Interview with Blushdesk

Blushdesk is a generative AI-powered virtual assistant built specifically for sex workers by a sex worker to manage client communication. You set it up by defining your rates, boundaries, FAQs, and tone of voice, so it knows exactly how you want to present yourself and what is and is not allowed. Instead of acting like a generic chatbot, it is designed to sound like you and follow your rules.
Once configured, the AI handles incoming messages across WhatsApp, Telegram, and email. It replies instantly to common enquiries, filters out timewasters, and enforces your boundaries, for example deflecting inappropriate requests or answering routine questions like availability and rates. It never negotiates on your behalf or breaks the limits you have set.
You stay fully in control. You can review conversations at any time, step in whenever you want, and set triggers for when the AI should escalate a message to you, such as booking requests, VIP clients, or anything unusual. When it does hand over, you get the full conversation context, not a summary.
Technically, it runs on a self-hosted AI model with a strong focus on privacy and control. Messages are handled quickly, usually within a couple of seconds, and nothing requires the client to install anything. Clients simply message you as normal, while Blushdesk manages the conversation in the background.
Well the oracle certainly was wrong. BlushDesk can help you sort through the noise while you enjoy a beautiful book like Children of Strife. I personally use it and can tell you the mental breathing room I've gotten is more than worth it.
Using your example, you could read for hours and the only times the AI would call for you would be for something urgent or something beyond it, e.g. a first-time client who feels off, an unusual request, but for all the other messages where you pick up your phone and immediately realise it could have waited, e.g. "What are your rates?", "Will you be available on Friday next week?" The AI handles it easily across email, booking forms, WhatsApp and Telegram. It really is the secretary we all deserve with the tons of timewasters and noise that floods our DMs.
It works with both personal and business. You're right that Meta bans anything related to the sex work industry from using their API but what we did here is use the same technology you'd use to connect to WhatsApp Web on desktop, using a trusted open source library called Baileys. This bypasses the possibility of being banned by WhatsApp for being a sex worker. This doesn't mean it is without risk though. The biggest is that it's unofficial, so if WhatsApp changes something it could stop working for a few days (you will be notified). The other risk would have been your number being banned due to spamming, but that's a moot point in our case because the AI never reaches out first and doesn't send one-to-many messages. Clients reach out first, then the AI responds. We highlight these risks no matter how small in the app. It also goes without saying not to use your personal WhatsApp number for this. Only WhatsApp has this risk though, Telegram doesn't.
Sadly, sex workers would not be the end of big tech companies. The reason they can afford to ban us and hide us out of sight is because most people outside the industry are either vaguely supportive or apathetic to our plight. A more likely cause of their downfall would be data theft backlash, kids' mental health litigation or EU regulation with teeth, and when that day comes, for whatever reason, I will be breaking out my finest bottle of champagne.
With directories though, I would say their days are numbered. Their entire business model is us. If we leave, they die, and the reasons to leave keep piling up: the opaque ranking schemes where the price to be seen keeps going up, ratting providers out and we only find out after the fact, genuinely horrible support. The day most providers take back the one edge directories have over us, which is discoverability, is the day their relevance ends. With more and more providers getting their own websites, we're gradually inching closer to this.
On our part, I knew what I'd love to see from these platforms and never got it, so I built BlushDesk with that in mind. Things that would be unusual for most apps in this space, e.g. a warrant canary that lets you know if we get data requests from government agencies, and if we don't update it for more than a month you know we've been told to stay quiet (we update it monthly). There's also a page dedicated to explaining every single tech choice we made, from what AI system we use to what we use for emails to our encryption. Nothing hiding behind vague reassurances: Why you can trust Blushdesk.
Short answer: no. I wouldn't want it on the app store.
While BlushDesk could benefit massively from being an app, being on the app store means playing by Apple and Google's rules, plus it puts us out there. Right now it's an if-you-know-you-know situation and you can't accidentally find it, but with app stores that could happen, ruining the low profile we want.
If sideloading takes off though, which I expect in the next couple of years thanks to lots of countries shaking off the US's yoke, BlushDesk would be ready. Here's what it would look like: biometrics like Face ID to open, the app working offline so you can update your calendar in an underground parking garage.
5) When do you think we'll see artificial general intelligence and what do you think it will be like? What would Blushdesk 2.0 powered by AGI look like? I noticed in your FAQ “When it encounters something outside its scope, it hands off to you with full context.” What kinds of things are likely to be outside Blushdesk's current scope and assuming a future AGI incarnation could be almost limitless in scope, what would that future look like for sex workers?
Predicting AGI is like a weekend in Vegas, no one knows what will happen. My honest guess is anywhere from five to twenty years and we'll argue about whether it's "really" AGI for at least another decade after that. I've also learned over the past few years that the most useful AI is often the boring kind that just does one thing well. A master of one, so to speak.
Right now the things outside BlushDesk's scope are:
a. Emotional nuance. It can't tell if a client is nervous and in need of reassurance versus guarded and worth watching.
b. Negotiations where the answer is a maybe. Right now it's very good at "no" and very good at "yes, here are the details." It's deliberately bad at "maybe, depending on how I feel about this person."
c. Risk judgment. When someone says all the right words and something is still off, that's a human call. The AI reads what's written. It doesn't read what's underneath.
What does an AGI-powered BlushDesk 2.0 look like? Honestly, I'm just going to tell you my dream: I tell it I want to tour Berlin and it handles the logistics end to end, flights, apartments, local regulations, local red flags.
Obviously this needs a lot more personal data to function, which brings us back to privacy and jurisdiction. If AGI runs on OpenAI servers in the US, we can't use it. If it's open source on a Swiss box we control, that's a completely different conversation.