Stories that remember you.
What a decade behind the DM screen taught me about what an AI Game Master needs to do, and what it should never try to do.
I've been the DM for our homebrew D&D campaign Syrinway for nearly 10 years. Now I'm taking it digital with an AI narrator named Algus. Here's what a decade behind the screen taught me about AI Game Master design.
Read post →Why an AI RPG can fit alongside your tabletop campaign, not replace it.
AI RPGs aren't meant to replace your in-person D&D games. They're meant to live between sessions. When your group can't meet, your tabletop campaign sits frozen. Branching Skies is a pocket campaign with a GM that doesn't forget.
Read post →Five common failure modes, and how we built around each one.
Every AI RPG in 2026 fails at memory in one of five common ways. Here's what each failure mode looks like in practice, and how Branching Skies addresses each one.
Read post →The architectural reason almost every AI RPG fails at memory.
Most AI RPGs forget your story because they treat each turn as a fresh prompt with no persistent memory. Here's the architectural reason, and how we built Branching Skies to fix it.
Read post →A seven-point checklist for evaluating any AI RPG.
A seven-point checklist for evaluating AI RPGs on memory depth. Most products miss at least three. Here's what to look for and how to test for it.
Read post →And why it's time for something built for your phone.
Two failures back to back: AI DM tools that demand more focus than you have time for, and AI DM tools that forget your campaign. Here's why the trend keeps producing both, and what I built when I gave up on the trend.
Read post →The three product shapes hiding behind the same term.
AI RPG in 2026 means three different product shapes calling themselves the same thing. Here's the taxonomy: open-ended generation platforms, chat companions, and curated campaigns.
Read post →A curated 2026 guide for solo TTRPG players who want memory, agency, and a world that remembers.
AI Dungeon got the genre off the ground. It also showed us where the genre breaks. Memory fails. Story Cards get ignored. The world drifts mid-session. Here are five alternatives in 2026 for solo TTRPG players who want a campaign that holds together.
Read post →Voyage builds worlds from prompts. Some players want a world built by hand. Here's what to try if that's you.
Voyage launched in 2026 with serious funding and a serious bet: infinite generated worlds beat one curated world. Some players want that bet. Others want the opposite. Here are five alternatives for the players who want a story authored before they showed up.
Read post →Six mobile AI RPGs ranked by how well they remember what you did. The criteria most reviews miss.
Most mobile AI RPGs forget your story. A few don't. The difference comes down to architecture, not marketing copy. Here's a buyer's guide that ranks six apps in 2026 by the criteria that actually decide whether your campaign holds together.
Read post →The architectural reason your AI Archmage casts Hold Person six times in a row, and what a companion combat system should actually do.
Your AI controlled Archmage casts Hold Person against a five enemy encounter. The next round he casts Hold Person again. The architectural reason this keeps happening across AI RPG products is the same problem behind every memory failure: LLMs are bad decision-makers and great storytellers. Here's the fix.
Read post →