Notes on writing faster, on your Mac
How Tab-to-accept autocomplete works, why it stays on your Mac, and what it's like to build a privacy-first writing tool.
What macOS 26 Changed for People Who Type
macOS 26 brought Apple's on-device writing engine to the Mac, and it quietly changed the rules for everyone who types for a living. Here is what shifted and where it leaves the field.
Read the post →
Apple's Engine or a Local Tier: How to Actually Choose
WriteAmp can run on its own local tiers or Apple's on-device model. Here is the honest decision guide for which one you want, without the marketing.
Read the post →
WriteAmp Can Run on Apple's Model. Here's the Trade.
WriteAmp can route its suggestions through Apple's on-device model on macOS 26 instead of its own local one. Here is the honest trade table for that choice.
Read the post →
Lifetime Updates When You're the Only Employee
Lifetime means something different when the company is one person. Here is the honest meaning of lifetime updates, what keeps them going, and what happens if I stop.
Read the post →
I Read My Own Prediction Logs So the App Stays Honest
Every suggestion WriteAmp shows or hides is logged locally, and I read those logs to keep the app honest. Here is what the logs actually say and why they matter.
Read the post →
The Hardest Part Was Teaching It to Stay Quiet
Making the model suggest the right word was the easy part. Making it know when to say nothing took most of the work. A maker's war story about the feature you never see.
Read the post →
Founders Type the Most and Notice It the Least
Founders write more than almost anyone, and think about writing tools less than almost anyone. Here is what that gap costs, and what happened when I finally noticed it.
Read the post →
Drafting in a Second Language at Work
Half the world does their job in a language that isn't their first. Here is what writing support in your working language actually takes, and why it is harder than it looks.
Read the post →
Apple Intelligence for Writing: What It Does and Where It Stops
Apple's on-device writing engine is real and getting better, but it is patchy about where it shows up. Here is what it does, where it stops, and how a system-wide tool fits alongside it.
Read the post →
Why the Trial Is 30 Days and the Price Is $29
The trial length and the price are not marketing guesses. Here is the maker's view of why WriteAmp is 30 days free, then $29 once, and what each number is doing.
Read the post →
Building the Honest Alternative: Where WriteAmp Came From
WriteAmp's core descends from an open-source project called KeyType, and I have never hidden that. Here is the honest story of lineage, license, and what got built on top.
Read the post →
Support Teams Type the Same 40 Sentences a Day
Support work is mostly sentences. The same polite openings, the same explanations, the same closes, forty times a day. Here is what autocomplete does for support.
Read the post →
The Line Between Fast Enough and In the Way
Autocomplete is judged by feel, not specs. The whole game is landing under the line where a suggestion still feels like it belongs, and knowing why that line exists.
Read the post →
Notarized, Updated, Boring: The Trust Stack You Never See
Before a Mac app reaches you, it is signed, notarized, and wired into an update channel. Here is what that stack actually is, and why the boring parts protect you.
Read the post →
Code Comments Are Writing, and Nobody Optimizes for Them
Developers write prose all day in the middle of code, and nobody treats it like writing. Comments, TODOs, and PR messages are where autocomplete quietly earns its keep for developers.
Read the post →
Notion AI Wrote My Paragraph. I Still Typed It Myself.
Notion AI can write you a paragraph, and then you rewrite it to sound like you. Here is what that round trip actually costs, and what autocomplete does instead.
Read the post →
Apple's Built-In Predictions: Good, Patchy, and Not a Habit
Apple has shipped text prediction for years, and it is genuinely decent where it shows up. Here is what it does, where it stops, and how it differs from a real autocomplete habit.
Read the post →
Stop Retyping the Sentence You Wrote Yesterday
The sentences you retype are not really the same, which is why snippets fail and autocomplete works. Near-repeats are the quiet majority of your typing.
Read the post →
Fast Notes Beat Perfect Notes
The best notes app is the one that keeps up with your thinking. Autocomplete in your notes is not about speed, it is about not losing the thought.
Read the post →
Async Teams Run on Written Updates
Remote and async teams do their talking in text, and the people who write the updates carry the team. Here is what autocomplete does for that work.
Read the post →
TextExpander Does One Thing Better Than Me. Use Both.
TextExpander is for boilerplate you want byte-identical every time. Autocomplete is for near-repeats that vary. They are not competitors, they are two halves of the same problem.
Read the post →
WriteAmp vs Grammarly: The Honest Version
Grammarly is genuinely good at grammar. WriteAmp finishes your sentences on your Mac. They are not rivals, and here is the honest case for each.
Read the post →
Day One You Notice It. Day Three You Don't.
The adoption curve of autocomplete is a reflex, not a workflow. What the first three days feel like, and why the best sign is that you stop noticing it.
Read the post →
How a Menu-Bar App Knows What You're About to Type
A tiny app in your menu bar somehow predicts your next words in any app. Here is the whole pipeline, from the key you press to the gray text that appears.
Read the post →
The NDAs in Your Inbox: Why Client Work Changes the Privacy Math
Sign one NDA and the tool you type into stops being a personal choice. Here is how client confidentiality reshapes the writing-software decision for consultants.
Read the post →
The Email-Forward Life
Some jobs are a series of emails. If you live your working day in the inbox, here is what autocomplete actually does for you, and where it stops helping.
Read the post →
The Small File That Keeps Autocomplete From Saying Nonsense
When WriteAmp downloads a model, it also builds a second file: a map of the model's vocabulary. If the two disagree, the app refuses to guess. Here is why that file matters.
