About me
Rich Pennington: compiler writer, retired, still curious.
I've been programming for a very long time. For about twenty years I worked at Introl, where I wrote the Introl C compiler. That meant code generators, assemblers, linkers and libraries for a long list of processors (eventually the 6801, 6809, 68HC11, 68HC12 and most of the 68K family) and a long list of customers who each needed something slightly different. I learned a lot from both.
Some highlights from those years:
- The 6809. My favorite, because it's what got me to write my first C compiler. To my mind it has the best instruction set of any eight-bit microprocessor, with a clean, orthogonal feel reminiscent of the PDP-11. I once had my 6809 compiler running natively under OS-9, generating fully position-independent code for itself. Writing a position-independent linker, plus startup code that fixed up initialized data at run time, was one of the most fun problems I've worked on.
- The NS32000. A beautiful, VAX-like design, and doomed.
- Modula-2. We also did a Modula-2 compiler, with extensions made specifically for GM. It was generally available as part of the Introl tools.
After Introl I spent about sixteen years at GE Healthcare programming CT scanners, then about six at Johnson Controls programming embedded building automation controllers. Writing compilers was much more fun, so while I was at GE I kept doing it on the side: I wrote ELLCC, a cross-compilation toolchain built on LLVM and Clang for building C and C++ programs for many targets.
These days I'm retired. I'm working on a secure IoT project, I run a small network at home that I built about eight years ago, and I study languages for fun. My German is decent; I first learned it at twelve, when I spent time in Munich.
Why AI, and why now
A few weeks ago I started using Claude, an AI assistant made by Anthropic, for real work: not toy questions but the kind of system administration that I'd been putting off for years. It turned out to be far more capable than I expected, and also capable of confident mistakes. Both of those things seemed worth writing about.
My half-joking reaction after the first two weeks was: maybe they should have given me this fifty years ago.
How the posts are made
Claude writes the first draft of each post from the record of what we actually did: our conversation, the commands that ran, and the project notes we kept along the way. I read it, correct it, and decide what goes up. If something here is wrong, the mistake is mine for letting it through.
I leave out passwords, addresses and anything else that would help someone attack my network. The stories are real; the specifics are deliberately vague.
Get in touch
Questions, corrections or your own stories about working with AI are welcome. Write to me at rich@pennware.com.