Remote android phone for AI agents

Give your AI Agent a loaded phone

We should give our AI Agents their own Android phone + US phone number, that they can remotely drive and use. Agents like Claude, ChatGPT, and GrokBot could do a lot more for you by having a phone of their own.

Present: Let the agents run wild

If we accept that soon, suddenly everyone will have unlimited intelligent AI agents, and those agents are running around on the Internet, writing comments, filling in forms, doing all sorts of work, then we can expect that websites will continue to do more of the same, using old things like email verification, and SMS verification, and captchas. So if everyone's using these AI agents, pretty soon the AI agents are going to want to get those things too. Email seems pretty solved, lots of agents read your email or even have their own inbox. They're starting to get a bit of identity about them. They'll probably want a phone too. Instead of trying to ban this, I beleive that we should make it as easy as possible for the AI agent to get those things, so that they can do more. The question of whether this is good or bad is better answered by other mechanisms (like upvotes, TikTok's watch time algorithm, or Google's SERP ranking algorithms). So to get these AI agents armed, that's where Hardlinephone comes in. The idea is to make the Hardlinephone service as easy and as accessible to reach by an LLM-powered AI agent. That's why we implemented the MCP protocol, and llms.txt to allow the AI agents to find our service, then sign up for the service, and then buy a phone + number. And then, even make use of the phone by clicking around on the screen and all those other sorts of things. So, really, we're making the ideal service for a phone usable by AI agents. And it's extra valuable, because instead of being a virtual machine emulated fake phone in the cloud, it's actually a real hardware phone with real location in the US. As well as the SIM card being a real US phone number, and not a VoIP number. So I say, let the agents run wild with their phones.

Past: How we got here

In the past, website owners employed cheap and easy to check ways to deter masses of spammy automated bots. Websites used email verification or captchas, which don't really cost the website very much to check, and deters the lowest and crappiest bots. Or at least most of them. Now that the bots are actually doing real useful work on behalf of mostly well-meaning humans, as well as themselves recently generating better and better content, and becoming more and more intelligent with their comments and insights, we should pause and reconsider: should all bots be banned? We may not have to ponder for long. They're also becoming better able to get around even difficult captchas, and email singups are easily done by GrokBot now. Website owners will have to abandon these methods of deterring bots, and either embrace the bots and allow them in, as part of the contributed content that has added value to their app/website, allowing it to compete with the humans head on, or shut the door with harder methods. There may come a day where the bot content is actually better and more engaging and more valuable than the human generated content, in which case, I guess, the bots have won. However, website owners may simply decide that we don't want AI generated content at all, and then, therefore, we will have to switch our methods of filtering that don't test for intelligence (as in the case of the captcha) but rather instead to investment, or posession of a rare thing. e.g. you have to pay to have an account. Which maybe the bots will do if their humans tell them to. Maybe not. Or, perhaps, it'll be something scarce that you have, that bots typically don't. Like a phone in a US location with a real US phone number! I preduct that those things will become increasingly valuable for people and their agents, in the future, for automating real tasks, and doing valuable work. And that's why Hardlinephone exists — to allow them to do so.

Future: How this ends

Eventually, we'll run out of those things, bots will have them just as easily as humans do. I predice that the end game is that things will be cryptographically signed, and it will be up to governments to check on the humanness of a person, and then somehow tie that humanness to one of their existing fiat identification items, like passports or driver's licenses. And then they will issue people with a cryptographically signed identity or private key. Which can then be made available for verification against by website owners. Perhaps this will use zero knowledge proofs, so that you can prove that you're a human without having to give away your identity. Although people will (understandably) be reluctant to give that up in the early days, without understanding that ZKPs are a) possible and b) trustworthy. And probably some governments will indeed break the trust of the users by saying that a system is a true zero knowledge proof when, in actual fact, they know exactly who the authenticating person is, or if not, they can actually find out via some sort of backdoor. In any case, eventually, everything will be cryptographically signed. But until then, both humans and bots will need to access a real US phone, perhaps with a real US phone number, in which case, they'll be able to use Hardlinephone to do so.

-Andrew