Augmenting Technical Writing with AI: 10 Key Principles

I was telling my Technical Communication students about Tom Johnson this week in regards to his online primer for learning API documentation. In this post, he's sharing ways that technical writers can use GenAI, and need to use GenAI, as part of their jobs and how tech comm as a profession is changing, and honestly the profession has always been changing with changing technologies.

Tom hits the nail on the head here. So many people have said, and it is true, that AI can't ever totally replace technical writers. This article offers practical advice for those of us working with AI to augment our work and spend our time more efficiently.

I'm curious about memory retention and reading comprehension rates for text written by AI versus text written by humans. What's the point of writing something so smooth that it's immediately forgettable?

⚡ RAG vs Fine-Tuning — Which One Should You Use for LLM Applications? One of the most common questions in Generative AI today is: 👉 Should we use Retrieval-Augmented Generation (RAG) or Fine-Tuning? Both approaches improve Large Language Models, but they solve different problems. #GenerativeAI #LLM #RAG #FineTuning #ArtificialIntelligence #MachineLearning #AIEngineering #AIArchitecture #DataScience #VectorDatabase #AIInnovation #TechTrends

Read what's new on the Next Gen Learning blog » Beyond the Pencil: Using AI to Bridge the Gap for 2nd Grade Writers https://lnkd.in/etFiMv-N

I've been using frontier AI models to explore the sixth moment of the Riemann Zeta Function — one of the toughest open problems in analytic number theory. I wrote a 25-post thread on X to share my workflow: finding and killing false resolution paths, compressing the frontier barrier object, and being honest about what AI can and can't do for hard math. The tl;dr, while powerful AI models like GPT Pro and Codex won't prove theorems for you, they can be remarkably effective in reducing the problem space until theorems have few good places to hide. Post linked in the first comment!

We’re systematically reviewing 250+ generative AI research papers. That means multiple coders independently analyzing the same studies and then checking whether they agree. Inter-rater reliability is one of the most time-consuming parts of rigorous evidence synthesis. So we asked: could AI help? We ran two rounds of experiments with Claude, GPT, Gemini, and NotebookLM. Not to code papers themselves, but to audit how humans coded them. What worked: AI quickly identified where coders agreed on clear-cut fields. It even caught an error a human reviewer missed: one coder had accidentally flipped min and max values across multiple papers. What failed: When we asked each system to calculate overall agreement, we got results ranging from 62% to 90% on identical data. One model later admitted its Krippendorff's α values were "estimates," not calculations. Bottom line: AI is useful for first-stage triage. It can surface where disagreements cluster. But it cannot explain why they exist or determine which interpretation is correct. The hardest part of the work (and the part that determines validity) still requires human judgment. Full methodology, prompts, and lessons learned in our latest blog post: https://lnkd.in/ekXA-y_Z Subscribe to The Hubdate for monthly updates: https://lnkd.in/eYcU-WpV #AIinEducation #ResearchMethods #OpenScience #EdTech

I’m really grateful that alongside my PM responsibilities on this project, I’ve had the chance to actually work hands-on with LLMs and see how they perform in real workflows. There’s a lot of noise right now about AI replacing jobs. I don’t have a crystal ball, but we also thought we’d all be driving flying cars by now. What I’m seeing instead is much more practical. The value right now is in learning how to use these tools well and understanding where they actually hold up in practice, not just trying to predict where they’ll end up. This project is giving me the chance to test them in a structured way. I did my best to set up thoughtful experiments, but I’m not a researcher or an LLM expert, and I definitely felt that at times. When I mentioned as much to John Whitmer a while back, he said something along the lines of, “welcome to research,” which felt about right. One of the biggest takeaways is that AI can help with parts of the process, but it doesn’t remove the need for judgment; just ask ChatGPT how many R's are in the word strawberry. In some cases, it just shifts where the effort goes. I’d love feedback on the prompts (linked in the post, many of which were shaped with the help of the same models I was testing), and I’m curious how others are using AI in their work. Has it actually improved your process, or just changed it? Full write-up here: https://lnkd.in/e_FzAjVx

Ever thought how an AI model makes sense of human language? We often talk about "embeddings" and "vector space," let's learn the concept.. One of the simplest document I developed explains about "embeddings" and "vector space," with a classroom analogy that gives you simple ye effective understanding. I’ve put together a presentation deck that deep-dives into this "Learn & Fix" loop. If you’re interested in how to make complex AI concepts accessible for students or stakeholders, let’s connect! #ArtificialIntelligence #MachineLearning #DataEngineering #AIExplained #TechEducation