AI Policy

Why this document exists

Mostly so you don’t have to guess. I always appreciate it when other people or companies publish an AI policy. (Outside’s AI Policy is by far the sharpest and most resolute one I’ve come across!)

This page will keep changing as the tools change, as my own thinking changes, and – inevitably – as I run into more blurry edge cases I hadn’t accounted for.

My general stance on AI

There is no ethical way to use AI right now. None of the negative effects of this technology get cancelled out by using it more thoughtfully.

I hold that position the same way I hold most of my other convictions while living an imperfect life under capitalism: I strive to be a ‘mindful consumer’ while acknowledging that nothing I do will pass a purity test. I aim high, compromise constantly, and try to be honest about it rather than pretend otherwise.

AI is no different. I use it in small, deliberate ways, while staying sceptical of the capitalist undercurrents that turn it – like so many technologies before it – into an extractive, exploitative, dehumanising force.

I publish a newsletter about (among other topics) technology. Refusing to touch anything AI has come near isn’t really a workable option for me. So when I do choose to engage with these tools, I try to keep one question in the back of my mind: is this sharpening and enhancing our thinking and doing, or simply replacing it?

This document is my attempt to draw that line honestly and in public. It’s blurry in many places, and I won’t always get it right. The point isn’t to convince you that my take is the correct one – it’s to show you where I’ve landed at the moment, so you can decide for yourself whether that’s a DD you want to support.

How I use AI

General research. For background on an author, a company, an app, an artist without a bio, etc., I sometimes turn to Jan, Thaura.ai or Claude, in addition to classic web searches via DDG.

Summarising. I consume a lot of media mostly through Readwise. I always read/listen/watch in full. Sometimes I use Readwise’s built-in AI feature to help me structure summaries, organise highlighted sections or find related essays I’ve read in the past.

Creative writing. All of my writing starts as a Google Doc, Apple Notes file or, for shorter texts, directly in VS Code. I occasionally dictate first drafts using a free, open-source tool called Hex, which employs limited AI to make the transcription more accurate. From there, I sometimes use Claude to help with light edits (e.g. refine a section, test an alternative wording, point out inconsistencies or repetitions). I also use Claude to find attributable quotes for the newsletter’s header that match the issue’s main theme. With English being my second language, I’ve long run my writing through grammar/proofreading tools such as DeepL Write, which now also lean on AI.

Copywriting. I’ve occasionally used Claude for smaller, sales-flavoured bits of website copy, like the advertiser page and summaries of reader surveys. Mostly because I find that kind of writing very tedious, if not uncomfortable.

Curation. All items in DD are chosen by me, pulled from a mix of RSS feeds, newsletters, social accounts and other sources. There is no automation involved. If I haven’t read or properly inspected an item, it won’t get featured.

Coding. I love writing markup and still do pretty much all HTML/CSS myself. But for backend work – the DD Lounge and various internal admin tools running behind the scenes – I make use of Claude Code’s VS Code extension to help build and troubleshoot.

How I don’t use AI

In short: AI doesn’t speak for me, and it doesn’t decide anything on my behalf.

It doesn’t write lengthy texts. It doesn’t touch my communication – no AI-drafted emails or social posts. It doesn’t curate – what gets featured in DD is chosen by me. It doesn’t generate images or videos for me. AI also doesn’t touch reader data for targeting or profiling. And I don’t run productivity apps, generative tools or agents that stitch different SaaS products together behind the scenes. The only AI service I pay for, on and off, is Claude.

AI in the items I feature and in sponsorships

The same approach that guides how I create DD also guides what goes in it. I don’t automatically rule out a tool or product because AI was involved in building or running it. (Sometimes it’s impossible to tell, anyway.)

The screening is based on a mix of fact-checking and personal taste: some basic background checks on the people or company behind it, a read on the quality signals that are often pretty obvious once you look closely, and a fair amount of good old ‘gut feel’.

If something looks like it was vibe-coded into existence purely to catch a hype and make a quick buck, it doesn’t clear the bar. (Some vibe-coded, free and fun experimental projects can, though!) If it’s thoughtfully built and an AI feature earns its place, I’ll allow it. What I’m looking for, in short, is restrained, thoughtful use of AI in service of human work, not as a simple replacement for it.

The same applies to sponsors. I don’t always get a full, detailed look at what’s being advertised – sometimes a product hasn’t officially launched yet, sometimes the AI integration is more opaque than I’d like. I try, as best I can, to understand what’s being sold and whether it holds up against the standards above.

Some categories are an automatic no – like crypto, financial products, most physical products that encourage consumption. I filter out clichéd sales pitches and shady marketing techniques.

Beyond that, it’s judgement, case by case. I don’t always get it right. To be honest, finding sponsors is becoming harder as it is. Though, that’s not a reason to lower the bar, just an admission that it’s a real constraint I’m working within.

Pitching your project to DD

If your pitch is fully AI-generated, automated and dressed up as a note from ‘a long-time reader’, I’ll spot it. These emails tend to follow the same tired pattern, and they go straight to the bin, usually unread. I’m genuinely happy to hear about your work, just write to me as yourself.

Questions?

Got a comment, a question, or spotted a use of AI I haven’t disclosed? Get in touch. (It has a habit of sneaking into apps.) Reply to any issue, or email me directly.

Last updated: Oct 2026