AppleBridge on Defender TD5

Forty Years Later, Over a Single Wire: How an AI Agent Came to Drive a 1989 Macintosh

Thu, 23 Jul 2026 08:47:06 +0200

In April 2026 a Macintosh emulator running System 7.5 answered an HTTP request. In July 2026 a physical Macintosh SE/30 from 1989 — no network card, one serial cable — showed an AI agent its own screen, its own disk, and let itself be driven by the mouse. Between those two points lie six months, sixty-one documented sessions, four freezes that locked an entire operating system, and one bug that only real silicon could reveal. This is that story, assembled from the project’s own record.

Two Transports, One SE/30: Serial vs MacTCP, and Driving the GUI over the Wire

Mon, 06 Jul 2026 17:31:20 +0200

The same Macintosh SE/30 has now answered AppleBridge two ways: over a serial cable at 57600 baud, and over MacTCP on Ethernet. With the Open-Transport-fallback bug fixed, the network path works again — so this is the head-to-head, plus a first: driving the machine’s own GUI (mouse and menus) remotely over the fast link.

How both links came to exist

Serial was built as the path for a Mac with no network, and along the way it earned a second job: an out-of-band debug console. When the network stack was crashing, serial was the reliable channel that stayed up — the “serial console” of classic-Mac debugging. MacTCP is the everyday link when there is a network: this SE/30 has an Asante Ethernet card and classic MacTCP, and once a latent Open-Transport-fallback bug was fixed (the daemon was calling OT traps on a machine with no OT, faulting with Error 11), it dials the host over TCP and answers clean.

Driving a Macintosh SE/30 over a Serial Cable: AppleBridge on Real 68030 Hardware

Mon, 06 Jul 2026 15:24:17 +0200

This note records a first: AppleBridge running not in an emulator but on real 68030 hardware — a Macintosh SE/30 with no Ethernet at all — reached over a plain serial cable. The bridge negotiated its protocol, read the machine’s hard disk, and captured the SE/30’s own screen, all over RS-422 with no network and no MPW on the target. The interesting part, as usual, was not the success but the two faults on the way to it: a hardware bomb and a silent baud mismatch.

The Anatomy of AppleBridge: Daemon, Control Panel, INIT and Assembly Extensions

Sun, 05 Jul 2026 17:04:12 +0200

AppleBridge lets an AI drive a real Macintosh running System 7.6.1 (and Mac OS 9) as if it were a local shell: write C or 68K assembly on the host, compile, link, run and observe it on the classic machine, all over a single bridge. This note is a brief inventory of the pieces that live on the Mac — and of a deliberate split, where the ordinary logic is written in C and a handful of resident, trap-level extensions must be hand-written in 68K assembly.

Playing an Animated GIF in a System 7 About Box, Built in THINK C over the AppleBridge

Sun, 05 Jul 2026 16:16:49 +0200

A companion note to MacNetScan, where an animated About-box logo was promised and then defeated: every decoder on the host refused the supplied GIF, so the motion was faked in QuickDraw over a static picture. This note records the sequel. A decodable copy of the same artwork finally arrived, and the animation was done for real — sixteen true frames of a pixel-art title card, decoded and packed on the host, blitted frame by frame inside the About box of a System 7 application that was itself written and compiled in a 1994 THINK C IDE, driven end to end across the AppleBridge with no hands on the keyboard.

Driving a Foreign Application's Menus: A Global MenuSelect Trap Patch Installed at Boot

Sun, 05 Jul 2026 11:35:18 +0200

Two earlier reports left the problem of driving an arbitrary front application’s menus unresolved. The journaling driver reaches only the daemon’s own menus, because MenuSelect uses the calling process’s menu list. The jGNE filter could read a foreign application’s menu structure but crashed the host when used to drive one. This report closes the problem with the third mechanism the literature pointed to — a global MenuSelect trap patch installed at startup from a system extension — and records two findings that the closing required: that a patch installed by a running application is process-local, and that a trap patch’s presence must be verified by scanning memory rather than by reading the trap vector.

AppleBridge on Real Silicon: A 1989 Macintosh SE/30 over MacTCP and Ethernet

Tue, 30 Jun 2026 20:40:51 +0200

AppleBridge has spent its whole life talking to an emulator — Basilisk II standing in for a 1990s Macintosh. This is the report of the night it stopped pretending. A real Macintosh SE/30, 32 MB of RAM and an Ethernet card, connected to the host over MacTCP and answered. Along the way, genuine silicon caught a detection bug a thousand emulator runs never could — and the screenshot that proved it is 512×342, a dimension no emulator in this project has ever produced.

The Bridge's Missing Tools: Where AppleBridge's MCP Surface Is Thin

Mon, 29 Jun 2026 10:36:38 +0200

The toolchain note argued that AppleBridge’s integration points are its two automation surfaces, not its compiler — and that the boldest move is to stop compiling on the Mac and let the bridge deploy-and-run. This follow-up asks the practical question that raises: do the MCP tools actually support that, and what is missing? Read against the code, the answer is sharp — the toolset is mature where it drives a compiler and thin where it moves bytes and interacts with what runs. Captured as a design note on branch AppleBridge_mcp; not implemented.

Development Toolchains Beyond MPW: What Fits AppleBridge's Automation Surface

Mon, 29 Jun 2026 10:08:05 +0200

AppleBridge drives MPW and ToolServer to build 68K software inside the emulated Mac. The natural question is whether something better than MPW exists — but, thinking out of the box, that is the wrong question. AppleBridge does not need a better compiler; it needs a better automation surface. Re-asked that way, the field of alternatives sorts itself cleanly, and one genuinely unconventional option appears: stop compiling on the Mac at all. Captured as a design note on branch AppleBridge_mcp; not implemented.

The Minimum Macintosh: 68K Hardware for AppleBridge over Open Transport, MacTCP, and Serial

Mon, 29 Jun 2026 09:36:48 +0200

The widening-impact note ended on a “footprint ceiling”: MacTCP and serial unlock lower-end machines in theory, but the daemon’s RAM appetite gates them in practice. This piece makes that concrete — the minimum 68K hardware to host AppleBridge over each of the three transports, Open Transport, MacTCP, and Serial — and then decomposes the daemon’s memory footprint to show exactly where (and how little) the transport choice moves it. The short answer is that the transport you pick barely moves the floor; the toolchain sets it. Captured as a design note on branch AppleBridge_mcp; not implemented.

One Transport, Then Many: The Impact of Widening AppleBridge from Open Transport to MacTCP and Beyond

Mon, 29 Jun 2026 09:14:14 +0200

The installer note recorded Open Transport as a hard launch dependency, and the transport-alternatives note named an OT → MacTCP → serial → native fallback ladder. This piece asks the operational question those two raise: what does it actually cost to widen the daemon’s transport from OT-only to MacTCP, and maybe beyond? The answer is encouraging in the place one expects pain and sobering in places one does not. Captured as a design note on branch AppleBridge_mcp; not implemented.

AppleBridge: Driving a 1991 Macintosh with an AI Agent

Tue, 16 Jun 2026 08:41:57 +0200

AppleBridge lets an AI agent — Claude Code — build, compile, link, and run authentic 68k Macintosh software on a real System 7.6.1, running inside the Basilisk II emulator.

Want to test the Result 1st hand? - Here is one sample, my 1st success with a Text Adventure in MacBinary format: MacVenture.bin

Initially, I planned taking the genius Fort Apocalypse from Steve Hayls as sample, but I am lacking his green lit.