--- schema: 1 type: video title: 'How We Got LLMs to Recommend Our Open Source Library — Christopher Burns, Inth' description: >- Their onboarding form asks how you heard about us. On April 13th the answers started spiking, and the single largest source of inbound for c15t is now an LLM telling someone to install it. Christopher Burns is not a researcher and says so twice. He founded Inth, built c15t, the open source consent banner library, and reckons he has been hacking on this only slightly longer than the room has. The Collison brothers used to install Stripe by taking your laptop off you; going through Y Combinator, Burns found himself handing people a prompt instead. Good developer experience primitives turned out to be agent primitives. No single trick covers it, so the optimizations got abstracted into a framework neutral docs pipeline that generates the agent facing files from MDX. The rest is practical. Write llms.txt by hand rather than generating it, because forty good lines beat a thousand lines of noise. Agents fetch, they do not browse, so hand them links and a line on what each page is for. Serve markdown instead of HTML, three ways, since not every agent can set a header: a .md suffix, content negotiation, and a query parameter. The part Burns thinks matters most: coding agents mostly never open your documentation site. They read the repository and node_modules, working from stale training data and compiled source. Ship bundled markdown and an AGENTS.md inside the package and he measures close to half the tokens saved. He closes on a caution he applies to his own slides: the ground moves weekly and nothing stays perfect. Speaker info: - https://x.com/burnedchris - https://www.linkedin.com/in/burnedchris - https://github.com/burnedchris - https://burnedchris.com Timestamps: 0:00 - Not a scientist, just hacking on it 2:23 - The spike, and where the inbound came from 4:20 - No single fix, so they built a docs pipeline 5:52 - Write llms.txt by hand, not generated 7:07 - Ship markdown instead of HTML 9:17 - Web MCP: letting an agent ask your docs 10:09 - Agents never visit your site, they read node_modules 12:53 - Testing whether your site is agent ready 14:36 - Q&A: where to start on a plain website --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/V_5bn4q-vAI-youtube-eafc26c08ed0d88bd3c0 Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- canonical: >- https://liquidgold.cc/channel/aidotengineer/explore/video/how-we-got-llms-to-recommend-our-open-source-librar-b81775e58efc markdown: >- https://liquidgold.cc/channel/aidotengineer/video/how-we-got-llms-to-recommend-our-open-source-librar-b81775e58efc.md llms: null parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T04:39:46.251Z' language: null children: [] videoId: V_5bn4q-vAI durationSeconds: 987 publishedAt: '2026-08-26T15:30:07.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 24 endSeconds: 150 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=24' heldReason: rights - ordinal: 1 text: null startSeconds: 150 endSeconds: 162 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=150' heldReason: rights - ordinal: 2 text: null startSeconds: 162 endSeconds: 261 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=162' heldReason: rights - ordinal: 3 text: null startSeconds: 273 endSeconds: 353 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=273' heldReason: rights - ordinal: 4 text: null startSeconds: 365 endSeconds: 434 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=365' heldReason: rights - ordinal: 5 text: null startSeconds: 434 endSeconds: 448 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=434' heldReason: rights - ordinal: 6 text: null startSeconds: 448 endSeconds: 568 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=448' heldReason: rights - ordinal: 7 text: null startSeconds: 580 endSeconds: 615 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=580' heldReason: rights - ordinal: 8 text: null startSeconds: 626 endSeconds: 774 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=626' heldReason: rights - ordinal: 9 text: null startSeconds: 626 endSeconds: 774 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=626' heldReason: rights - ordinal: 10 text: null startSeconds: 785 endSeconds: 883 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=785' heldReason: rights - ordinal: 11 text: null startSeconds: 893 endSeconds: 987 watch: 'https://liquidgold.cc/watch/V_5bn4q-vAI?t=893' heldReason: rights --- # How We Got LLMs to Recommend Our Open Source Library — Christopher Burns, Inth Their onboarding form asks how you heard about us. On April 13th the answers started spiking, and the single largest source of inbound for c15t is now an LLM telling someone to install it. Christopher Burns is not a researcher and says so twice. He founded Inth, built c15t, the open source consent banner library, and reckons he has been hacking on this only slightly longer than the room has. The Collison brothers used to install Stripe by taking your laptop off you; going through Y Combinator, Burns found himself handing people a prompt instead. Good developer experience primitives turned out to be agent primitives. No single trick covers it, so the optimizations got abstracted into a framework neutral docs pipeline that generates the agent facing files from MDX. The rest is practical. Write llms.txt by hand rather than generating it, because forty good lines beat a thousand lines of noise. Agents fetch, they do not browse, so hand them links and a line on what each page is for. Serve markdown instead of HTML, three ways, since not every agent can set a header: a .md suffix, content negotiation, and a query parameter. The part Burns thinks matters most: coding agents mostly never open your documentation site. They read the repository and node_modules, working from stale training data and compiled source. Ship bundled markdown and an AGENTS.md inside the package and he measures close to half the tokens saved. He closes on a caution he applies to his own slides: the ground moves weekly and nothing stays perfect. Speaker info: - https://x.com/burnedchris - https://www.linkedin.com/in/burnedchris - https://github.com/burnedchris - https://burnedchris.com Timestamps: 0:00 - Not a scientist, just hacking on it 2:23 - The spike, and where the inbound came from 4:20 - No single fix, so they built a docs pipeline 5:52 - Write llms.txt by hand, not generated 7:07 - Ship markdown instead of HTML 9:17 - Web MCP: letting an agent ask your docs 10:09 - Agents never visit your site, they read node_modules 12:53 - Testing whether your site is agent ready 14:36 - Q\&A: where to start on a plain website --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/V_5bn4q-vAI-youtube-eafc26c08ed0d88bd3c0 Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- Canonical: https://liquidgold.cc/channel/aidotengineer/explore/video/how-we-got-llms-to-recommend-our-open-source-librar-b81775e58efc · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.