Ancient Web: Sensation Bot Went From BBC BASIC to a Year-2000 Chat Window

Part of the Ancient Web Collection.

Modern chatbots have trained everyone to expect fluency. Sensation Bot offers a better historical question: what could a conversational program do when it was running on a BBC Micro and matching words in BASIC?

Visit Sensation Bot.

The site's unusually detailed development history starts around 1993. The first program ran locally on an eight-bit BBC Microcomputer and answered selected phrases with hard-coded replies. It could also converse with a second copy of itself, though understanding was not part of the bargain.

From FidoNet to the browser

In about 1995, a DOS version written in QuickBASIC began reading and replying to messages on FidoNet, the network connecting many bulletin-board systems. That let the bot encounter a wider audience without yet being a web application.

The author says a subsequent standalone edition used the MoonRock programming language. By 1998, a new matching scheme scored candidate responses instead of checking only fixed phrases, and the program began saving human input for future responses.

In mid-2000, the project moved online in a mixture of C, Unix shell scripts, and PHP. The developer registered the sensationbot.com domain later that year. Its initial personalities included a general mode and an uncensored one that recycled unfiltered user input. The site's retrospective says that could produce offensive nonsense and occasional unexpected switches between languages.

A 2010 rewrite moved the server implementation to PHP and introduced a chat-window interface instead of reloading the page for every line. The site now links several simulated personalities, including a general-audience option and clearly labeled adult-only ones.

A chatbot that explains its own trick

The FAQ is refreshingly specific about the program's limits. It describes keyword matching, a restricted dictionary, scoring replies, spelling corrections, pronoun swaps, and fallback random answers. It says outright that the bot does not genuinely understand the conversation.

The FAQ also dates its antenna-like logo to July 1998 and points to a January 2001 archived capture. That evidence is unusually helpful for distinguishing the historical project from modern chatbot branding.

The site's current chat feature was not tested for this article. Its real value here is the surviving developmental record: a practical bridge between hobby BASIC, BBS culture, and web-based conversational experiments.

Read the bot's own history and how its matching system works.

Part of the Ancient Web Collection.