I wanted to build a graveyard for apps.
A quiet place for projects that promised to change our lives, but only changed their passwords and pricing.
Each grave would have a name, dates and an honest cause of death.
“Another task manager. Organised everything except its own business model.”
“Meditation app. Sent eleven notifications a day to remind you to be calm.”
“Social network. Founder last seen online on Tuesday.”
For the particularly promising ones, a “Resurrect” button. After email verification, obviously. Some traditions are stronger than death.
While I was thinking about epitaphs, someone in San Francisco had already held the funeral.
They buried the whole category.
Including the ones that were still working.
Startups have a wonderful attitude towards competitors: if they’re not dying quickly enough, you can always begin the ceremony.
On 17 September, Photon held its App Funeral, doubling as a developer event. According to the company’s own recap, more than a hundred founders and developers attended. Guests received beta access to Photon 2.0. [1]
Inc. describes a church venue and a coffin containing app icons. [2]
Some companies prepare a presentation.
These people made final arrangements.
I respect attention to detail. Especially when the organiser’s confidence stands in for checking whether anyone has actually died.
Imagine being the app. You’re still processing payments. You have three open support tickets. Someone onstage is already remembering you fondly.
Photon raised $4.5 million in seed funding. Gradient and A* co-led the round, with Vercel among the other participants, according to co-founder Daniel Tian’s announcement. [3]
The money was for the business, not the coffin.
But a coffin is considerably more memorable than “messaging infrastructure for AI agents”.
The cemetery turned out to have excellent networking.
First, he just didn’t want to reply himself
According to the founder’s account published in Inc., Tian connected AI to his iMessage conversations so it could reply in his style. His friends, he says, didn’t notice the substitution. That’s the founder’s story, not an independently conducted experiment. [2]
I’m not sure which part is more insulting.
That a machine is answering for you.
Or that your friends are perfectly happy with the arrangement.
Perhaps they thought: finally, he’s actually reading the messages.
Maybe the future of communication began when someone realised that maintaining a friendship didn’t require his personal attendance.
But the party trick had a practical side. The founders had been building consumer apps and repeatedly hitting the same problem: persuading people to try them. Hence the idea of bringing a service into the conversations people already have. [2]
And here, I understand them.
I might want a new service. That doesn’t mean I want a new account, a seven-screen induction and an email asking me to rate our introduction.
We haven’t done anything together yet, and you already want to know whether I’d recommend you to my friends.
Calm down. I haven’t even memorised the password.
Then comes the request to allow notifications.
I love this stage of a relationship: we barely know each other, and you already want permission to interrupt me.
A funeral with an App Store link
Photon builds tools connecting agents to familiar communication channels. Its 15 October announcement of Photon 2.0 lists iMessage, WhatsApp Business, Telegram, SMS, RCS, email and voice calls. [4]
The proposition makes sense. Instead of another place you have to visit, the service appears in a conversation.
Then you get to the details.
The company that held a funeral for apps advertises iMessage mini apps on its homepage. [5]
I wanted to see how literally to take that.
Quite literally, as it turns out.
The public Spectrum code includes a file called packages/imessage/src/remote/app.ts. It contains the identifier of Photon’s own iMessage extension and its App Store ID. The code builds a card that opens a supplied URL inside that extension. [7]
The documentation explains that, for these cards, the recipient downloads the Spectrum iMessage App once. Mini apps then open directly inside Messages. This applies to that particular feature, not to every ordinary text conversation with an agent. [8]
There really is a Spectrum — Apps in Messages listing in the App Store, with Photon Codes named as the developer. [6]
Apps are dead. To access selected afterlife features, please download the app.
If the deceased won’t open, check your iOS version.
This doesn’t make Photon’s idea ridiculous. One container for many small services can be more convenient than dozens of separate installations.
There is simply some distance between “people don’t need another app for every task” and “apps are dead”.
About one marketing ceremony’s worth.
There’s a server running underneath the coffin
In its technical announcement, Photon describes macOS clusters, Apple Account registration and problems it had to tackle: kernel leaks, inconsistent system state and encryption-related failures. The team built a component called Astrolabe to manage macOS state. [4]
This is more interesting than the funeral.
Because there’s actual engineering here. Without the black ribbons.
The user sees a simple message. Someone else makes sure it arrives, that the reply comes back, and that the system remembers who is talking to whom.
