Sponsored by

As soon as I spot an AI tell in an email or LinkedIn post:

(no judgement if you've written one)

Today is about two skills: spotting AI writing in your own drafts, and editing it out cleanly when you do.

The em dash got all the attention, but the harder tells have survived all word-level edit.

Those are mostly what we're working on. But first, some catch up:

NEWS NEWS NEWS

Tools That Caught My Attention

1. AgentRail: AI coding agents are good at writing code, less good at everything around it. AgentRail handles the full ticket-to-merge loop with source-available infrastructure, TypeScript and Python SDKs, and live integrations for GitHub, Linear, CircleCI, and GitHub Actions.

2. Nimbus: An AI browser companion built around the Claude Code UX, designed to navigate sites, fill forms, and extract data on your behalf. Mac-only for now, with the first 500 users getting it free forever.

3. Resemble.ai: Voice cloning, real-time text-to-speech, and speech-to-speech across 149 languages, plus a deepfake detection suite that runs on calls in Meet, Teams, Zoom, and Webex. Powers branded voices for customer service, gaming, and entertainment, with API access and an open-source model (Chatterbox) for teams that want to self-host.

Quick reference

The PDF at the bottom has the full list. Three rough buckets:

  • Word-level: em dash, "actually," "quietly," "X, not Y" closers, "dive deep," rule of three in fragments.

  • Sentence-level: announcing the point, hedging stacks, tautologies that sound profound and say nothing.

  • Structural: the reveal cascade, setup-list-philosophy sandwich, fortune cookie close, and performative casual maintained as a costume

Training your eye

You can't read articles about AI tells and become good at spotting them. The skill comes from reps.

1. Read a lot

I read 20 to 25 newsletters a week.

Mix in distinctly human writers (Patti Smith, Paul Graham, Anne Lamott) so you stay calibrated to what real voice sounds like.

The gap between that and how you write now is what you're trying to close.

2. Notice without fixing

A few weeks of marking-but-not-editing trains your gut to catch the same patterns in your own drafts as you type them.

3. Read out loud, all the time

Once you've heard it enough times in other people's writing, you hear it in your own the moment it appears.

4. Squint at the silhouette

Human writing is irregular: some paragraphs one line, some half a page.

If your draft looks like a stack of identical bricks, the words alone won't fix it.

5. Edit other people's writing first

A hundred reps of this and your own drafts get much easier.

Editing them out

Once you can see them, the workflow:

The Bottom Line

That isn't about memorising more tells.

It's about volume: reading more, marking without fixing, reading out loud, putting in the editing reps.

The skill works like any other muscle.

If you do one thing from this email, pick something you wrote a month ago and read it out loud right now.

Mark every sentence that doesn't sound like you and don't fix anything yet.

That's the first rep.

One more thing

Full reference doc with every pattern I've spotted this year. Around 40 of them, by category.

Free, no signup.

Comprehensive Analysis of AI-Generated Writing Tells.pdf

Comprehensive Analysis of AI-Generated Writing Tells.pdf

314.20 KBPDF File

Until next time,
Vaibhav 🤝🏻

If you read till here, you might find this interesting

#AD 1

How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

#AD 2

Get two hours of your day back.

Lindy texts you a meeting brief before every call: who you're meeting, what you discussed last time, what's still open. 1 minute setup. Try it free.

Reply

Avatar

or to participate

Keep Reading