← All AI Engineer talks

AI Engineer World's Fair 2026

How We Got LLMs to Recommend Our Open Source Library — Christopher Burns, Inth

Read the talk

Making c15t Easier for Agents to Discover, Understand, and Install

Christopher Burns explains how concise indexes, Markdown delivery, documentation tools, and package-bundled guidance helped agents work with c15t—and where the evidence remains provisional.

From a talk by Christopher Burns

At a glance

Ideas worth remembering

  • Optimize both discovery and use: concise indexes help agents find material, Markdown reduces the cost of reading it, and bundled documentation serves coding agents inside the installed package.

  • Offer Markdown through .md URLs, request-header negotiation, and mode=agent because clients differ in how they can request content. Discovery-header support remains uncertain.

  • Burns reports LLM recommendations becoming the largest inbound source and almost 50% token savings from local documentation. These observations support further experimentation, but the talk does not establish causal attribution or a reproducible benchmark.

  • For an ordinary website, begin with manually curated indexes and add Markdown pages where feasible. Treat readiness scores as evolving feedback and continue making concrete improvements.

Developer experience becomes an acquisition channel

Christopher Burns, founder of Inth and creator of the open source cookie consent library c15t, presents his work as practical experimentation. His questions are whether agents can understand the library and whether they can do so efficiently. He explicitly disclaims a scientific background and warns that the surrounding technology changes quickly.

The business context is a library designed for developers in a category Burns describes as serving marketers and lawyers. That positioning made a strong developer experience central to the product. He reports 45% month-on-month growth and 2.8 thousand production websites. His download figures are less clear: he first describes growth from 1.2 thousand downloads to nearly 2 million, then cites 3 million NPM downloads without explaining the difference.

The signal that motivated the talk came from onboarding responses asking how people discovered the product. Burns says those responses began spiking on April 13, and recommendations from Claude, ChatGPT, Codex, and Gemini became the largest source of inbound. This is evidence of users attributing discovery to LLMs; it does not isolate which documentation changes caused those recommendations.

0:190:21
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

0:01 · section reference included

Many small improvements, one documentation pipeline

Burns describes an accumulation of small improvements beneath the visible library: llms.txt, sitemaps, RSS feeds, and robots files. Each serves a different part of discovery or access. His premise is that no single tool fixes the whole problem, so familiar web infrastructure and newer agent conventions need to work together.

The installation path has also changed. Burns asks how many audience members have accepted an agent’s suggestion to install a library and observes many raised hands. He compares the Collison brothers’ hands-on Stripe installations with his own experience in Y Combinator, where he could give someone a prompt. In that workflow, the developer experience must support an agent carrying out the installation as well as a person reading instructions.

As the team built more documentation websites, it extracted these improvements into an open source, framework-neutral documentation pipeline. Its basic operation is to take MDX files and generate the additional files needed for agent access. This makes the repeated work reusable across sites. Burns reports that other developer companies were seeing similar results, but describes the system as documentation infrastructure rather than a guaranteed way to improve recommendations.

2:392:42
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

2:39 · section reference included

Give agents a concise entry point and a useful map

A documentation site with hundreds of pages creates a selection problem: which page will answer the agent’s question? Burns recommends writing llms.txt by hand, with deliberate attention to the answers an LLM needs. He reports that about 40 useful lines beat 1,000 lines of noise in the team’s testing. The practical decision is to curate the entry point instead of assuming that a larger generated file will be more helpful.

His second recommendation addresses retrieval. Burns characterizes the agents he is targeting as fetching resources rather than browsing through a site. He describes a fuller companion index that works like a sitemap: page links accompanied by short descriptions of what each page is for. Those descriptions help an agent decide what to fetch before spending tokens on the page itself. The concise introduction and the broader page map therefore serve different purposes.

5:596:02
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

5:59 · section reference included

Make Markdown available through three access paths

Once an agent selects a page, the representation it receives affects the cost of reading it. Burns calls HTML expensive and proposes a Markdown twin for each documentation page. His example is a Next.js quick-start page whose URL gains a .md suffix to return Markdown. The page can continue to serve its normal website experience while exposing a text representation for agents.

The Markdown version also needs to be discoverable. Burns describes a header that advertises an alternative Markdown representation when an agent visits the website. However, he is explicitly uncertain about which agents recognize it. Advertising an alternative creates an opportunity for a client to use it; the talk does not establish reliable support across clients.