Read the post →
Typing on a Plane: WriteAmp With No Network
The real test of local software is a flight with no Wi-Fi. WriteAmp works offline, and here is exactly where its network dependence ends.
Read the post →
Autocomplete vs Autowrite: Two Tools, Two Jobs
One finishes the sentence you are typing. The other writes it for you. Autocomplete and autowrite are different jobs, and mixing them up causes most of the disappointment.
Read the post →
What 60 Words a Day Cost You This Year
You retype the same sentence openings all day and never notice. Sixty words a day adds up to real hours a year. Let me show you the math.
Read the post →
The Open-Source Autocompletes Are Good. Here Is What a Paid App Adds.
The free, open-source autocompletes are genuinely good, and one of them is where my app came from. Here is the honest case for what $29 actually adds.
Read the post →
What You Are Actually Downloading When You Pick an Engine
When you pick a local engine you download a model file. Here is what those files actually are, why the sizes differ, and what lands on your disk.
Read the post →
The Battery Feature Nobody Asks For
You unplug your Mac and it gets quieter about the heavy lifting. The engine downshift is the feature nobody asks for and the one I am proudest of.
Read the post →
The Apps WriteAmp Refuses to Read
There is one kind of field where WriteAmp will never, ever offer a suggestion. Here is why your password manager is the safest place on your Mac.
Read the post →
The Boring Privacy Win: A Local Model Has No Reason to Phone Home
The strongest privacy argument is structural, not moral. A model that runs on your Mac has no reason to want your sentences, and that changes everything.
Read the post →
What WriteAmp Turns On by Default (and How to Turn It Off)
Fresh installs get history, clipboard awareness, and screen context switched on. Here is exactly what those defaults do and where every switch lives.
Read the post →
Your Writing History Is an Encrypted File. The Key Is in Your Keychain.
When WriteAmp learns from you, it writes one encrypted file to your disk, and the key that unlocks it sits in your Keychain. Here is exactly how that works.
Read the post →
Open This One Settings Screen Before You Trust Me
Before you trust a word I say about privacy, open this one settings screen. It takes two minutes and shows you exactly what the app learned from you.
Read the post →
Why Code Editors Got Autocomplete a Decade Before Email Did
Developers got Tab-to-accept autocomplete years ago. Everyone else is still retyping the same sentences. The reason is not technical, it is economic.
Read the post →
What a Subscription Autocomplete Costs Over Five Years
Eight dollars a month does not sound like much until you add up five years of it. Let me do that math with you.
Read the post →
I Published My Benchmarks, and the Numbers Are Not All Flattering
I measured my own autocomplete and put every number online, good and bad. Here is what v1 found, including the parts I would rather not dwell on.
Read the post →
Eight Languages I Don't Speak, and the Gate That Keeps Me Honest
WriteAmp supports eight languages beyond English. I speak exactly none of them, which is why the whole feature is built around a quality gate instead of my instincts.
Read the post →
The Screen Recording Permission, Explained Like You Are a Person
An autocomplete app asking for Screen Recording sounds terrifying. Here is what it actually does with that permission, and how to keep it on a short leash.
Read the post →
Knowing When to Shut Up Is the Feature
The most important thing my autocomplete does is nothing at all. Here is why showing nothing is the whole product.
Read the post →
Is WriteAmp Worth It? An Honest Answer
The direct question deserves a direct answer. Here is who WriteAmp pays for itself, who should skip it, and the limits I will not hide.
Read the post →
The 2026 Mac Autocomplete Map, Drawn Honestly
A category that barely existed two years ago now has a dozen apps all doing ghost text and Tab. Here is who is who, priced honestly.
Read the post →
Follow the Draft: The Privacy Math of Cloud vs Local Writing Tools
Take one email draft and trace where it goes with a cloud writing tool and where it goes with a local one. The two trips are not remotely the same.
Read the post →
The Three Times WriteAmp Uses the Network
The homepage says nothing leaves your Mac. It is a true sentence with three small asterisks. Here is exactly what they are.
Read the post →
Ghost Text: The Quietest Interface on Your Mac
It is the interface you stop seeing the moment it works. Here is what ghost text is, why it took a decade to leave code editors, and when it earns its keep.
Read the post →
Every Text Field Deserves a Tab Key
Code editors got Tab-to-accept a decade ago. Every other app is still typing on paper. Why the same habit should work everywhere, and what it costs.
Read the post →
Mini, Midi, or Max: Choosing a Local Model Like a Grown-Up
Three local tiers plus Apple Intelligence on macOS 26+. The reflex is to pick the biggest. That's usually wrong. Here's how to actually choose.
Read the post →
Personalization Is a Black Box Until You Force It Open
Your writing tool "personalizes" by hoarding what you type. WriteAmp shows you every phrase it learned and lets you wipe it. Here's how.
Read the post →
Stop Mailing Your Drafts to a Server: On-Device Writing Tools, Rated Honestly
Cloud writing assistants upload everything you type. I built the on-device version and ran the comparison. Here's who each is for, with the tradeoffs stated plainly.
Read the post →
The Tab Key Is the Most Underrated AI Interface on Your Mac
Startups built chat boxes to rewrite your work. The most useful AI interface is the Tab key you already own.
Read the post →
Subscribe with your favorite reader: RSS feed. Questions: [email protected].