Lemma reads all of your agent conversations, finds the failures nobody knew to look for, and opens the PR that fixes them. ### Why this role exists Every team building agents in production has the same problem and most of them don't have language for it yet. Their agent fails silently, they find out when a customer churns, and they assume it's their fault for not writing better evals. Naming that problem clearly is most of our marketing.
The company that explains why agents fail becomes the company people call when theirs do. This role owns that. Not swag, not conference booths, not a newsletter nobody reads. Writing that engineers actually finish, built on real traces and real failures, published often enough to compound. ### What you'll do * Write the technical content: failure taxonomies, teardowns of real production incidents, honest posts about what we got wrong * Build things that demonstrate rather than describe.
Open source tools, reproducible failure cases, benchmarks people can run * Own our developer-facing surfaces: docs, quickstart, the first ten minutes of using Lemma * Be in the communities where agent engineers actually complain, and be useful there before you're promotional * Bring what you hear back to product. You'll often know what's broken before the team does ### What we're looking for * You're an engineer who writes, not a marketer who codes.
We'll ask for both * Published technical writing we can read. Blog, docs, talks, threads, doesn't matter where * You've built with agents in production and can talk about specific failures, not abstractions * Strong TypeScript, enough to ship demos and contribute to docs and product * Comfortable being early. There's no content engine here yet, no audience, no calendar. You're building all of it We move fast on hiring. Target is an offer within two weeks of first contact.
Onsite in San Francisco. We sponsor visas.
Job details are sourced from the employer's original posting.
Open job posting