The story
Two decades, one idea
Digital assistants through every era of the field: scripted intents, word-sense disambiguation, SIML, and now large language models. Half a working life on one idea, now brought to the public.
School days
The talking face
Speech synthesis
Where the obsession started
Small desktop apps with a round emoji face and a speech synthesizer. Whatever you typed, the face spoke back and tried to lip-sync. Built for fun, shared with nobody, and never really put down since.
Highlights
-
Typed text became a spoken, animated reply.
-
The idea took hold: a computer character you talk to.
Jargon
An assistant for one
Scripted intents
The first real assistant
A personal assistant built for its author and shown only to friends, never released on the internet. When Iron Man hit theaters, friends had a nickname ready: the guy already had his own digital assistant.
Highlights
-
Built for one user, refined by daily use.
-
Fully scripted intent matching, years before mainstream voice assistants.
2013
The first public video
IronPython NLP
The first public demo
The first assistant to reach YouTube. A 3D head placed inside an existing game engine, with facial and display animations, and IronPython doing word-sense disambiguation to pull terms and entities out of user sentences. All of it scripted, none of it a language model.
Highlights
-
3D head and facial animation running in a game engine.
-
Word-sense disambiguation for entity extraction, pre-LLM.
Syn Virtual Assistant
The shipped product
SIML
From hobby to product
Syn Virtual Assistant was the shipped product: sold on Steam as a paid desktop assistant, rebuilt generation after generation on the game engines of its day. It ran on SIML, the Synthetic Intelligence Markup Language: its author's own bot language, a step beyond the static markup languages of the era, letting authors train a small local model for intent matching instead of writing only fixed rules.
Highlights
-
Sold on Steam as a paid desktop assistant.
-
Multiple generations across two decades of engines.
The framework years
VA Framework and Oscova
Bot framework
From product to platform
Syn was reworked into the VA Framework, shipped on Steam, which let anyone build an AI assistant of their own. Alongside it came the Oscova Bot Framework: not an assistant but a bot development framework in its own right, with dialogs, intents, and entity recognition, distributed through the Syn.Bot package on NuGet since 2015. The Oscova brand was born there and is registered today; simlbot.com and oscova.com are still live, and oscova.com will soon point here.
Highlights
-
VA Framework on Steam: assistant tech in other people's hands.
-
Oscova Bot Framework, on NuGet since 2015, gave the brand its name.
Today
Oscova Agents
LLM era
The same instinct, the modern stack
Oscova Agents is the LLM-era successor: an agent OS for home and office, running local models on hardware you own, with the privacy and egress controls a household assistant deserves. At heart it is the same private assistant its author has always built.
Highlights
-
Durable agents as services, not chat tabs.
-
Local models first, your data stays home.