Burns consequently recommends three ways to request the same content. A .md URL is explicit and easy to hand to an agent. Content negotiation lets the normal URL return Markdown when a request header indicates that the client accepts it; he describes configuring this behavior in Next.js. A mode=agent query parameter supplies another URL-based option because some agents cannot attach request headers. Supporting all three accommodates differences in how agents fetch content.

7:077:09
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

7:07 · section reference included

Let an agent ask the documentation

Burns next moves from fetching documents to interacting with them through tools. He describes early WebMCP support exposing three operations: search docs, get pages, and ask docs. These give an agent ways to locate relevant material, retrieve a page, or submit a documentation question. The pipeline brings the supporting context together, although Burns does not detail the implementation of the answering operation.

He treats this as an emerging direction and briefly speculates that agents may also communicate over email. That possibility extends his interest in making information accessible through agent interactions, but it is a forecast rather than an outcome demonstrated by this documentation system.

9:159:18
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

9:15 · section reference included

Put documentation where coding agents already look

For library authors, Burns considers the local package the most important surface. His observation is that coding agents often work from the repository and node_modules without visiting the documentation website. They combine potentially stale training knowledge with whatever they can infer from compiled source. A better website alone cannot resolve the information gap in that workflow.

The team’s solution is to ship bundled Markdown documentation inside the installed package, alongside an AGENTS.md file that points to it. This gives the agent a local route from a question to explanatory material. Instead of having to reconstruct intended usage from compiled code or locate the right documentation online, it can read the documents already present in its working environment.

Burns reports almost 50% token savings across multiple models from this approach, compared with spending effort searching the web and finding the right resources. He sees particular value for libraries that change frequently. The result is promising, but he supplies no task set, model list, or measurement protocol that would make the percentage reproducible or establish it as a general expectation.

He says the bundled documents already help without additional skills. Skills can nevertheless direct an agent to inspect node_modules, and project guidance can ask it to read the relevant bundles and verify that they match the library it is working with. This adds explicit direction to the local documentation path without making a skill a prerequisite for using it.

10:0410:06
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

10:04 · section reference included

Use readiness tests as changing feedback

Burns broadens the application beyond developer documentation. He also uses the package on the company’s marketing website, where every part has a Markdown file, and suggests that blogs or other Markdown-based sites can use the same approach. The pipeline generates the extra outputs from the material those sites already maintain.

Evaluation tools were still emerging as he prepared the talk. Burns mentions a Cloudflare readiness test and favors a tool called Aura AI that examines a website and produces recommendations. He shows a score of 59 and says it had been much higher three weeks earlier. Without a stable scoring basis or a breakdown of the changes, that number is best understood as feedback from a changing test rather than a fixed measure of agent effectiveness.

His closing advice is to keep improving as agents, models, and expectations change. He returns to his opening stance: these are lessons from working on the problem, not definitive answers. Small improvements remain useful even when a perfect or permanent implementation is unavailable.

12:1112:12
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

12:11 · section reference included

Where to start on an ordinary website

An audience member from a website agency asks which techniques matter when building a startup’s ordinary website rather than a developer tool. Burns first emphasizes providing a .md file for every page. His rationale is token efficiency as more agents access websites. He also identifies the implementation obstacle: many content management systems are not built to produce Markdown versions of their pages.

Burns says he built his own CMS, ChrisCMS, to support his needs, while joking about his team’s reaction. His practical starting order is less demanding: create llms.txt first, then the fuller companion index, maintaining them manually if possible. He still recommends those files when the underlying system does not use Markdown and suggests creating the additional files as needed. A site can therefore begin with a useful description and page map while working toward Markdown delivery across its pages.

14:3214:35
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

14:32 · section reference included

Read the complete timestamped transcript
  1. 0:01

    [music]

  2. 0:12

    >> The talk title, we'll see if it lines up

  3. 0:15

    by the end of it, but when we put this

  4. 0:17

    talk title in

  5. 0:19

    just to be honest with you, so much

  6. 0:21

    changes in like 3 days at this point.

  7. 0:24

    We'll see how it goes. So, yeah, the

  8. 0:27

    whole point of it was that

  9. 0:30

    how I got LLMs to understand my open

  10. 0:32

    source library and what I did to do it

  11. 0:36

    well. Is it some kind of scientific

  12. 0:38

    background? Am I from a lab? No.

  13. 0:43

    That's

  14. 0:44

    My slidey clicky thing's not working.

