Photon Wants Your Next App to Live Inside a Text Thread

A coffin for app icons, a church in San Francisco, speeches from Vercel, Stripe, and OpenAI executives — Photon's September 17 "app funeral" was staged, but the business case underneath it is not. The startup has quietly signed up more than 40,000 developers, grown revenue tenfold in four months, and now carries $4.5 million in seed funding to keep building the toolkit it hopes will replace the app install screen with a text thread.

That makes Photon a useful specimen for anyone running product, growth, or engineering at a company deciding whether its next customer-facing release should be an app at all — or an agent that lives where customers already are.

A Funeral as a Business Plan

Co-founder and CEO Daniel Tian isn't shy about the timeline problem. "Honestly, it's going to take a while, but I think, directionally, it's inevitable," he told TechCrunch about the shift from apps to agents. The funeral doubled as a developer day, which tells you something about Photon's actual strategy: it isn't waiting for app stores to collapse. It's selling infrastructure to builders who want to hedge now, while apps and agents coexist.

Analysis: the theatrics matter less than the underlying thesis — that discovery, not functionality, is the real cost center of mobile software. Photon's pitch is a direct response to that cost, not a prediction about when apps die.

Meet the Operator Rethinking the App Install

To see why this resonates, picture a product lead at a fintech or insurance startup — the kind of company already on Photon's customer list, which includes Corgi Insurance, Rho, Boardy, Ditto, Fliptexts, and Slashy. This operator owns onboarding and retention. Every new feature currently requires a release cycle, an App Store review, and a push notification begging users to update. Each of those steps leaks users.

Note: this is an illustrative composite built from Photon's publicly named customer categories, not a quoted interview with any specific company.

For that operator, the question isn't philosophical. It's a line item: how many users will complete a native-app install flow in 2026 versus simply replying to a text?

The Old Workflow: Build, Launch, Beg for Downloads

Photon's founders lived this problem before they named it. Tian and CTO Ryan Zhu spent their student years building consumer apps at hackathons and kept hitting the same wall — not building the product, but getting anyone to find it. The old sequence looked like this for most small teams:

• Build a native or web app
• Submit to app stores and wait on review
• Spend acquisition budget to get installs
• Fight notification fatigue and uninstall rates
• Repeat the whole cycle for every feature update

Their workaround was personal and small: a bot that texted friends back as Tian, running over iMessage. When they open-sourced it on GitHub, thousands of developers adopted it almost immediately — evidence, in hindsight, that the discovery bottleneck wasn't unique to them.

The New Workflow: Messaging as the Interface

The managed version of Photon, launched this past April, turns that scrappy fix into infrastructure. It bundles a unified API, an extensible channel framework, a command-line interface, and an observability suite, letting a team build one agent experience and deploy it across iMessage, WhatsApp, Telegram, SMS, RCS, email, and voice without rebuilding the logic for each channel.

The new operator workflow collapses into three steps instead of five:

• Define the agent's logic once, against Photon's unified API
• Point it at whichever channels the target customer already uses
• Monitor and iterate through the observability suite, pushing updates without an app-store review cycle

For regulated industries — insurance, banking, healthcare — the managed platform also carries SOC 2 Type II and HIPAA compliance, which is the difference between a hobby project and something a compliance officer will actually approve.

A Four-Step Rollout Sequence

Based on how Photon itself scaled from side project to platform, a sensible adoption sequence for an operator looks like this:

1. Prototype on the open-source layer. The free version still accounts for 98% of Photon's actual usage, making it the low-risk way to test whether a messaging-native agent fits a specific customer journey before any procurement conversation.
2. Graduate past ten users. Photon's free tier caps out there; anything larger requires moving to one of three paid subscription tiers, which also unlocks the 99.95% uptime guarantee the open-source version can't match on a local machine.
3. Layer in compliance if needed. Teams in healthcare, fintech, or insurance should confirm SOC 2 and HIPAA coverage early, since that's the gate for handling sensitive customer data over text.
4. Connect the surrounding stack. Photon already integrates with Vercel, LangChain, Mastra, Convex, Render, Railway, and Telnyx, so operators building on those tools can plug in the messaging layer rather than building channel-specific connectors from scratch.

The Metrics That Actually Prove It's Working

Photon doesn't publish hard revenue figures, but the proxy metrics it does share give operators a template for judging their own rollout:

• Churn under 3% — a signal that once teams adopt the messaging-native pattern, they don't retreat to native apps.
• Messaging volume up 5x in a single month — a sign that usage, not just sign-ups, is compounding.
• 40,000-plus developer sign-ups with 10x revenue growth in four months — evidence the pattern generalizes beyond Photon's own founders.

For an operator evaluating a similar shift, the equivalent dashboard should track message-thread retention against app-session retention, cost per activated user across channels, and how many feature releases ship without a store review — the metric native apps structurally can't win.

What Could Go Wrong

Analysis and risk framing: betting the front end of a product on iMessage or WhatsApp means betting on platforms Photon doesn't control. Apple and Meta set the rules for those channels, and Meta's own Muse app — which has climbed to No. 1 on app stores — is effectively testing whether the dominant consumer AI agent will be built by the platform owner itself, not a third-party layer like Photon. Tian's own hedge is that developers will still need infrastructure regardless of who wins the consumer agent race, and that the next expansion is agent-to-agent communication — his example being a travel-planning agent that pulls in a flight-booking agent on a user's behalf.

For operators, the practical risk is simpler: a messaging-native agent still depends on a third party's API stability, rate limits, and policy changes. The payoff — faster iteration, no install funnel, access to users inside apps they already check dozens of times a day — has to be weighed against that platform dependency on a case-by-case basis, not treated as a foregone conclusion.

Sources

• TechCrunch AI — Photon held a funeral for mobile apps. Now it has $4.5M to help replace them with agents