Dachary Carey on AI Agent Documentation Research

Technical writers — quick question: Are your docs actually readable by AI agents? In my latest Technical Writing Uncensored interview, I sit down with Dachary Carey to talk about what’s really happening when AI tools crawl documentation sites — and why many modern docs quietly fail. We cover: • Agentic AI and developer workflows • Mechanical truncation risks • HTML vs Markdown for AI consumption • What writers should change right now If you work in developer docs, this is a conversation you don’t want to miss. 🎥 Watch here: https://lnkd.in/ghGe6NME What changes (if any) are you making to prepare docs for AI?

I recently sat down with Diana Payton for a chat about all the things docs-and-agent-related that I’ve been digging into. (And more!) If interesting chats about what’s going on in the industry today are your jam, check it out. Bonus: Diana has chatted with a lot of other interesting folks about interesting things, so if you’ve been looking for a way to keep tabs on what’s going on, the Technical Writing Uncensored episode list is a goldmine 😁

I quite enjoyed the recent conversation between Fabrizio Ferri-Benedetti and the Technical Writing Uncensored podcast for its grounded, "post-hype" look at how LLMs are actually shifting the landscape for documentarians. The core message? AI isn't an autonomous intelligence; it’s a hammer. And a hammer only builds a house if a skilled carpenter is swinging it. Here are the key takeaways for those of us navigating the "depressing middle ground" of AI implementation: 1. AI as a Second Language Instead of aiming to be "AI native," we should treat these tools as a second language that we pick up to enhance our primary skills. The Efficiency Gain: AI excels at "soul-killing" chores, like changing metadata across a thousand files or converting SQL case formatting. The Cognitive Cost: There is a real risk in delegating too much; writing from scratch is how a writer gains an intimate, "manual" understanding of a product. High-Quality Inputs: AI produces derivative content; humans must still provide the high-quality "organic" thoughts at the start of a project to prevent future "information slop." 2. Transitioning from Writer to Editor-Manager The pipeline is shifting toward a model where engineers or AI agents generate initial drafts, and the technical writer steps in as the "prompt architect." Owning the Prompt: To ensure safety and quality, writers must own the instructions and prompts that teams use to generate documentation. The "Picky Editor" Advantage: While AI is surprisingly good at coding because compilers are deterministic, it struggles with the nuance, tension, and truth required for great prose. Managing Artificial Entities: Documentarians are becoming managers of LLMs, which requires a new set of "career growth" requirements. 3. The "Mental Gym" and the Analog Garden To keep our thinking sharp in an era of "addictive" AI assistance, we need spaces where the machines aren't allowed. Cognitive Fitness: Engaging in "mental gym" exercises—like writing a blog or coding from scratch without help—is essential to maintain our unique value. The Human Premium: As AI content becomes a commodity, there will be an increasing market for "certified organic thoughts"—content that carries the human emotion and context a machine cannot replicate. Demystifying the Tech: Learning how machine learning actually works (the "statistical machine" under the hood) is the best way to protect our mental health and avoid "AI delusions." The Bottom Line AI frees us from the drudgery, but only if we use that extra time for high-value curation and systems thinking. We aren't being replaced by AI; we are being challenged to prove why human-led documentation is the "high-signal" truth the machines depend on. Looking for a little more high-signal truth in your LinkedIn feed? Follow Fabrizio! https://lnkd.in/giwjsjUf

"Are we heading toward an AI utopia, an AI dystopia… or something in between? In this episode of Technical Writing Uncensored, I sit down with technical writer and blogger Fabrizio Ferri-Benedetti to talk about how AI and LLMs are changing technical writing — and what writers should actually be worried about. Fabri writes the excellent Passo Uno blog, where he explores technical writing, AI, and the future of documentation. Read his blog here: https://passo.uno We discuss whether AI will replace writers, how technical writers can stay relevant, and why writers might actually become more important in an AI-driven world. Fabri also explains why writers should think of AI as “a second language”, not something that replaces human expertise. We also discuss: Why AI is more like a tool (a hammer) than an autonomous intelligence Why good documentation is essential for AI systems Whether tech writers will become editors and prompt architects Why writers need a “mental gym” to keep their thinking sharp This conversation is a thoughtful look at the real impact of AI on technical writing, without the hype. #TechnicalWritingUncensored #TechnicalWriter #TechnicalWriting #Documentarian #Documentation #WordNerds #TechComm #ContentStrategy #WritingTips #CareerAdvice #UserExperience #DocsLikeCode #TechWritingUncensored #TechCommunity #aiforwriters #ai #llm #artificialintelligence " https://lnkd.in/ehjUvAXe

