In the latest spectrum-ts release, we added opt-in OpenTelemetry tracing for Spectrum’s messaging layer.

Astrolabe is a declarative macOS configuration framework. You describe the desired state of a machine in Swift — packages, services, system settings, and related setup work — and Astrolabe keeps the machine moving toward that state over time. This update adds a built-in way for the Astrolabe binary itself to update after it has been installed.

Spectrum 1.6 is a fundamental redesign. We collapsed the entire platform surface from seven action methods down to two primitives — an input stream and an output dispatcher — because Spectrum is a tool for building agents, and the old surface did not reflect what agents actually are.

Spectrum’s terminal provider now runs a fully asynchronous chat UI in your terminal, with support for reactions, replies, and inline image rendering. Test agents without leaving your editor.

The developer world has decided LeetCode is dead. AI can solve any hard problem in seconds. Grinding algorithms is a hazing ritual from a bygone era. We have heard the arguments. We disagree — and we are doubling down. At Photon, every engineering hire goes through a live LeetCode interview. Not because we are nostalgic. Because we are testing for one specific thing that matters more now than it ever has: **the ability to find the meta**.

Agents are powerful, but they are stuck behind dashboards and apps most people will never use. Spectrum is an open-source SDK and cloud platform that connects your agents to iMessage, WhatsApp, Telegram, and the messaging interfaces where human life already happens.

Every Mac has a hidden expiration date. After exactly 49 days, 17 hours, 2 minutes, and 47 seconds of continuous uptime, a 32-bit unsigned integer overflow in Apple's XNU kernel freezes the internal TCP timestamp clock. Once frozen, TIME_WAIT connections never expire, ephemeral ports slowly exhaust, and eventually no new TCP connections can be established at all. ICMP (ping) keeps working. Everything else dies. The only fix most people know is a reboot. We discovered this bug on our iMessage service monitoring fleet, reproduced it live on two machines, and traced the root cause to a single comparison in the XNU kernel source. This is the full story.

MimiClaw built a personal AI assistant that runs on a $5 chip. With Photon, it now lives natively in iMessage - the app 1.3 billion people already text from every day

At Photon, we only hire 10x engineers. We shipped a full SDK in a week, then built the infrastructure to support it in a month. That kind of work usually takes big companies a year. But as we moved faster, one thing started slowing us down: the CI/CD pipeline—releases, versioning, testing, and binary builds. In 2026, we're writing software at a speed we never imagined possible, but our CI/CD is still stuck in 2015.

At Photon, we keep asking ourselves a simple question: what should AI and agents actually look like in the future? Will we really open a browser, type in a URL, and use agents as if they were just another SaaS tool? Will we download yet another app just to “chat with an AI boyfriend or girlfriend”? This feels unexciting and it certainly does not match the futures we see in sci-fi scenes. We believe that in the world we are heading toward, AI should not appear as a “feature” or a “tool.” It should feel like a type form of life, deeply woven into our social structures. Our kids will not be surprised when they see AI as we were. They will not treat it as a cold, external program as we treat our macs and phones. They will look at AI the same way they look at friends and classmates. In that world, AI becomes a first‑class citizen in our society. With that in mind, we started asking what we can build today that moves us in that direction. One answer we kept coming back to was iMessage. In the United States, almost everyone uses iMessage every day. Millions of messages flow through it constantly. It might be the most natural and native interaction surface for agents in this era: the agent shows up in your conversation list like a friend, and even joins your group chats. So we decided to turn this idea into reality and build infrastructure that lets AI exist in a truly “human” way inside iMessage. That is how we arrived at imessage-kit - an open‑source TypeScript SDK for controlling iMessage. It lets developers send, receive, and orchestrate iMessage messages through code. Along the way, we had to work through a lot of technical constraints and ended up re‑imagining how AI can communicate with people. This blog focuses on the technical side of that journey: how iMessage works under the hood, what it takes to build a reliable SDK around it, and how that unlocks new interaction patterns for agents. We will save more speculative interaction design and UX experiments for future posts.