Posts

Showing posts from October, 2026

Stop Repeating Yourself to Claude Code: Write One Rules Page

Image
Hi there. Do you explain the same things to Claude Code at the start of every new chat? I'm not a developer, so I write my project rules once and let Claude read them every time. Here is how I use one rules file in the project folder. This is for anyone tired of repeating the same setup talk. What the rules file is If a file named CLAUDE.md sits at the top of your project folder, Claude Code reads it when it starts. It is a plain text file you can write in Notepad. Anything you put there you do not have to explain again in a new chat. What I put in mine My file has five parts. Who I am: a non-developer who needs plain-language explanations What the project is: one or two short paragraphs Do and don't: publish outside only after my approval, never type passwords in chat Folder layout: which file lives where What "done" means: how to check that the work is finished Tips for writing it Short and clear beats long. Too long and the important rules get buried. Ins...

Text in AI Images: Which Tool I Use, and How I Check the Spelling

Image
Hi there. When AI makes a picture, the text inside it is often garbled or misspelled. I'm not a developer, and I make every blog thumbnail with AI. This post is not a fair head-to-head test of tools. It is a record of how I split images with text from images without text, and how I check them. It is for anyone who wants accurate text in AI images. First, a disclaimer I did not test ChatGPT, Gemini and Claude on text accuracy under the same conditions. So I do not claim that one is more accurate than another. What follows is how I actually work. How I split the work Thumbnails and explainer images that contain text: made with ChatGPT through Codex. Scene images with no text: made with Google's image tools. Real photos: taken from a free stock site that allows commercial use, with the photographer credited under each photo. I keep text images in one place because that is where text mistakes pile up and need fixing. Photo: Matheus Bertelli on Pexels How I get the tex...

Subscription-Only AI Workflow: How I Split Work to Avoid Usage Limits

Image
Hi there. If you run AI on subscriptions only, usage limits are the biggest worry. I'm not a developer, and I run blog and video work on subscriptions without any paid API. Here are the rules I use to split work so I hit limits less often. This is for anyone who wants to automate with a subscription-only AI setup. First, a disclaimer My subscriptions are Claude Pro, ChatGPT Plus and Google AI Pro. Plan limits change often, so I do not list numbers here. Check each service's official page for current limits. This post covers only how I divide the work. Rule 1: Give AI only the work that needs AI Posting to the blog at a set time is not done by AI. Windows Task Scheduler starts a small program at the scheduled time, and that program does not call any AI. AI writes the text and makes the images. Publishing on time runs without AI. Photo: Markus Winkler on Pexels Rule 2: Match the model to the job Simple jobs get a light model, ordinary jobs a middle model, and only hard...

Blogger Post Not on Google? How to Request Indexing in Search Console

Image
Hi there. Did you publish a post on a new blog and find that Google does not show it yet? I worried about the same thing with my new blogs. I'm not a developer, but I follow a short routine in Google Search Console: inspect the post's address, then request indexing. Here is how to request indexing for a Blogger post in Search Console. What to know first Indexing means Google adds your page to its search results. Requesting it does not guarantee a ranking or a date when the post will appear. I have no data that lets me promise a timeline. But telling Google about your post's address is the one reliable first step you can take yourself. Before you start: your site must be in Search Console I registered my whole domain as a Domain property, verified ownership, and submitted the sitemap. If you have not, add the property in Search Console first and finish the ownership check. Photo: Markus Winkler on Pexels The steps Sign in to Google Search Console with the account ...

When Your AI Agent Says "Done": 5 Checks I Run Before I Believe It

Image
Hi there. When you hand work to an AI agent, the answer that comes back is usually "done". How do you know it is true? I'm not a developer, but I don't take a completion report at face value. I run five checks first. Here is how I verify an AI agent's completion report. Why checking matters An agent tends to say "success" when a command ends without an error. But a command that finished is not the same as a result that changed. In my own work, an agent set a YouTube Shorts thumbnail through the API and reported it done, yet the Studio screen still showed the plain video frame. I wrote that down in my failure log, and now I upload Shorts thumbnails myself. Photo: Sanket Mishra on Pexels The five checks Where did you look? The answer should be "I read the result back", not "the command succeeded". After posting to the blog, I reopen the public page and compare the title and text. Does the report contain names and numbers? Po...

Launch Your AI-Built App With One Desktop Icon on Windows

Image
Hi there. Did you build a small app with an AI coding tool and now have to type a command in a black window every time you start it? I'm not a developer, and that was the part I liked least. So I set it up so that double-clicking one desktop icon starts the app and opens the browser. Here is how to launch an AI-built app with one desktop icon on Windows. All you need is one small file The key is a small file called start.vbs . It does three things: It checks whether the app is already running. If not, it starts the app quietly, with no black window. When the app is ready, it opens http://127.0.0.1:4747/ in your browser. My app is a work dashboard that opens only on my own computer. That is why the address starts with 127.0.0.1, which means "this computer". How I made it I did not write the file myself. I asked Claude Code for it. My request was one sentence: "Make it so a double-click starts the app and opens the browser." Then I opened the finished fil...

