👋
The thing about AI right now is that most people are using it the same way they used Google in 2005. They type a question. They get an answer. They move on.
That is fine. But it is leaving most of the value on the table.
This week, I am running a series called Get AI Native in 2026. Not a course and definitely not theory. Every issue, I will walk you through one or two things you can actually set up this week that will change how you work. Step by step.
This issue is about two tools that genuinely changed how fast I work. One is about how you input. The other is about what runs in the background so you do not have to.
Let us get into it.
CT
Typeless: Cut time writing, refining, searching and copy-pasting
What it does: lets you dictate across your computer, edit selected text by speaking, translate, summarize, search, and draft without moving everything into a separate AI chat.
I tested it by: running a full email thread through it. Instead of typing instructions into Claude, I selected the email I received, held the shortcut (Alt or Fn), and said:
Draft a response to this email based on the thread selected addressing issues raised, keep it direct and cut the last paragraph.
Done in about four seconds.
The verdict: If you are spending more than 30 minutes a day in Claude or ChatGPT, this is the highest ROI thing you can install right now. The real unlock is not transcription. It is being able to edit with your voice. You can select any text on your screen and speak the change you want. It removes the friction of figuring out how to phrase your prompt. You just say what you mean.
If you use it this week: install it, do the 10-minute setup, and then spend one day forcing yourself to use voice instead of typing every single time you open an AI tool. It feels awkward for about an hour. Then it does not.
Claude Skills: Teach Claude your way of working
What it does: a Skill is a folder containing instructions and, when useful, reference files, scripts, or other resources. Claude loads the relevant Skill when a task matches it. Think of it as packaging the way you do a repeated job.
I tested it by: packaging my company’s brand guidelines and document rules into a Skill.
Previously, I would spend at least 20 minutes every session providing repetitive context: our color palette, design style, and the specific tone we use for documentation. It was a tedious process of manual alignment every time I started a new project.
Now, I can provide the raw input, whether that is an email thread, a rough file, or a quick list of ideas, and say:
Process these thoughts into a document.
The Skill is not a model trained on the brand. It gives Claude the instructions, references, and reusable process it should apply. That distinction matters because the files remain readable and editable. When the brand changes, I update the Skill instead of hoping the model learned the right lesson.
The verdict: this is the closest thing to giving an assistant a dependable procedure instead of explaining the job again every time. The first setup takes longer than another prompt. The return appears when the same task comes back next week.
The people who will get the most out of this are anyone doing the same type of output repeatedly: reports, client briefs, content drafts, meeting summaries.
If you use it this week: pick the one thing you ask Claude to help with most often. Write down exactly how you want it done. That is your first skill. Ask Claude to help you create that skill and you will see it come to life.
This week
Install Typeless (the free tier is enough) and build one Claude Skill for a task you repeat at least twice a week. That is it. Do not try to automate everything at once. One skill, done properly, will show you exactly what is possible.
What changed after I published this
The underlying idea became more useful, not less.
Typeless now presents voice input as a system-wide layer for dictation, selected-text editing, translation, search, and quick AI actions. Claude Skills are now a documented folder format that can carry instructions, scripts, and reference material across Claude.ai, Claude Code, and the API.
The products changed. The useful habit stayed the same: remove friction at the input, then package repeated judgment so the next run begins with context.
The Plug’s Note
The pattern I keep seeing with people who are genuinely fast and adaptive with AI versus people who just use it sometimes is this: they set up systems. Not complicated systems. Just small things that mean they do not have to make the same decision twice.
Skills are that for Claude. Voice input is that for prompting. Neither of them requires you to learn anything technical. They just require you to spend 10 minutes one afternoon setting them up instead of starting from scratch every time.
That is what this series is about.
See you next week.
CT
This was originally published as Issue #49 of Sunday Scoop.