←

Shape Rotator Graduation Day

Shape Rotator’s first cohort completed a ten-week accelerator focused on applying cryptography and AI research to working products under real constraints.

During Graduation Day the teams demoed their projects one last time while the cohort's own systems watched, scored, and visualized the night.

Project Presentations

24 presentations

Etherea

Turns speech, music, and other live signals into projected visual environments intended to make shared spaces more responsive and socially engaging.

Presented by James Barnes

The Brainrot Box

Transforms live meeting transcripts into reactive visuals using private inference and a feedback-driven improvement loop.

Presented by Andrew Miller

Pixel Rotator

A creative workspace for photo, video, and live production, with shared controls and a future agent layer that can learn from each creator's editing process.

Presented by OGNodeFather

Pitch Rotator

Gives early-stage founders a short, repeatable camera loop for practicing their pitch with focused feedback.

Presented by Albiona Hoti

Conclave

A private transcription workspace designed to keep meeting records, voice fingerprints, edits, and redactions under participant control.

Presented by Prakhar Ojha

Synclave

Helps developers move privacy-sensitive applications from managed hosting into confidential-computing environments without rebuilding them for a new platform.

Presented by LSDan

TinyCloud

Builds infrastructure for private agent applications in which users retain ownership of their data and grant apps limited authority to use it.

Presented by gbafa

Topos

Local-first personal database that turns private data into useful signals with controls for how agents access it.

Presented by Jonny Johnson

Smithers

Agent workflow system designed to make long-running work inspectable, resumable, and adaptable as models change.

Presented by Fucory

BitRouter

A context-aware LLM router that continuously improves agent workflows.

Presented by kel

elizaOS

Framework and product stack for autonomous AI agents that think, learn, and act autonomously.

Presented by Shaw

Novel Tokens

Explores whether repeated interactions with a physical object can provide richer identity and provenance signals than words alone.

Presented by NOVEL

Trustless TEE and DevProof

Explores how machine-checked proofs can make hardware security assumptions explicit and reject designs that violate them.

Presented by Quintus Kilbourn

πCreds: Privately Inferred Credentials

Infers privacy-preserving credentials from private, provenance-bound data inside a TEE, extending credentials beyond simple database predicates.

Presented by Sam Breckenridge

Pramaana

Derives pseudonymous credentials from identity data through hardened commitments and a document-to-face enrollment check.

Presented by Vaishnavi Dasika

NDAI and Conditional Recall

Lets agents reason over private context inside a TEE and reveal only outcomes or memories allowed by explicit conditions.

Presented by Quintus Kilbourn

DealProof

Negotiate data deals between agents over sealed evidence inside an TEE, with human approval before the result is finalized.

Presented by Kristian Koci

Crossroads and Infinite Drip

Crossroads provides policy-controlled transfers across chains; Infinite Drip uses that infrastructure to exchange CPU mining work for tokens.

Presented by James Austgen

Aomi

Turns natural-language intent into policy-gated blockchain actions, with tools for simulating and signing across EVM and SVM chains.

Presented by Cecilia Zhang

Daedalus Anchor

Maps portfolio risks to tradable event contracts, then explores custom underwriting where public markets cannot provide enough coverage.

Presented by Freya Zhang

WikiGen

Lets researchers ask questions over private biological data, run approved analysis inside a TEE, and release only the agreed result.

Presented by Wiktoria Leks

CoordinationOS

Turns team transcripts into permission-aware memory, linking people, projects, and agents through provenance and human review.

Presented by Shashank

Private LLM Search and Flashnet

Combines encrypted indexes with anonymous transport to protect both query contents and the searcher's network identity.

Presented by dmarz

Vibe Coding Room

Turns spoken ideas into a shared visual project tree that people can shape together before sending selected branches to coding agents.

Presented by Ron Turetzky

Talks & Panels

8 presentations

Panel: Past, Present, and Future of Agentic Coordination

The panel asks how long-running coordination problems change once software agents can communicate, act, and represent people at machine speed.

Panel: Solarpunk / Lunarpunk Product Design

The panel uses Solarpunk and Lunarpunk as contrasting lenses for deciding when AI products should maximize cooperation through disclosure and when they should preserve autonomy through privacy.

Agency Engineering and the O3 Framework

Agency Engineering and the O3 Framework explores how people can reclaim action from centralized platforms by delegating narrowly scoped control of their own logged-in browser sessions through open code and trusted execution environments.

Presented by Andrew Miller

Privacy Habitat and Infrastructure

The talk argues that privacy must be designed as the habitat of a system, including its metadata and external side effects, rather than added as a narrow content-encryption feature.

Presented by Phil Daian

LLM Account Pooling and Token Subsidies

The talk treats premium-model subscriptions as a temporary, heavily subsidized bridge for continuous agent workloads and proposes owned, metered compute as a more durable foundation.

Presented by shad0w

Workspace Design and Working Memory

An agent workspace can function as external working memory, but it compounds useful knowledge only when the user deliberately controls what the agent investigates, postpones, challenges, and stores.

Presented by Joshua Pham

Everything Is a TEE

Trusted execution is available across far more hardware than cloud GPUs, but useful private systems require complete attestation, careful integration, and explicit reasoning about each device's residual trust and side-channel assumptions.

Presented by Andrew Miller

Shape Rotator Study / Storydeck

Storydeck compresses a longitudinal transcript corpus into summaries, metadata, and knowledge-graph triplets so a language model can surface grounded stories and exploratory relationship patterns across a community.

Presented by Brandon