
Promptless
An AI teammate that automatically updates your customer-facing docs.
Thesis
- 01
AI coding ships 5–10× more code. Docs don't follow. GitHub Copilot users showed 12–15% higher activity; 73% of OSS respondents use AI for coding or docs.[9] Code velocity is decoupling from doc velocity. The outer loop — keeping the README, the changelog, the API reference, and the integration guide aligned with the diff — is the new bottleneck.
- 02
Docs are the highest-leverage marketing for a dev tool. Mintlify raised $18M Series A from a16z to capture the docs surface; >20% of the most recent YC batch ships on Mintlify.[2] Docs are how every dev tool gets discovered, evaluated, and adopted. The CMS war is settled; the content war just started.
- 03
Coding agents now read docs to recommend products. "People would come to our booth and tell us that they'd learned of Vercel because ChatGPT told them to use Vercel."[8] LLMs ground on the cleanest, most current docs. The llms.txt standard (Jeremy Howard, September 2024) is the formalization.[13] Stale docs aren't a UX problem — they're a distribution problem.
- 04
The system of work, not the editor. Mintlify, ReadMe, and GitBook are publishing platforms.[2] [5] [6] Promptless is the AI teammate that proposes the change. The wedge isn't where docs live; it's who writes the next version when the code ships.
Problem
"Someone should update the docs." The most important non-urgent task at every software company.
Every engineering team has the same Slack message. It is the most important non-urgent task at every software company, and the one that always loses to a shipping feature. The result is the universal experience of reading documentation: examples that don't compile, parameters that don't exist, screenshots from a UI shipped three months ago, and a quickstart that contradicts the API reference two pages later.
The traditional answer was a technical writer chasing PRs in Linear. That model breaks at modern velocity. Engineering teams now ship multiple times a day. AI coding assistants are about to take the same teams from one PR per developer per week to ten.[9] No technical writer in the world keeps up with that flow rate, and most teams don't have one to begin with — docs maintenance is a tax paid by the founding engineer at 11pm.
The cost compounds in three directions. Customer support tickets balloon when the docs lie. Sales loses deals to demos that don't match the integration guide. And every coding agent grounding on stale docs recommends the wrong API call, the wrong product, or no product at all. The docs were always a moat. They are now an acquisition channel.[8]
73%
of OSS devs use AI for coding or docs
GitHub Octoverse 2024 — code velocity now ahead of docs velocity
12–15%
activity lift for daily Copilot users
More PRs per engineer, no proportional rise in docs output
20%+
of the latest YC batch on Mintlify
The CMS war is settled. The content war just started.
GitHub Octoverse 2024[9] · Mintlify Series A announcement[2]
Why Now
AI is breaking the docs the same way it broke search.
Three trends collided in the same eighteen months: AI coding made the docs gap permanent, LLMs became the new product-discovery layer, and llms.txt formalized the channel. The AI teammate that owns the docs surface owns the channel.
People would come to our booth and tell us that they'd learned of Vercel because ChatGPT told them to use Vercel.
Guillermo Rauch[8]
CEO · Vercel
I don't love doc writing, but holy ****, it is so much better when I have Promptless.
Aaron Levin[11]
Founding SE · Vellum
I trust Promptless more than my team's technical writer.
Hunter Spinks[11]
VP Product · FIS
Three preconditions converged in the same eighteen months.
AI coding tools went mainstream. 73% of OSS respondents now use AI for coding or docs; 1M+ open source maintainers, students, and teachers received free Copilot access; daily users show 12–15% higher activity.[9] The cohort writing more code per engineer is also the cohort writing less docs per engineer. The asymmetry is the opportunity.
LLMs became the new search engine for developers. Guillermo Rauch reports exponential signup attribution from ChatGPT recommending Vercel — booth attendees tell him the LLM sent them.[8] The implication is structural: docs are the new SEO. The product whose docs are clean, current, and machine-readable gets recommended. The product with stale docs is invisible to the next generation of buyers.
The llms.txt standard formalized the channel. Jeremy Howard's September 2024 proposal makes it explicit — websites should ship LLM-readable markdown, not just HTML.[13] When the distribution channel becomes a structured surface, the product that auto-maintains the structured surface owns the channel. Promptless is that product.
People would come to our booth and tell us that they'd learned of Vercel because ChatGPT told them to use Vercel.
How It Works
Monitor the system of work. Draft inside the existing CMS. Review like a pull request.
The integration surface is the moat.
