AI companions
Tolan
Tolan is Portola’s AI companion app, pairing voice conversation with an animated alien that responds to touch and remembers its relationship with the user. Available for iPhone, it lets people discuss interests, relationships and goals, share photos and design a virtual planet. Its purpose is to help users navigate everyday overwhelm while encouraging connections with people in their lives. Characters gradually share their own interests and experiences, with that progression deliberately tuned to model healthy attachment.
Portola’s founders are Quinten Farmer, its CEO, Evan Goldschmidt, its CTO, and Ajay Mehta, its president. Tolan soft-launched in 2024, combining underlying AI models with systems for listening, responding and remembering. Its character-design work chose nonhuman aliens to support friendly interaction without inviting romantic roleplay. The initial animation system used a state machine to coordinate authored keyframes and procedural movements: this gave characters an expressive presence across conversations while native model processing of emotion and tone could not yet reliably drive animation.
The company’s website reports more than 100,000 subscribers and over two million hours spent with its guides. Portola offers paid features and recurring subscriptions. Its product blends reflection with practical help, such as organizing a day or discussing homework, while its service terms classify the experience as informational and recreational rather than professional healthcare or counseling.
Explore the recordings
Tolan’s supplied archive contains one recording: Paula Dozsa’s voice-first AI companion talk. The paths below explore different parts of that recording rather than separate talks. Technical results, usage figures, and product observations are claims reported in the recording; this catalog does not establish Tolan’s current implementation or product status.
Start with the voice interaction
For the clearest introduction, follow Dozsa’s discussion of latency, interruptions, and turn boundaries. Identified in the recording and organizer record as a Tolan iOS engineer, she explains why spoken conversation needs different handling from text. Her reported latency regression and turn-taking improvements provide concrete examples of how interface behavior affects the user experience.
Trace the pipeline and model choices
For an engineering path, focus on the stage-by-stage voice pipeline and model routing. Dozsa describes measuring everything from end-of-utterance detection to playback, reports faster time to speech after a model/API change, and discusses routing emotionally serious turns differently from lighter exchanges. Treat the cost and retention findings as reported experiments, not general guarantees.
Explore memory and character continuity
For companion design, follow the material on retrieved memories, nightly compression, and per-turn context. The useful connection is between memory maintenance and character behavior: Dozsa describes separating stable memories from volatile prompt content, while tone monitoring adjusts delivery without changing character identity.
Finish with agents in the development workflow
For team practices, turn to implementation and review agents, PR shepherding, bug triage, and character evaluation. Dozsa describes MCP-connected tools and a multi-perspective find–fix–verify process. She also reports improved reliability alongside increased AI coding; the recording does not establish that AI coding caused those improvements.
1 talk
Newest first
1 speaker at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.