For me to install nothing, somebody else has to install and maintain an awful lot.
That’s a perfectly useful arrangement. We’ve been paying to avoid other people’s complexity for years.
The magic usually ends where the engineer’s on-call shift begins.
But disappearing from my screen isn’t the same as ceasing to exist.
This also applies to subscriptions, incidentally.
Who inherits the deceased?
If an agent reaches me through iMessage or WhatsApp, those apps don’t become less important.
They become the entrance to even more services.
Photon’s standard terms explain that message delivery depends on third-party platforms it doesn’t control. Their rules and limits can change. [10]
So Apple and Meta aren’t in the coffin at this funeral.
They own the venue.
And they can change the lease while you’re reading the eulogy.
Here’s the question I find worth discussing: if ordering, paying, searching and working increasingly begin with one conversational assistant, who decides which services it offers us?
Will we see alternatives? Can we check what it actually did? Will we notice when helping becomes selling?
These are questions about the model’s future, not misconduct I’ve established at Photon.
But the difference matters.
I used to open a shop and know I was in a shop.
Now the shop can text: “Hi, I’ve been thinking about how to help you.”
Very sweet. Why does the help come in my size and three colours?
Intimacy, on the business plan
There’s another detail on Photon’s pricing page.
Business costs $250 per line per month. Among its features is cold outreach to up to fifty new contacts a day. Meanwhile, the standard terms prohibit spam and unsolicited messaging and require the necessary consent. How the company distinguishes those scenarios needs explaining. The phrase “cold outreach” alone doesn’t establish unlawful messaging. [9][10]
As a glimpse of the future, though, it’s revealing.
We’re pleased the assistant will be where it’s convenient for us.
Businesses are also pleased they’ll be where it’s convenient for us.
I wanted useful services closer. Sales heard an invitation.
Messages are a particular kind of place. Your children write there. Your friend. The person whose reply you’re waiting for.
That interface already has an emotional reputation. You don’t have to build it from scratch, as you would for an unfamiliar website.
It can make a good service easier to reach. It can also make a commercial offer feel more personal than it deserves.
You see “typing…” and briefly hope somebody has missed you.
Possibly. Your monthly payment, at least.
The shop window and the contract have different voices
Reading the public pages turned up a less amusing discrepancy, too.
Photon’s homepage advertises support for healthcare teams compliant with HIPAA, the US framework governing protected health information. Its standard terms, however, say the service is not intended to meet HIPAA requirements and Photon is not a Business Associate under HIPAA. [5][10]
There may be a separate agreement for healthcare customers. Without that agreement or an explanation from the company, a conclusion that it has broken the rules would be premature.
But a buyer still needs to establish which promise applies to them.
The future may be button-free. The small print looks likely to outlive us all.
There’s no plot for it in my cemetery. That one will be burying us.
What if they’re right?
Photon has a strong point.
Not every service deserves a separate icon on my phone.
To move an appointment, check an order or get a short answer, a conversation really can be better than a menu.
But there are things I want to see.
A map. Several options side by side. Payment history. The field I’m changing. The result before I confirm it.
“I’ve done everything” is a lovely sentence.
Especially when there’s a way to check.
Because the gap between “I did everything” and “I did exactly what you asked” can accommodate an entire support department.
So I don’t see proof that apps are dying. I see a fight over where my next action begins: with an icon, a search, a message or a request to an assistant.
And that’s a big fight. Big enough to manage without a pretend death certificate.
Incidentally, Photon’s September recap names BailMeOut among its competition winners: a demo agent designed to rescue people from bad dates and meetings. [1]
This is my favourite detail.
Onstage: the end of an era.
In the demos: finally, someone can help me go home.
Humanity taught a machine to write code, analyse data and hold a conversation. Then, quite reasonably, asked it to come up with a reason to leave.
Technology becomes much more persuasive when it stops saving civilisation and tackles a specific uncomfortable evening.
My app graveyard may still be useful.
For services that forgot their users. For lovely ideas that never found a business model. For yet another product that wanted my address but not my opinion.
I’ll simply add a separate section for those declared dead by their competitors.
Convenient access to iMessage.
And an “I’m still here” button.
Because burying apps that people still need isn’t predicting the future.
It’s booking the plot with remarkable confidence.
Preferably, check the pulse before the presentation.