  15. 0:46

    So, I just like to say again, I'm just

  16. 0:48

    like you. I'm just this side of the

  17. 0:51

    stage. I've just hacking hacking it

  18. 0:54

    together, figuring out what is useful,

  19. 0:57

    what is token efficient, these kind of

  20. 0:59

    things. And again, I am British. Please

  21. 1:02

    don't think my accent makes me an

  22. 1:04

    expert.

  23. 1:07

    So, for quick contacts, I'm Christopher

  24. 1:09

    Burns. I'm the founder of Inth. I

  25. 1:12

    created a open source cookie banner

  26. 1:14

    library called C15T. That really

  27. 1:17

    annoying thing on the internet. That is

  28. 1:19

    me.

  29. 1:20

    Uh I spoke at Next Conf uh after it

  30. 1:23

    started taking off and it had 1.2

  31. 1:26

    thousand downloads at the time. Now it's

  32. 1:27

    closer to 2 million.

  33. 1:29

    In terms of like statistics, so we just

  34. 1:32

    checked that, you know, this is not

  35. 1:33

    theoretical. This is actual

  36. 1:37

    uh something that is succeeding.

  37. 1:39

    We have 3 million NPM downloads.

  38. 1:41

    4.5 uh 4.5 45% month-on-month growth.

  39. 1:46

    2.8 thousand websites using it in

  40. 1:49

    production from Minify to Z to

  41. 1:52

    Inphysical.

  42. 1:54

    And

  43. 1:55

    the whole concept of this talk was

  44. 1:58

    uh it goes back to we were doing all

  45. 2:00

    these things to make our library more

  46. 2:03

    efficient. You know, we were batting

  47. 2:05

    upwards

  48. 2:07

    compared to every other tool. Every

  49. 2:09

    other tool was built for marketers and

  50. 2:11

    lawyers. We were built for the

  51. 2:12

    developer. So, we had to make sure we

  52. 2:15

    had a very good developer experience.

  53. 2:17

    And we had an onboarding format said,

  54. 2:20

    "How did you hear about this?" And we

  55. 2:22

    started to get spikes that from April

  56. 2:25

    13th, you know, now it is our number one

  57. 2:30

    source of inbound is

  58. 2:33

    Claude, ChatGPT, Codex, that is ChatGPT,

  59. 2:37

    Gemini recommending us.

  60. 2:39

    And I like to think of this as, you

  61. 2:42

    know, the iceberg. You know, we start

  62. 2:44

    with the top of C15T

  63. 2:46

    and

  64. 2:48

    there's many many tools that go into it

  65. 2:51

    from, you know, LLMs.txt to site maps to

  66. 2:54

    RSS feeds to robot.txts.

  67. 2:56

    So many micro optimizations that you can

  68. 2:58

    do from old methods of running the

  69. 3:01

    internet to new methods.

  70. 3:03

    And uh

  71. 3:05

    how many of you have, you know,

  72. 3:09

    made these kind of tools? How many of

  73. 3:11

    you have, really put simply, said, "Hey,

  74. 3:16

    agents, we need to this to be done." And

  75. 3:21

    yeah, it said, "We should install this

  76. 3:23

    library." And you've gone, "Okay." Raise

  77. 3:26

    your hands. How many people have done

  78. 3:27

    this?

  79. 3:29

    Pretty much most people.

  80. 3:32

    That's a lot of hands.

  81. 3:34

    So, what's really funny is that we went

  82. 3:37

    from wizards installing our software to

  83. 3:39

    agents installing them. And I just went

  84. 3:43

    through Y Combinator.

  85. 3:45

    And what's really interesting is it if

  86. 3:48

    you know who these two people are. These

  87. 3:50

    are the co-founders of Stripe, the

  88. 3:52

    Collison brothers, and they had a really

  89. 3:54

    classic saying of like a Collison

  90. 3:57

    brothers install.

  91. 3:59

    And they would hand you their laptop,

  92. 4:01

    and they would install Stripe. These

  93. 4:03

    days, it's kind of like just a prompt.

  94. 4:06

    Being in Y Combinator, we just give

  95. 4:09

    people a prompt. And really what that

  96. 4:11

    means is that our very good developer

  97. 4:14

    experience primitives are now hitting

  98. 4:17

    agent primitives.

  99. 4:20

    So,

  100. 4:21

    as we was pulling all these things

  101. 4:23

    together, there is no one tool that

  102. 4:26

    fixes everything. I like to think about

  103. 4:29

    these problems like, you know, Batman's

  104. 4:31

    utility belt. Loads of really small

  105. 4:33

    things targeted in different areas to

  106. 4:36

    get it done.

