Post for: #archaic (10 Posts)

First code. Two sketches. One just listened. The other tried to talk. Neither worked well. But I could see signals on the bus.

Swapped the CD4049 for a CD4050 I ordered. I'd been using the inverting buffer because it's what I had in my stash, negating the signal in software. Worked, but added confusion to every debugging session. The CD4050 passes signals straight through. One less thing to think about.

Next day I wrote a dozen more. Tried interrupts, then polling, then fuzzers that sent every byte from 0x00 to 0xFF to see what the typewriter does. All based on FozzTexx's documentation, from reverse engineering the communication between the typewriter and the PWP. Getting closer, but nothing reliable.

Day after I fried the typewriter, FozzTexx replied on Mastodon. "What model Smith Corona do you have?"

Told him what had happened. SD-250 Spell Right II. Fried it by shorting pin 3 and pin 4. Turns out it could also function as a smoke machine.

He created a GitHub repo and shared the schematic for Typarona, his Smith Corona project - an adapter bridging the typewriter and a modem - to get it to behave as a teletype able to dial into his BBS.

Hardware schematics only. Said he'd been in the middle of a massive firmware rewrite two years ago and needed to hunt it down.

Found it that afternoon. Pushed it to the repo. Said he's not sure if it even builds.

Didn't matter. I could read it.

SD-265 arrived from eBay yesterday. Spacebar torn out. Suboptimal, but usable for developing and testing.

While the replacement was in transit, I had time to think.

FozzTexx's documentation proved the typewriter could have a two-way conversation with a computer. Not RS-232. Something proprietary. But documented. Solvable.

So: what if the computer on the other end was ChatGPT?

Type a message on the typewriter. An ESP32 reads the keystrokes, sends them to the ChatGPT API over WiFi, and prints the response. Character by character. On paper.

A 1986 typewriter. Talking to an AI.

That was the simple version. It lasted about a day.

The RS-232 adapter arrived from Amazon. Connected it. Nothing.

Tried different baud rates. Different settings. Nothing.

The DB-9 connector looked like a serial port. It wasn't. Smith Corona used a proprietary protocol. The connector shape lied.

Finding Smith Corona documentation is hard. The company closed in 2013. 127 years of making typewriters, then gone.

Most of what remains online: eBay listings, repair manuals, nostalgia forums. Not protocol specs.

But people document obscure things. Found a trail. Reddit post linking to a Twitter thread linking to a Mastodon account. All leading to one person: FozzTexx.

He'd already done the hard work and very meticulously. Years ago. Reverse-engineered the entire protocol. Built a working BBS terminal out of a Smith Corona. Published everything on the Mastodon thread in real time.

I learned things I didn't know existed. Electric typewriters had PWP modules that turned them into word processors. An entire era of technology between manual typewriters and PCs.

And FozzTexx documented what every pin on the DB-9 does. Data in, data out, clock, request. And pin 3: +36 volts, to provide power to the PWP module.

Walked my son to Aikido. He's ten. Usual route.

Old computer on the curb. Heavy. Clean.

"Dad, come on, we'll be late."

"You know me. Let's take it home and take it apart."

I opened the case. It was a typewriter! Full keyboard, LCD display, of all things. Print head on a rail. Not a mechanical typewriter. Something else.

He knows me. I put it in the reusable shopping bag, below the Gi. He volunteered to help carry it, each holding a handle. We still made it to Aikido on time.

Got home. Plugged it in. Expected nothing.

It turned on.