When I read something these days,I stop to ponder for a moment.. is it an AI perspective or a real human experience.. there are moments I wish to hear the real human tone in writing, enjoy the writer’s unique style … every writer to his or her own unique style…very different to the colder and machinist tone of an AI generated words.. sentences..para.. story - the machine generated monotonous tone story that we get to hear more often these days! #humansvsmachine

My standards in work and quality have only grown in this era and I am more than happy to express my standards. If you cannot pitch a CREATIVE idea to me without the use of generative means, shame. If you cannot TALK to your team and conflict resolve without consulting an AI, shame. If you cannot TEACH me without referring me to a program with errornous information that turns my copy into generic writing, shame. I'm here to create. I'm here to communicate. I'm here to learn the proper way. I'm a human being. If you want the human touch, a machine cannot do this for you.

It's hard to separate the "AI as gajillion-dollar market darling" conversation from the "AI as internet-ruining slop tsunami" conversation from the "AI as career-swallowing black hole" conversation from the "LLMs as neat and useful tools" conversation. I joined Matt Stauffer, CEO of Tighten, on his new podcast "Pragmatic AI" to talk about each of these, but mostly to explore how AI impacts writing and creative work. We covered a lot: how it's right there in the name that Large Language Models were always going to upend the careers of writers and programmers first; how the uniqueness of your creative process determines the uniqueness of your creative output; why expertise, taste, and curation matter more than ever in the next era of the internet; and why we need more media decentralization with fewer Big Tech chokepoints and gatekeepers. But mostly we talked about AI as a tool, how tools shape us, and the hopes and fears I have that follow from that. I'm a professionally curious person. I'm also a sicko: I *like* the grunt work it takes to understand something and compellingly communicate it. Google Search is functionally useless (a generational own goal), and although I still search daily for ten blue links on UDM14, LLMs are the tools I reach for now when I need to learn faster, follow my curiosity further, and understand more than I need in order to chisel away at the ideas. I would never let an LLM write sentences for me because writing is the residue of my thought process. Plus, I like making sentences. I prefer the way they taste when I make them. And that's the Jordan Keller promise—our recipe hasn't changed in 39 years: 1. Start from your unique domain of expertise, curiosity, tastes, and interests 2. Notice what you notice 3. Ask interesting questions Watch/Listen: https://lnkd.in/gBGNwk_R

The issue I see with leaving the AI giveaways in your writing is that those of us who are sensitive to words and sentence structure will very often scroll past as soon as we notice these tells. Why? Because in a world where our attention is divided in a million directions at once, each of us (consciously or unconsciously) creates our own standards for where we put our attention. I, for one, naturally gravitate toward the work of humans who are deep in the vulnerable and (to me) admirable endeavor of true creative self-expression. I do experience the process of working with AI on other tasks to be fun and creative. But when it comes to content, I want to be sure that the time I’m spending reading your work is matched by the time you spent developing it. Even if you start with an AI draft, I want to know that you have invested your own creativity and innovation in editing it into your own words.

Readers want to engage with people who see and understand them, not with bots. Be that author your reader is looking for to help them live better lives. I'm Erin, editor of transformational books and book coach for soul-centered healers, therapists, and coaches who want to write a transformational book with confidence and without overwhelm. I'm here to help you write your best book. 😊 ✍ 📚

If you're preparing for AI Engineer interviews, one topic that is rapidly becoming a favorite is Approximate Nearest Neighbor (ANN) Because in real-world LLM and RAG systems, retrieval speed and scalability matter more than textbook theory. Here’s why ANN is interview-critical 🔹Core of Modern RAG Systems Interviewers know that every production RAG pipeline depends on fast vector search. If you understand ANN, you automatically show: Practical GenAI knowledge System design maturity Production thinking If you want to transition into Data Science or AI, feel free to book a 1:1 session with me. I’ll help you with roadmap, projects, and interview preparation Link: https://lnkd.in/dez6Ji7E