  107. 4:38

    And we built all of these things into

  108. 4:41

    C15T because we wanted C15T to be the

  109. 4:44

    best developer framework in this tool.

  110. 4:47

    Think of it like Stripe Docs. And as we

  111. 4:50

    was building more and more tools, more

  112. 4:52

    and more documentation websites, we

  113. 4:54

    actually started abstracting these tools

  114. 4:57

    into a side quest

  115. 4:59

    that we call Lead Type. So, all of the

  116. 5:03

    things that we're going to talk about

  117. 5:04

    now are things that we have already

  118. 5:07

    solved with this open source framework.

  119. 5:10

    We have our friends at other developer

  120. 5:12

    companies implementing it and seeing

  121. 5:14

    similar results about how to like

  122. 5:16

    optimize for the agent experience.

  123. 5:19

    So, again, this isn't a magic SEO tool.

  124. 5:23

    It's actually

  125. 5:25

    a very non-sexy title, but a

  126. 5:27

    framework-neutral docs pipeline.

  127. 5:30

    Complex.

  128. 5:32

    But really, all it basically does is

  129. 5:33

    take your .mdx files, you run Lead Type

  130. 5:37

    generate, and it will spit out

  131. 5:38

    everything for

  132. 5:41

    um optimized agent experience for your

  133. 5:43

    websites.

  134. 5:45

    And the rest of this talk is going to

  135. 5:47

    look a bit like a BuzzFeed list, to put

  136. 5:50

    simply,

  137. 5:51

    of these problems, because again, not

  138. 5:53

    everybody knows even how to put an

  139. 5:56

    LLM.txt on their website. So, you know,

  140. 5:59

    that comes to the first problem of if

  141. 6:02

    your docs have hundreds of pages,

  142. 6:05

    and how can it navigate them to find the

  143. 6:07

    right questions?

  144. 6:10

    The first solution is obviously an

  145. 6:12

    LLM.txt.

  146. 6:14

    What we found in our research is that

  147. 6:17

    it's much better not to just generate

  148. 6:19

    this. It is much better to write your

  149. 6:22

    LLM.txt from hand. Obviously,

  150. 6:25

    AutoRaptor, but write it as you are

  151. 6:28

    trying to get the answers across to the

  152. 6:31

    LLMs.

  153. 6:33

    For about 40 good lines beats 1,000

  154. 6:36

    lines of noise from our testing.

  155. 6:39

    And that comes to the second issue of

  156. 6:42

    agents don't know how to browse. They

  157. 6:44

    know how to fetch.

  158. 6:46

    So, you then need the second part of the

  159. 6:49

    solution of the LLMs full.

  160. 6:51

    Again, think of this as a sitemap, where

  161. 6:54

    it takes the actual page and the links

  162. 6:58

    and a short description of what each

  163. 7:01

    page is for the LLMs to reference.

  164. 7:03

    Again, most people have heard these two

  165. 7:05

    solutions.

  166. 7:07

    But, where things are starting to get

  167. 7:09

    very complicated and we're seeing a lot

  168. 7:11

    of optimizations right now, is that HTML

  169. 7:14

    is expensive, and why can't we just ship

  170. 7:18

    markdown to the agents?

  171. 7:20

    And we can. And you've seen that

  172. 7:24

    everybody has started creating twin MDs.

  173. 7:28

    So, that's taking the normal website,

  174. 7:30

    such as Next.js quick start, and then

  175. 7:33

    having a dot MD on the end of it.

  176. 7:37

    And when you load that,

  177. 7:40

    it goes to the markdown

  178. 7:43

    But, what's really important here, and

  179. 7:45

    it's really worth noting, is this line

  180. 7:47

    at the bottom. If you look at all the

  181. 7:49

    best documentation websites, Minify,

  182. 7:53

    Vercel, C15T, pat myself on the back. Um

  183. 7:56

    they all have this in the header. This

  184. 7:59

    is saying to the agents whenever they

  185. 8:01

    visit the website that there is an

  186. 8:04

    alternative

  187. 8:05

    version of this in mark markdown. Again,

  188. 8:09

    who actually supports it? Don't ask me.

  189. 8:12

    Perplexity, some of the agents, it's all

  190. 8:15

    up in the air.

  191. 8:17

    And then, the second thing as well

  192. 8:20

    is that taking the .mds,

  193. 8:24

    you need to make sure that they're

  194. 8:25

    available through multiple methods. So,

  195. 8:28

    one of them is like the .md, so as you

  196. 8:31

    like copy it to an agent, you say .md.

