June 3, 2026
Office gossip, now with AI
Launch HN: Hyper (YC P26) – Company brain to power agentic development
Startup says it can remember your whole company — and commenters want receipts
TLDR: Hyper is pitching a tool that remembers what your company knows so AI assistants can do smarter work. Commenters were polite but immediately zeroed in on the trust problem: if the system sees conflicting info, can it really tell what’s true, or will humans need to step in?
A fresh Hacker News launch post just dropped, and the pitch is big: Hyper wants to become your company’s all-seeing memory, pulling together scattered notes, chats, emails, and calendars so AI helpers stop acting clueless. In plain English, the founders are saying today’s AI is smart enough to do more work, but still forgets the messy human context hiding in Slack messages, random docs, and hallway decisions. Hyper’s answer? Build one giant shared brain for the company.
But the real action was in the replies, where the mood was a mix of “congrats!” and “okay, but what happens when this thing gets the facts wrong?” The founders got a warm welcome, with multiple users tossing in cheerful launch-day applause, but one early question immediately cut to the anxiety underneath the hype: if different sources disagree, who decides what’s true? That’s the spicy core of the thread so far — not whether the idea sounds cool, but whether anyone wants an AI-powered office memory making a “best guess” about reality. One commenter basically demanded to know if there’s a human referee in the loop before the robot starts rewriting company history.
There wasn’t a full-on flame war yet, but the tension was there: future-of-work excitement vs. trust issues. Also adding a tiny whiff of drama? One comment was straight-up flagged, which in internet terms is basically someone knocking over a drink off-camera while everyone else pretends not to notice.
Key Points
- •Hyper is a startup founded by Shalin and Kanyes that aims to improve AI agents and automations by building a shared, persistent company knowledge system.
- •The founders argue that AI models are increasingly capable, but their effectiveness is limited by fragmented company information and the shortcomings of session-based context retrieval such as MCP.
- •Hyper ingests data from company tools including Docs, Slack, Email, Calendar, and Granola, then converts it into a knowledge graph supported by embeddings for semantic search.
- •Its memory architecture separates raw source items called episodes from extracted facts stored as timestamped subject-predicate-object records with typed relationships such as derivation, tension, and supersession.
- •Hyper includes source provenance, per-user access controls, search that combines embeddings with Postgres full-text search, freshness mechanisms using webhooks or polling, and integrations with tools such as Claude Code, Cowork, Codex, and Cursor.