Most docs tools own the editor. Promptless owns the workflow into the editor. Every integration — GitHub, Slack, Linear, Jira, Zendesk, Intercom, the CMS — is one more reason the system stays in place when the docs platform underneath changes. Each new signal raises the floor of the draft and lowers the marginal cost of the next docs surface.[1]
No prompt engineering. No new tool to adopt. Promptless is positioned as an AI teammate, not a docs co-pilot. The team's existing workflow stays intact; the new collaborator shows up in the existing review surface. This is the Grammarly-next-to-Word pattern: the platform doesn't change, the floor of the output rises.[4]
CMS-agnostic by design. The docs CMS market is fragmenting — Mintlify, ReadMe, GitBook, Document360, Notion, Confluence — and every team has a different stack. Promptless is positioned to be the constant across the stack: the AI teammate that writes into whichever CMS the next customer is on.[2] [5] [6] [7]
Docs Are the Highest-Leverage Marketing
The buyer thinks they're paying for writer hours. They're actually buying a distribution channel.
The wedge looks like a productivity tool. The compounding return is a distribution channel. Three customer patterns make this concrete.
Three customer patterns, one structural thesis.
Vellum — the customer-visible answer. Aaron Levin, founding solutions engineer, was hitting the docs every time a customer asked an API question. Promptless made the customer-visible answer current. The work he was doing in support tickets now happens before the ticket exists.[11]
Latitude.sh — paid in support tickets, not writer hours. Latitude.sh's CTO told the founders Promptless "basically pays for itself." The ROI line item is reduced support volume, not writer FTE. The pricing logic is anchored on the symptom — broken docs — not the headcount.[11]
Runpod — the writer stays, the throughput multiplies. Runpod's senior technical writer credits Promptless with slashing time-to-first-draft. The existing writer kept their job, just shipped 5× more. The product raises the floor of the existing team; it doesn't replace them.[11]
I love your product. It works incredibly well and basically pays for itself.
Market
The densest buyer pool is already on the modern docs CMS.
Mintlify alone powers >20% of the most recent YC batch and reaches 20M developers annually.[2] The modern docs CMS layer (Mintlify, ReadMe, GitBook, Document360) sits underneath 400k+ companies maintaining customer-facing documentation.[4] Every one of those companies has the same problem on the same publishing layer. Promptless is the AI teammate the publishing layer was waiting for.
The medium-term ICP is every Seed-to-Series B technical company with a docs surface that customers depend on. They ship product weekly, can't afford a dedicated writer, and their head of DevX is the buyer. The early reference customers (Vellum, PostHog, FIS, Runpod, Latitude.sh) span both the YC-AI cohort and the Fortune 500 fintech motion — the same product wedges into both.
Every dev-tool company is one stale doc away from losing the next LLM recommendation. Promptless is the answer the publishing layer was waiting for.
Competitive landscape
Four neighbors. Promptless is the only one positioned as the AI teammate, not the docs platform.
Every incumbent owns the editor. Promptless owns the workflow into the editor. The CMS layer commoditizes over the next decade; the AI teammate composes across all of them.
Mintlify won the editor. ReadMe won the API hub. The AI teammate that writes into all of them — and watches the system of work behind them — is the next layer. That layer is Promptless.
Founder deep dive
Two founders who had already paid the docs-maintenance tax — at petabyte scale and at acquisition exit.
Founder & team
Risks & mitigations
What we're watching
References
- [1]Promptless — Product homepage
- [2]Mintlify — Series A announcement ($18M, a16z-led; >20% of latest YC batch on Mintlify; Anthropic, Perplexity, Cursor, Pinecone, Zapier customers)
- [3]Y Combinator — Promptless company profile
- [4]Y Combinator Launches — Promptless: An AI teammate that automatically updates your customer-facing docs
- [5]ReadMe — Product overview (interactive API documentation; Series C, Accel-led)
- [6]GitBook — Product homepage (documentation platform; $70M+ raised)
- [7]Document360 — Knowledge base platform overview
- [8]Sequoia Training Data — Guillermo Rauch on ChatGPT as fastest-growing acquisition channel for Vercel
- [9]GitHub Octoverse 2024 — AI coding tool adoption (73% of OSS respondents use AI for coding or docs; 1M+ free Copilot users; 12–15% activity lift)
- [10]Mintlify — Customer outcomes (Coinbase 20min → 60s docs updates; HubSpot 50% engineering reduction; Zapier 3× faster docs)
- [11]Orange Collective customer reference and public testimonials — Vellum (Aaron Levin), FIS (Hunter Spinks), Runpod (Mo King), Bazel (Alan Mond), Latitude.sh (Eduardo Soubihe), Prove (Nicholas DeWald)
- [12]Frances Liu — LinkedIn (ex-Cloudflare, OneSignal, Graft; Stanford CS PhD drop-out)
- [13]Jeremy Howard — llms.txt proposal (September 2024; standard for LLM-readable site content)
- [14]Anthropic — Claude.ai docs (case study: AI assistants ground product recommendations in current documentation)