  197. 8:34

    Another one is just taking the normal

  198. 8:38

    um link and then adding a uh redirect

  199. 8:42

    into your like your Next.js config, so

  200. 8:44

    that if it detects an agent has the

  201. 8:47

    header of accepting markdown, instead of

  202. 8:50

    returning the HTML, it will return the

  203. 8:53

    markdown. And then, the third one is

  204. 8:56

    that not all agents can append header

  205. 8:59

    tags. So, there's also a URL query of

  206. 9:03

    mode equals agent.

  207. 9:06

    So, they're the ones that pretty much

  208. 9:08

    everybody knows. Um and it's pretty

  209. 9:12

    basic internet knowledge at this point.

  210. 9:14

    Um

  211. 9:15

    but one of the really interesting ones

  212. 9:18

    is where we're going next. And our

  213. 9:20

    tooling is also helping this is that an

  214. 9:23

    agent can't ask your website anything.

  215. 9:28

    So, we need to think about the web MCP.

  216. 9:31

    And this is still very early, but our

  217. 9:34

    tool is already uh

  218. 9:36

    um exposing three different tools to

  219. 9:40

    WebMCP. Search docs, get pages, and ask

  220. 9:44

    docs. Again, um our library lead type is

  221. 9:48

    pulling all of that context together so

  222. 9:51

    an agent can easily ask it the right

  223. 9:52

    questions. I think we'll even see a

  224. 9:55

    future where communication happens over

  225. 9:58

    email and there's companies in San

  226. 10:00

    Francisco building that today.

  227. 10:04

    But this is actually the most

  228. 10:06

    interesting one and I think the most

  229. 10:09

    important one that anybody who has any

  230. 10:12

    type of developer module surface, NPM

  231. 10:15

    modules, cargo, Python, whatever.

  232. 10:19

    Is that the uncomfortable truth is that

  233. 10:22

    coding agents are actually never

  234. 10:24

    visiting the website if you have a

  235. 10:26

    library.

  236. 10:27

    They're actually visiting the node

  237. 10:30

    modules. They read the repo and they

  238. 10:33

    read the node modules. They They have

  239. 10:36

    previous stale training data

  240. 10:39

    and they're trying to work it out on

  241. 10:41

    what it can do from the the compiled

  242. 10:44

    source.

  243. 10:46

    So, again, following what people like

  244. 10:49

    Vercel are doing and people who are

  245. 10:51

    thought leaders in this industry is that

  246. 10:54

    we take the bundled markdown documents

  247. 10:59

    and then we also put them in the node

  248. 11:02

    modules

  249. 11:03

    with an agents.md

  250. 11:06

    file.

  251. 11:07

    And the agents.md file basically says,

  252. 11:09

    "If you've got a problem, if you've got

  253. 11:11

    a question,

  254. 11:12

    all the documents are here. Grab them."

  255. 11:16

    And we actually see that this has

  256. 11:19

    surprisingly real effects. We can see

  257. 11:22

    that between many different models

  258. 11:24

    almost 50% token saving on instead of

  259. 11:28

    trying to search the web, find the right

  260. 11:31

    tools, pulling the markdown files from

  261. 11:34

    your code base. So, if you have a

  262. 11:36

    library that's forever changing,

  263. 11:40

    then having the node modules built in is

  264. 11:43

    a very effective solution. This is also

  265. 11:45

    working without any skills, but if you

  266. 11:48

    want as well, you can add skills to it

  267. 11:50

    to say, "Look at the node modules and go

  268. 11:52

    from that."

  269. 11:54

    And again, just uh doubling down into

  270. 11:57

    this point, looking at like the

  271. 11:59

    agents.ai and DFile, you can say like

  272. 12:01

    when working with Z15 T Next.js library,

  273. 12:04

    read the bundles and

  274. 12:07

    verify that they match and go from

  275. 12:09

    there.

  276. 12:11

    So,

  277. 12:12

    that's really like how we've done it. I

  278. 12:16

    don't want to say this is like

  279. 12:17

    prescriptive, that I know the answers.

  280. 12:21

    If you have documentation websites, or

  281. 12:23

    if you have any type of markdown, if

  282. 12:25

    you're running your own blog, you know,

  283. 12:27

    I've been using our package as well on

  284. 12:29

    our marketing website. Every part of our

  285. 12:31

    marketing website also has a markdown

  286. 12:34

    file. It can be something that's used

  287. 12:36

    for many things. We're currently just um

  288. 12:40

    most people are just using it for

  289. 12:41

    documentation.

