Image: The Letter Company
Why it mattersGrowth work has run inside the product code base and taken attention from product engineers. A separate harness for it keeps the main repository focused on the product.
Growth work used to be a weekend task for whoever had time. Now it is a job. The work still runs on the same code base as the product.
The Letter Company has launched Letterpress-1 on Y Combinator. The company calls it the first coding agent, harness and IDE for growth engineering. The launch page carries 13 votes at the time of writing. The founder listed is Mathew Pregasen.
What it does
Letterpress-1 is a coding agent that edits a company's website. It opens pull requests for the changes. It can set up A/B tests. It can make SEO fixes. It can write landing-page copy. It can also build tools for AI agents that visit the site. The YC page says it is built for teams that use git. The product is reached through a web app, a Slack bot, a command-line tool, and an agent interface.
The Letter Company says the platform rests on five parts it has built for growth work. Proofs are evidence chains that show a customer outcome. Simulations are copies of the live site. A change can run in a copy before it reaches real visitors. Benches are tests that the agent runs on its own output. The tests can be hosted inside the company and repeated. Kernels are the writing steps that produce the on-page copy. Phantoms are automatic publications used to test a strategy.
Who is using it
Letter's home page at letter.dev lists more than 50 customer logos. Cloudflare is among them. Other named logos include Credal, Vellum, SpeakeasyAPI, Thunder, Zip, Oso, Extend and Flox. Letter has not published a customer count or a price. The home-page strap line reads "Forward-thinking companies build on Letter".
The pitch is that growth is now an engineering problem. The YC page says modern engineering tools are not built for it. The work reacts to outside signals. Those signals include how AI agents read the site. They also include how search rankings move. They also include what visitors do after they land.
Why a separate harness
Growth work and product work compete for the same reviewer time inside one repository. Letter says these two kinds of work want different repositories, different tests, and a different harness. A coding agent built for the growth side can then run with less supervision. The product repository stays focused on the product.
The YC page notes the agent can also build tools for other AI agents that visit the site. A buyer may send a shopping agent to the site. A search engine's crawler may send a reader agent to a product page. The site needs a shape those agents can read. Letter says Letterpress-1 can produce that shape.
Source
- Letterpress-1 on Y Combinator, The Letter Company's launch page
- letter.dev, the product site
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
