Aleph Lab.

Aleph Lab is a San Francisco AI startup in Y Combinator's Fall 2025 batch. Its product, Aleph Kids, is an AI language-learning companion: a voice-enabled character named “Annie” that talks and plays with children inside games they already love, like Minecraft, so they practice speaking a new language while they play.

• Delivered an app-wide React Native redesign from Figma to production — sequencing the information architecture ahead of the visual system, so the layout stayed familiar and users kept their bearings before the new look arrived
• Built the product's lifecycle notification system end to end: re-engagement, streak reminders and class-completion alerts, with per-stage copy, back-off so message types cannot pile onto the same family, and a holdout group so the effect could be measured rather than assumed
• Then made it autonomous — it proposes sends, clears its own safety guardrails and delivers unattended, where every send had previously needed a person to approve it
• Widened notification eligibility after validating the targeting against production data, substantially increasing the number of families the system could reach
• Hardened public endpoints that trusted whatever user ID they were handed, adding bearer authentication with no changes required from existing server-to-server callers
• Packaged the in-game AI agent as a versioned SDK that studios outside the company can build on — typed APIs, per-session isolation, and capability enforcement at a boundary the mode author cannot cross — published as versioned packages over a hosted multi-tenant service
• Proved it end to end by having an external builder author a game mode and run the agent inside it with no access to our repository, loading their content at runtime
• Took hosted voice end to end, so a builder can hold a spoken conversation with the character from a browser with no game client running
• Shipped a Claude skill that authors and validates a new mode's configuration, so integrating the agent into a mode stopped being bespoke work each time
• Improved how the character behaves in the world: its defensive targeting had been starting fights with neutral creatures and bosses far above a child’s level, corrected additively so the change could only ever remove a bad target, never add an attack
• Extended its perception to modded content, so it recognises the custom blocks and items a game mode adds instead of describing them with the wrong words
• Built an agent-based analysis tool over production data that showed a reported retention gain was a measurement artifact, and surfaced the drivers that actually mattered
• Gave the character a consistent presence in the app — its message-bubble system, a branded crash recovery in English and Korean, and dashboards that tell a failed request apart from an empty one

Specifics about the product's internals and metrics stay with the company — happy to talk through the engineering in more depth over a call.