  290. 12:43

    But, you can literally run it and it

  291. 12:44

    will pull out all of these extra files.

  292. 12:48

    And

  293. 12:49

    one of the big things was when I put

  294. 12:51

    this talk together,

  295. 12:52

    you know, we were seeing the results

  296. 12:54

    that Claude was recommending, but there

  297. 12:57

    was not really any like test suites yet,

  298. 13:00

    or test harnesses, on like is your site

  299. 13:03

    agent ready? And Cloudflare brought one

  300. 13:05

    of them out.

  301. 13:07

    But, my favorite

  302. 13:10

    is actually one called Aura AI. Um this

  303. 13:13

    is brand new and it tests a lot.

  304. 13:17

    I'm happy to show off score of 59

  305. 13:20

    because it's constantly changing. 3

  306. 13:22

    weeks ago, it was a lot higher. And

  307. 13:24

    again, this is a forever changing area.

  308. 13:27

    So, aura.ai, put in your website, and it

  309. 13:30

    will start giving you recommendations.

  310. 13:33

    It's forever changing. Again, we can

  311. 13:35

    just stay on top of it. And yeah, this

  312. 13:39

    is like one of my final slides is that

  313. 13:41

    the slide the

  314. 13:42

    the market agents, LLMs, everything is

  315. 13:46

    forever changing. There is no such thing

  316. 13:48

    as perfection. When I started making

  317. 13:51

    these slides, I got so caught up of

  318. 13:53

    like, everyone expects me to be the

  319. 13:55

    expert here, but I've just been hacking

  320. 13:57

    on this problem a little more than you

  321. 13:58

    guys have so far. So, never get caught

  322. 14:01

    with being perfect. Every small little

  323. 14:04

    increase really does matter. Every small

  324. 14:07

    little thing you add really does matter.

  325. 14:11

    Thank you so much. You can find me on X,

  326. 14:14

    Burn Chris, and LinkedIn, and

  327. 14:15

    everywhere.

  328. 14:16

    >> [applause]

  329. 14:17

    >> WOO!

  330. 14:19

    I THINK

  331. 14:21

    I THINK WE HAVE TIME for one or two

  332. 14:23

    questions. Yeah, of course.

  333. 14:32

    >> So, if you were building um

  334. 14:35

    uh we're a website agency. We work with

  335. 14:37

    a lot of startups building like their

  336. 14:38

    own websites.

  337. 14:39

    >> Mhm.

  338. 14:40

    >> If you were just building a website, not

  339. 14:41

    necessarily like developer tool, but

  340. 14:43

    just a website to be found, which of

  341. 14:45

    these methods like would you concentrate

  342. 14:47

    on if you're starting from scratch?

  343. 14:49

    >> Yeah, I think the most important ones,

  344. 14:53

    and we're starting to see this more and

  345. 14:55

    more, is

  346. 14:57

    trying to provide a dot md file for

  347. 14:59

    every single page. A lot of CMSs are not

  348. 15:02

    built in this way. Um and we see this

  349. 15:05

    optimization happening more and more

  350. 15:07

    where I didn't put in the slide, but

  351. 15:10

    we're seeing more and more websites

  352. 15:13

    being visited by agents instead of real

  353. 15:15

    humans. So, in terms of even like trying

  354. 15:18

    to be proactive and token efficient, you

  355. 15:21

    should provide a markdown file if you

  356. 15:24

    can. Again, a lot of CMS's are not built

  357. 15:27

    this way. I actually built my own CMS.

  358. 15:31

    My name is Chris and I built ChrisCMS,

  359. 15:33

    short for Christmas. It's a whole It's a

  360. 15:35

    whole thing my team wishes I never

  361. 15:37

    built.

  362. 15:38

    But, it does work and it does bring this

  363. 15:41

    like token efficiency up. So, yeah, I

  364. 15:44

    would say llms.txt

  365. 15:46

    is your first shout.

  366. 15:49

    llms.txt full form

  367. 15:51

    .txt. Second, if you if you can, just do

  368. 15:55

    them manually. Say you're not even

  369. 15:57

    working on systems that have markdown, I

  370. 16:01

    still recommend them.

  371. 16:03

    But, you can always get creative with

  372. 16:05

    creating these

  373. 16:08

    files on the on on the go.