Epicenter

What do they actually do

Epicenter makes small, open-source, local-first apps that all write to the same on-disk “memory” — a folder of plain text files plus a SQLite index — so your data stays portable and under your control. The main product live today is a keyboard-driven transcription tool (“press a shortcut → speak → get text”) that runs locally and saves transcripts into that shared folder for use by other tools you choose or build website GitHub repo releases.

Distribution is via GitHub and community channels (Discord, discussions). The project is very early and community-driven rather than a hosted SaaS; YC lists the company as S25 with a team size of one YC profile website GitHub discussions Discord link.

Who are their target customer(s)

• Journalists and interviewers: Need fast, reliable transcripts after interviews without uploading source audio to a cloud service; want editable files they can keep and reuse across tools repo README releases.
• Podcasters and meeting hosts: Want to capture conversations, attach notes, and search locally instead of juggling siloed apps; prefer one on-disk memory that multiple tools can read/write vision website.
• Privacy-conscious professionals (e.g., lawyers, therapists, executives): Cannot or prefer not to send sensitive audio/text to third-party clouds; need local control, inspectable storage, and optional local models website vision.
• Knowledge workers and researchers: Use notes, transcripts, and an assistant together but lose context when data is locked in one app; want interoperable tools over the same plain-text/SQLite memory vision.
• Developers and tinkerers: Want hackable, open-source tools they can extend and integrate with existing workflows; prefer community-driven distribution (GitHub, Discord) over closed SaaS repo Discord.

How would they acquire their first 10, 50, and 100 customers

• First 10: Personally invite active GitHub contributors and Discord members to install the release, run a short onboarding call, and gather quotes/feedback to publish repo Discord.
• First 50: Publish step-by-step workflows for interviews/podcasts, seed threads in journalism/podcasting forums, host two live demos (journalists, podcasters), and ship easy installers/templates for non-developers releases website.
• First 100: Add polished installers, short video walkthroughs, privacy/security one-pagers, paid onboarding for firms, and case studies; show up at journalism/podcasting events with demos to capture professional users vision YC profile.

What is the rough total addressable market

Top-down context:

Epicenter sits inside the global speech/voice recognition and transcription market, which 2024 estimates place at roughly $8–$16B depending on definition MarketsandMarkets Fortune Business Insights.

Bottom-up calculation:

Using anchored segments with published counts — ~45k U.S. journalists and ~1.32M U.S. lawyers — at 2–5% adoption paying $60–$120/year for installers/support implies roughly ~$1.6M–$8.2M in annual revenue; adding podcasters, researchers, and international users would increase this BLS ABA.

Assumptions:

• Monetization via optional paid installers/support for an otherwise open-source, local-first toolchain.
• Calculation uses only U.S. journalists and lawyers as a conservative base; excludes international users and other creator/research segments.
• Adoption of 2–5% and ARPU of $60–$120/year; excludes larger enterprise/on-prem contracts.

Who are some of their notable competitors

• Descript: Desktop/web editor with transcription and text-like audio editing aimed at podcasters and creators; overlaps on transcription but is closed-source and cloud-integrated rather than local-file–centric Descript features.
• Otter.ai: Cloud service for live and recorded meeting transcription with searchable storage and team features; competes on speed and accuracy but stores data in the cloud Otter features.
• Obsidian: Local-first notes app that keeps Markdown files on disk with a large plugin ecosystem; overlaps on file portability but is focused on note-taking, not a multi-app shared memory plus native transcription flow Obsidian.
• Logseq: Open-source, local-first knowledge graph/outliner storing data in files on disk; similar audience but centered on block/graph notes rather than a unified plain-text + SQLite memory across multiple small apps Logseq.
• OpenAI Whisper (and local Whisper apps): Open-source speech-to-text model widely used for on-device transcription; overlaps at the engine level but is a component, not a bundled UI plus shared on-disk memory ecosystem Whisper repo.