the template for AI startups launching fast. Suno modeled its first product on Midjourney's describe-it-and-it-appears bot: a crude Discord bot that turned a one-sentence prompt into 12.5 seconds of audio.
Discord meant no UI or auth to build, and users tried the product, gave feedback and told others in the same place. Mikey Shulman expected Suno to stay on Discord for a long time, as Midjourney did, but was wrong.
Midjourney experimented heavily with its free tier before requiring payment. It once offered slow free generations and now requires payment to generate.
Its mobile web is cited as the payoff of restraint and testing. Guillermo Rauch points to Midjourney's mobile web, which runs on Next.js, as proof that great mobile web comes from love, restraint and heavy testing.
His broader rule: say no nine times for every yes, because each feature grows into something you must maintain.
FLORA's most engaged users treat it as more than a better Midjourney. Users who adopt FLORA as a creative system show 100 to 200x more usage than those using it as a better Midjourney, so FLORA tracks nodes generated as its north star rather than weekly active users.
OpenArt faced the same backdrop: Midjourney, Stable Diffusion and DALL-E were already established at its launch, so it had to find a way to stand out.
Teams use Midjourney to cut mock-up work to a fraction of the time. Thrive Digital uses Midjourney and DALL-E for rough creative mock-ups in about 1% of the time, turning a week of work into an afternoon and re-prompting live on client calls.
Thrive has laid off no one; people still need to know what to ask and how to iterate. AI PMs also use it for logos, visual user stories and storyboards alongside prototyping tools.
Experimented heavily with its free tier, once offering slow free generations and now requiring payment.
Midjourney experimented heavily with its free tier before requiring payment. It once offered slow free generations and now requires payment to generate.
Suno modeled its first bot on Midjourney's describe-it-and-it-appears format. Shulman expected to stay on Discord for a long time, as Midjourney did, but was wrong.
Midjourney's Discord bot became the template for AI startups launching fast. Suno modeled its first product on Midjourney's describe-it-and-it-appears bot: a crude Discord bot that turned a one-sentence prompt into 12.5 seconds of audio.
Discord meant no UI or auth to build, and users tried the product, gave feedback and told others in the same place. Mikey Shulman expected Suno to stay on Discord for a long time, as Midjourney did, but was wrong.
Used as the contrast for FLORA: users who treat FLORA as a better Midjourney show far lower usage than those who adopt it as a creative system.
FLORA's most engaged users treat it as more than a better Midjourney. Users who adopt FLORA as a creative system show 100 to 200x more usage than those using it as a better Midjourney, so FLORA tracks nodes generated as its north star rather than weekly active users.
OpenArt faced the same backdrop: Midjourney, Stable Diffusion and DALL-E were already established at its launch, so it had to find a way to stand out.
Midjourney, along with Stable Diffusion and DALL-E, was already established when OpenArt launched, so OpenArt had to find a way to stand out in a crowded field.
FLORA's most engaged users treat it as more than a better Midjourney. Users who adopt FLORA as a creative system show 100 to 200x more usage than those using it as a better Midjourney, so FLORA tracks nodes generated as its north star rather than weekly active users.
OpenArt faced the same backdrop: Midjourney, Stable Diffusion and DALL-E were already established at its launch, so it had to find a way to stand out.
Runs on Next.js. Its mobile web is Rauch's example that great mobile web comes from love, restraint and testing.
Its mobile web is cited as the payoff of restraint and testing. Guillermo Rauch points to Midjourney's mobile web, which runs on Next.js, as proof that great mobile web comes from love, restraint and heavy testing.
His broader rule: say no nine times for every yes, because each feature grows into something you must maintain.
Helps PMs with logos and visual user stories, storyboarding ideas alongside prototype-building tools.
Teams use Midjourney to cut mock-up work to a fraction of the time. Thrive Digital uses Midjourney and DALL-E for rough creative mock-ups in about 1% of the time, turning a week of work into an afternoon and re-prompting live on client calls.
Thrive has laid off no one; people still need to know what to ask and how to iterate. AI PMs also use it for logos, visual user stories and storyboards alongside prototyping tools.
Showed four fast low-res images instead of one slow one because roughly one in four came out good, an example of fitting the interface to the model's hit rate. Its first use felt magical, as with DALL-E and Stable Diffusion.
Midjourney showed four fast drafts because about one in four was good. Instead of making users wait minutes for a single image that would be bad three times out of four, Midjourney returned four quick low-res images at once, since roughly one in four came out well.
Gustav Söderström uses it as the model case for measuring how often your ML is right and choosing a UI to match: at a one-in-five hit rate, show about five options and give users an easy escape hatch to something familiar. A single big play button only works at near zero prediction error. "If you don't understand the performance of your machine learning, you can't design for it."
Along with DALL-E, it produces creative mock-ups in about 1% of the time, cutting a week of mock-up work to an afternoon. Thrive has laid off no one.
Teams use Midjourney to cut mock-up work to a fraction of the time. Thrive Digital uses Midjourney and DALL-E for rough creative mock-ups in about 1% of the time, turning a week of work into an afternoon and re-prompting live on client calls.
Thrive has laid off no one; people still need to know what to ask and how to iterate. AI PMs also use it for logos, visual user stories and storyboards alongside prototyping tools.
AI assistantsAn AI research and product company behind ChatGPT, Codex and the Dots agent.
AI assistantsOpenAI's AI assistant app used for chat, plugins and agent work.
AI assistantsAn AI company that builds the Claude models and products such as Claude chat and Cowork.
AI assistantsAnthropic's AI assistant and model family, used for chat, coding and agentic work.
AI assistantsA voice AI company with text-to-speech, dubbing, speech-to-text and music models, plus consumer and creator apps.
AI assistantsGoogle's AI assistant app, available as a chatbot and moving toward a personal agent.
AI assistantsA Google Labs app that answers questions about your own sources and turns them into AI-hosted audio shows.
AI assistantsA voice-first AI companion app with an animated alien character that remembers your conversations.