You’ve heard you should try Claude for character work. ChatGPT is apparently great for brainstorming. Gemini has that massive context window. Someone in your writing group swears by Mistral for poetry.
So you do the math. Twenty dollars a month here, twenty there, maybe another ten over there. Suddenly you’re spending more on AI subscriptions than you are on coffee, and you’re an author, so that’s saying something.
Adam D’Angelo looked at this problem from the other side of the equation and saw something most people missed: the subscription trap wasn’t just annoying for users. It was bad for the AI models themselves. Brilliant, specialized tools were being locked behind their own little walled gardens, and most people would never find them.
His solution was Poe, which stands for Platform for Open Exploration. It’s one app, one subscription, and access to virtually every major AI model on the market. Think of it as the streaming bundle for artificial intelligence.
From Facebook’s CTO to the AI Bazaar
D’Angelo isn’t a newcomer to building platforms where information finds the people who need it. At Phillips Exeter Academy, he and his classmate Mark Zuckerberg built a music recommendation program called Synapse. He went on to study computer science at Caltech, then became Facebook’s Chief Technology Officer in 2006, when the company was still young enough to be scrappy. He helped build the technical foundation for what would become one of the largest platforms on the internet.
In 2009, he left Facebook to co-found Quora, a question-and-answer site built on a simple belief: the world’s best knowledge lives inside individual people’s heads, and building the right platform could unlock it. Quora grew into one of the most respected knowledge-sharing sites online, powered by real experts writing real answers.
Then GPT-3 arrived, and D’Angelo noticed something. AI could now answer questions instantly, at virtually zero cost. The answers weren’t always as good as a top Quora contributor’s, but they were immediate. Instead of trying to bolt AI onto Quora (he considered it, decided the quality gap was too wide), he realized the chat format itself was the breakthrough. People didn’t want to search for pre-written answers. They wanted to have a conversation.
“A chat kind of experience, where you can write a question and then get an answer instantly from AI, was more likely to be the best paradigm,” D’Angelo has said.
But he also noticed something else. Every AI company was building its own consumer app, its own billing system, its own iOS and Android clients. It was duplicated effort on a massive scale. What if, instead, there was a single platform where all of these models could live? A place where users didn’t have to choose between them?
“We’re more like a game console or an operating system, or a web browser,” he explained in 2023. “If you are building an internet product, you want it to work on all the web browsers.”
Poe launched in late 2022. By January 2024, Andreessen Horowitz had invested $75 million. The bet: that the future of AI isn’t any single model, but the ability to move between many of them.
What Poe Actually Does for Authors
At its simplest, Poe gives you a single interface to talk to GPT-4, Claude, Gemini, DeepSeek, Mistral, LLaMA, and dozens of other models. You sign in once, and the whole buffet is in front of you.
That alone is useful, but it’s not what makes Poe interesting. What makes Poe interesting is what happens when you stop thinking about AI models as individual tools and start thinking about them as a team.
Comparison without commitment. Working on a tricky scene? Send the same prompt to three different models and compare the results side by side. Maybe Claude nails the emotional nuance but GPT-4 gives you a better plot structure. Maybe Gemini, with its enormous context window, catches a continuity issue the others missed. On any other platform, you’d need three browser tabs, three accounts, and three subscription fees to do this. On Poe, it’s one conversation.
Model-hopping for different tasks. You might start a brainstorming session with one model, switch to another for research (pulling in Poe’s web search bot to check facts), then bring in a third to draft the actual prose. Each model has strengths. Poe lets you use the right one at the right moment without breaking your flow.
Custom bots. Anyone can create a specialized bot on Poe by writing a set of instructions and choosing a base model. No coding required. Want a “Cozy Mystery Plot Generator” that knows the conventions of the genre? A “First Page Critic” that evaluates your opening against what agents say they’re looking for? You can build these in minutes, or browse the more than one million custom bots other users have already created.
Poe Apps. A newer feature that lets users build small AI-powered applications inside the platform. These go beyond simple chatbots, allowing integration with multiple models and custom logic. For authors, this could mean a worldbuilding tool that cross-references your character sheets, or a research assistant that synthesizes multiple sources into clean summaries.
The @-Mention That Changes Everything
This is Poe’s Purple Cow, the feature that makes you stop and reconsider what an AI chat platform can be.
In a Poe conversation, you can @-mention any bot or model, and it joins the thread. Not in a separate window. Not in a new conversation. Right there, in the same chat, with full context of everything that’s been said.
