πΊοΈ Roadmap
Where Spicy Writer is headed
The context problem
AI models have hard limits on how much text they can usefully read at once. The obvious one is cost: bigger context means bigger bills, and those costs hit hard enough to where even ~trillion dollar companies like OpenAI limit it. Perhaps more importantly, coherence is a problem. Even when a company claims a 1M token context window, response quality drops well before that. Models start mixing up characters, hallucinating more, etc.
That means we can't just stuff everything about a world or story into every request and call it a day. Our vision for solving this on the website has three parts:
π File upload
LiveFor SUPER big lore dumps and reference material. Files are automatically searched "fuzzily" on each request, so only the chunks relevant to what you're writing right now get pulled in.
π Lorebook
LiveAlways-active entries the AI always sees, plus keyword-triggered entries injected automatically when matching terms appear in recent messages. Live, but being improved on continuously.
π Summary
In DevelopmentAn automatic, periodic summary of events so far, so the model keeps a sense of the story's arc even when older messages fall out of the live context window. Probably the hardest one to get perfect.
The goal is for all three to combine into a seamless, effortless writing experience, so you don't have to think about context management or restrictions, you just write.
What's coming
Heads up: this is NOWHERE NEAR exhaustive - not even all the big-ticket items, just a few I wanted to call out right now. Tons of other major features, smaller improvements, fixes, and experiments are happening alongside these nearly every day.
- Q3 SEPA and similar - local payment methods for countries where cards are not the norm.
- Q3 Restore Experimental - Restore the quality and value of Experimental Claude provider - difficult and still some unknowns but I think we can do it!
- ?? Auto-updating lorebook - the AI helps keep your lorebook in sync as your story evolves.
- ?? Summary feature - periodic, automatic event summaries to preserve long-arc coherence.
- ?? High quality text to speech - hear your story read aloud, in a voice worth listening to.