Most founders will tell you, without much prompting, that they should be posting more. They have opinions. They have a company worth talking about. What they do not have is the twenty minutes and the nerve to write the thing, publish it, and sound like themselves while doing it. That gap - between wanting to show up online and actually doing it - is the whole business Typa is trying to occupy.
Typa, a company in Y Combinator's Summer 2024 batch, is an AI ghostwriter. That description is doing a lot of quiet work, because "AI writes your posts" is a crowded, slightly tired category by now. What separates Typa is less about the model and more about the manners. It learns how a specific person writes, keeps a running memory of what they've said and shipped, and drafts content for their approval. The last four words are the product.
Send it a voice note on the walk to the office, or paste a link to something you read, and Typa turns it into a post shaped for X, LinkedIn, Bluesky, or your newsletter. Nothing goes live until you read it and say yes. The agent behind this is named Puff, and you reach it the way you'd text a chief of staff - iMessage, Slack, WhatsApp, or an actual phone call.
The pivot that made the product
Typa did not start as Typa. The legal entity is still Weavel, Inc., and the team's first YC-era product was Ape - a tool for prompt engineering, aimed at the narrow, technical audience of people who write instructions for language models for a living. It was a reasonable idea for 2024. It was also a tool for engineers.
Somewhere in the batch, the founders noticed the questions they kept getting were not about prompts. Other founders wanted help sounding like themselves online. So they moved. The pivot from Ape to Typa is the kind of decision that looks obvious in hindsight and terrifying in the moment: throw out the thing you built, follow the thing people are actually asking for.
There's a lesson buried in that move worth stealing if you're building anything: your users will describe the real product to you, usually while asking for help with something else. The trick is being willing to hear it. The Weavel-to-Typa jump narrowed the audience from "prompt engineers" to "anyone who has to write online," which is a much larger and much more anxious group of people.
What Puff actually does
The mechanics are less exciting than the framing, which is a compliment. Typa runs on three plain ideas. It keeps a knowledge base that updates itself as you feed it links, notes, and past writing, so it isn't starting cold every time. It maintains a model of your voice - cadence, the words you'd never use, the ones you always do. And it wraps both in an agent you can reach through the apps you already have open.
How a stray thought becomes a published post
or link
in your voice
or edit
your channels
The approval step, the yellow box in that chain, is the part Typa refuses to automate away. It would be trivial to let the agent post on its own. Plenty of tools do. Typa's bet is that the moment content publishes without a human reading it first, you get sludge - technically on-brand, spiritually nobody's. Keeping a person in the loop is slower. It's also the reason the output can carry someone's name.
Who it's for, and where it fits
Typa points at three groups. Solo founders and creators who want to publish without it eating a day. Companies that need one coherent story reaching investors, customers, and recruits at once. And agencies, who use it to run content for a stack of clients without cloning their best writer four times over. The same engine, pointed at three different scale problems.
On the map of AI writing tools, Typa sits between the scheduling-and-repurposing crowd - Taplio, Hypefury, Postwise - and the human ghostwriting agencies that charge a retainer to sit in your voice for months. It's cheaper and faster than the agency, more opinionated about fidelity than the schedulers. The wager is that "sounds like me" is the feature people will actually pay for, and the hardest one to fake.
Illustrative - channel coverage across Typa's supported destinations
The compounding argument
The founders frame content as an asset that compounds - the more Typa knows about you, the better it drafts, and the more consistently you show up, the more the audience builds. That's a tidy story, and it happens to be true of the boring, unglamorous version of content advice everyone ignores: post like yourself, keep doing it for years, don't disappear for three months.
Whether Typa is the tool that makes the boring advice doable is the open question. The team is four people, the funding is early - backed by Y Combinator and Krew Capital - and the category is loud. But the design choice at the center is a real one. In a year when the internet is filling with content nobody signed, a ghostwriter that insists you sign is at least pointed at a problem worth solving.
Ten years from now, the interesting measure won't be how many posts Typa drafted. It'll be whether the people using it still sounded like themselves - and what they built with the hours they didn't spend staring at an empty compose box.