AI Agent Silent Failures: 3 Cases and the Fix

Image
Hi there. The worst failures I have seen with an AI agent were silent failures, not the loud ones. I'm not a developer, and three times my assistant reported a result that did not match reality. Once a success looked like a failure, once nothing changed but it said done, and once an error looked like an empty result. Here are all three, and the one habit that now catches them. Failure 1: a success that looked like a failure My assistant ran a quick check on Windows and the command printed an ugly assertion error and exited with a failure code. Underneath, the check had actually worked. The program only crashed while closing, because it was told to exit while a network call was still shutting down. The script read the failure code and decided the check had failed. So it ran the check again, six times in a row. Fix: let the work finish and then end normally, or set the exit code instead of forcing an exit. Photo: Markus Winkler on Pexels Failure 2: nothing changed, but it ...

Two AI Agents, One Blog: A Shared Publishing Ledger

Image
Hi there. What happens when two AI setups can publish to the same blog? I'm not a developer, and I run two separate setups, one built on Claude and one built on a different AI assistant. Both can reach the same blogs and video channels. So I set up one shared ledger that both must check before anything goes live. The problem I was avoiding Drafting is safe to do in parallel. Publishing is not, because two setups could post the same topic, or change the same post, at the same moment. I wanted a rule both setups follow without needing to talk to each other. Photo: Markus Winkler on Pexels The ledger and the lock The ledger is a plain text file that lists every external action: what it is, where it goes, who made it and who published it. Next to it is a folder for locks. A lock is a small file that says "this destination is being changed right now". Here is the rule that both setups follow: Drafting, research, image making and local previews can happen at any time...

Free Stock Photos for a Blog: 6 Image Rules I Use

Image
Hi there. If you search for free images for a blog, the hard part is not finding photos. It is knowing which ones are safe to use and which ones fit your post. I'm not a developer, and I added photos to about twenty blog posts. These are the rules I ended up with and why. Where the photos come from I use one source only: Pexels. According to the Pexels license page (checked on 2026-10-07), the photos are free to use commercially, and attribution is not required. It also lists things that are not allowed, such as presenting identifiable people in a bad light, selling unmodified copies as posters, or implying that a person or brand endorses you. Please read the current terms yourself at pexels.com/license . Photo: Markus Winkler on Pexels My six rules Use photos from one free photo site only. Put the photographer's name and the site right under each photo. It is not required, but it is my rule. Use only photos that relate to the post. I first picked general photos, decid...

Codex Image Generation Not Working? A Fix That Worked

Image
Hi there. If you search for "codex image generation not working", this might be a different cause than the ones you have read about. I'm not a developer, and I run Codex from the command line to make blog images. One day six image requests in a row produced no images at all. Here is what was actually happening and the small fix. What I saw I asked for six images, one after another. Each run ended in under half a minute with the same message: no image was created. Nothing in that message explained why. Photo: Shantanu Kumar on Pexels How I found the cause I stopped looping and ran one request by hand. Then I read what Codex had actually said back. It had not tried to draw anything. Instead, it replied with a short note that recommended an environment and a model, and asked me to reply with a go-ahead word before it would start. It was asking for permission, and my script treated the lack of an image file as a plain failure. Why Codex asked The reply matched an ...

Claude Skills (SKILL.md): Make Your Workflow Portable

Image
Hi there. If you search for Claude skills, you will find a lot about the SKILL.md file format. I'm not a developer, and I use that format to keep my work procedures out of any single AI tool. Here is how my setup looks, what I check, and what still only works in Claude Code. Why I stopped writing procedures into chats Early on, I explained each job to the AI again in every new chat. It worked, but the explanation lived in one conversation and disappeared with it. I wanted the steps stored in files I own, so I could hand the same steps to a different tool later. Photo: Matheus Bertelli on Pexels What a skill file looks like A skill is one folder with one file named SKILL.md. I follow the open skills format ( agentskills.io/specification ). The top of the file holds a name and a description, and the body holds the steps. A few rules I keep: The name uses lowercase letters, numbers and hyphens, and matches the folder name. The description says when the skill should be used...

AI Agent Permissions: A Brief Is Not a Sandbox

Image
Hi there. AI agent permissions became much easier for me to think about once I separated instructions from access. I can tell an assistant to work on one article and leave everything else alone, but that sentence is not a locked door around my files. Here is how I now describe the boundary before a task starts, using a real publishing workflow rather than a promise that a separate setup is automatically isolated. A separate setup is organization, not a locked room My project instructions name an assigned task, an output folder, and files that must stay untouched. They also prohibit opening secret files or changing unrelated schedules. That gives the assistant a clear brief: what to do and where to stop. A different configuration folder can keep settings organized. It does not, by itself, prove that the process cannot reach other files on the computer. I treat those as separate questions instead of describing every organized setup as a sandbox. The distinction is practical. A rule ...

Updating 20 Live Blogger Posts at Once Without Changing a Single URL

Image
Hi there. At some point you may want to rewrite posts that are already live. Maybe the voice is wrong, maybe the photos don't fit, maybe you simply learned a better way to write. I'm not a developer, and I had to do this to 20 published pages in one go. Here is how it went, what I checked afterwards, and the two surprises I hit along the way. Why I had to edit 20 live pages Across my two blogs, I had 18 posts and 2 About pages at that point. I decided that three things needed to change across all of them: The voice. I wanted every page written plainly as "I", the way I actually work. A few references that didn't belong in an English blog. Photos. Each post needed two real photos that match its topic. Editing 20 pages by hand in the Blogger editor was possible. But I wanted every change checked the same way, and I didn't trust myself to do that 20 times. Photo: Sanket Mishra on Pexels One approval for the whole batch, not twenty Normally, nothing goe...