-
-
Ask Exponential anything—generate tasks, explore insights, or plan sprints in one click.
-
Use built-in templates to catch issues early, compare with competitors, or dive into feedback themes instantly.
-
Exponential turns raw complaints like “too slow” or “crashes often” into structured Facts and actionable Insights.
-
Competitive feedback turned into strategic insights—Exponential helps PMs close feature and usability gaps fast.
-
Exponential transforms raw interview quotes into clear Facts and Insights.
-
View and manage PM tasks powered by AI—summarized insights, user pain points, and competitor feedback, all in one place.
Inspiration
Product teams move fast, but PM busy-work hasn’t kept up. Triaging feedback, writing tickets, and chasing status across tools drains time and energy. We built Exponential at the Hackathon | Bolt.new to act as an AI co-pilot that soaks up all that context for you—so humans stay focused on strategy and creativity.
What it does
What It Does
Real-Time Insight Pipeline
• Listens to Slack and Linear in real time and vectors new messages in < 3 sec
• Gemini Flash 1.5 + hierarchical summarizer condense thousand-line threads into usable nuggetsFacts → Insights → Action Items
• The agent turns raw quotes into structured Facts, prioritised Insights, and team-specific Action Items
• Works with one-click templates (e.g., “Catch issues before they become fires”) or fully custom questions (“Are we behind Asana or Dovetail?”)Continuous PM Chat
• Ask: “What’s blocking Sprint 12?” or “Summarise all feedback on Feature X” and get contextual answers on the spotIntegration & Visibility
• Secure Slack + Linear OAuth today; file uploads, Amplitude, HubSpot on the roadmap
• Dashboards show data freshness, permission scope, and model spend
How we built it
| Layer | Tech & Notes |
|---|---|
| Front-End | React + Vite + TS + Tailwind — snappy dev loop |
| Back-End | Supabase Postgres with RLS, Edge Functions, Auth |
| AI Layer | LangChain agents + Gemini Flash 1.5 (summary) |
| Vector Store | pgvector — lightweight and co-located with data |
| Integrations | Slack Events API & OAuth, Linear GraphQL API |
Architectural Highlights
• Edge Functions absorb webhooks and push to the vector store in < 3 sec
• Hierarchical Summariser fits long threads into Gemini’s context window
• Permission-Aware Embeddings respect Supabase RLS at query time
• Token Budgets + Cache keep model costs predictable as usage scales
Challenges we ran into
| Challenge | Solution |
|---|---|
| Real-time indexing without lag | Batched ingestion + async streaming UI updates |
| Long-thread compression | Multi-level summariser (chunk → section → thread digest) |
| Supabase RLS edge-function pain | Workspace-scoped service-role JWTs |
| Cost control as traffic spiked | Per-request token ceilings + vector-cache hits |
Accomplishments that we're proud of
• Turned Slack chatter into a prioritised task — Facts / Insights / Action Items — under 30 sec
• Custom competitor analysis task (“Asana vs Dovetail”) generated on the fly
• Real-time indexing + summarisation demo ran live with zero dropped events
What we learned
• PMs love structured answers (Facts / Insights / Actions) more than raw summaries
• Small, tool-oriented LangChain agents are easier to test than monolithic prompts
• Predictable costs & permissions matter as much as model quality for enterprise buy-in
What's next for Exponential - Your AI PM partner
- New integrations: GitHub Issues, Amplitude, HubSpot, document uploads
- Fine-tuned feedback-specific embeddings for smarter clustering
- “What-if” sprint planner: simulate velocity, scope, and impact before committing
Built With
- auth
- bolt.new
- cursorai
- edge-functions
- gemini-flash-1.5
- github-actions-ci/cd
- langchain
- linear-graphql-api
- node.js
- pgvector)
- react
- slack-events-api-&-oauth
- supabase
- supabase-(postgres-+-row-level-security
- tailwind-css
- typescript
- vite
Log in or sign up for Devpost to join the conversation.