Picture this: you’re working through a complicated plot problem. You start with Claude, because it’s good at understanding character motivations. You lay out the situation, and Claude suggests three possible directions. Then you type @GPT-4 and ask it to evaluate which direction creates the most narrative tension. GPT-4 can see everything Claude said. It builds on that analysis, pushes back on one suggestion, and adds a twist you hadn’t considered. Then you @-mention a web search bot to verify that the historical detail your twist depends on is actually accurate.
Three models, one conversation, each one picking up where the last left off.
For authors, this is genuinely new. You’re not switching tools. You’re not copying and pasting between windows. You’re conducting a conversation where different AI perspectives build on each other in real time. It’s like having a writing group where each member has a different specialty, except this group is available at 11 PM on a Tuesday and nobody needs a snack break.
The Creator Economy (and What It Means for You)
Poe isn’t just a place to use bots. It’s a place to build and sell them.
In April 2024, Poe launched a creator monetization system. If you build a bot that attracts users, you earn revenue in two ways: a share of subscription revenue when users upgrade because of your bot, and a per-message fee you can set yourself.
This might seem like a detail for developers, not authors. But it has a practical consequence: it means there’s a financial incentive for people to build really good specialized bots. The writing-focused corner of Poe’s bot ecosystem is growing because creators can actually earn from building tools that authors want to use.
Browse the platform and you’ll find bots tuned for genre fiction, nonfiction structure, poetry critique, query letter feedback, and book marketing. The quality varies (as it does on any marketplace), but the best ones are genuinely useful, and new ones appear regularly.
The Honest Tradeoffs
Poe solves a real problem, but it makes some tradeoffs to do it.
You’re one layer removed from each model. When OpenAI releases a new feature for ChatGPT (like Projects or Canvas), it doesn’t automatically appear in Poe. You get access to the models themselves, but not always the native tools each company builds around them. If you want ChatGPT’s Projects feature or Claude’s Artifacts panel, you’ll need to use those platforms directly.
Message limits exist, even on paid plans. The free tier gives you around 100 to 150 messages per day, with tighter limits on premium models. The $5 Budget plan expands that, and the $19.99 Premium plan expands it further, but you’re still working within a points system. Heavy users doing all-day writing sessions may bump into ceilings.
No manuscript management. Like every general-purpose chat AI, Poe doesn’t know what a chapter is. It has no outline view, no story bible, no word count tracker. It’s a conversation platform, not a writing studio.
Quality varies on custom bots. With over a million custom bots on the platform, the range runs from brilliantly tuned to barely functional. Finding the good ones takes some browsing and experimentation.
It requires internet. No offline mode. If you write at a cabin in the woods (respect), you’ll need a connection.
Who This Is For (and Who It Isn’t)
Poe is ideal for authors who want to experiment with multiple AI models without paying for each one separately. If you’ve been curious about Claude but don’t want to commit to a subscription, or you’ve heard Gemini is good for long-context work but aren’t sure it’s worth the switch, Poe lets you try everything for the price of one membership.
It’s also a strong choice for authors who value flexibility over depth. If your AI use varies from day to day (brainstorming on Monday, research on Wednesday, marketing copy on Friday), being able to switch models based on the task is a real advantage.
Poe is less ideal if you want the deepest possible experience with a single model. ChatGPT’s native Projects feature, Claude’s Artifacts panel, and Gemini’s integration with Google Workspace are all things you can only get by going directly to those platforms. If you already know which model you prefer and use it heavily, a direct subscription may serve you better.
And if you’re looking for a purpose-built writing tool with story structure, chapter management, and AI designed specifically for fiction, tools like NovelCrafter or Sudowrite are built for that. Poe is a conversation platform, not a novel-writing environment.
The Bottom Line
Adam D’Angelo spent years building a platform where the world’s knowledge could be organized and accessed through human experts. With Poe, he’s trying something similar for AI: a single place where you can find the right model for the right job, without the overhead of managing a half-dozen subscriptions.
The multi-model @-mention feature is the real draw. Being able to bring different AI perspectives into a single conversation, where each model builds on what came before, is something no other platform offers in quite the same way. It turns AI from a one-voice tool into something more like a collaborative discussion.
Is it the best place to use any individual model? Probably not. Each model’s native platform will always have deeper features and tighter integration. But if you want range, if you want to compare, if you want the freedom to reach for whatever tool fits the moment, Poe is the only platform that puts all of them in the same room.
For authors still figuring out which AI assistant fits their workflow, that might be exactly where to start.
