Jan: Personal Intelligence

# Jan: Personal Intelligence > Nothing was your own, except the few cubic centimeters inside your skull. > — George Orwell, 1984 ## Overview Jan is a personal intelligence that is private, context-aware, and proactively helps you get things done. - Jan is designed to be self-hostable and private-by-default - Jan understands you and your big-picture goals - Jan understands is connected to your systems and understands your context - Jan has a "2nd brain", i.e. memory & personal knowledgebase - Jan can take on delegated tasks, and check in with you for approvals - Jan is proactive and takes initiative to identify actionables from noise Jan consists of a self-hosted server, and a consumer web/mobile app. In the long-term we believe most users will interact with Jan like a real person - via messaging apps, voice calls, or email. The Jan team previously built one of the more popular local AI apps with >4 million downloads, and trained two popular open-source agentic models with >100k downloads. ## Problem The biological human brain does not have sufficient attention and agency bandwidth to keep up with the modern world. We try to make progress on a few key important goals (e.g. job, family, health), against a blizzard of: - Planetary-level information firehoses - [Super-Dunbar number](https://en.wikipedia.org/wiki/Dunbar%27s_number) of communication channels - An accelerating world that generates more tasks that we can handle Human today rely on a limited, serial mental triage system and context window to provide agency, focus and persistence to work towards our goals. We are supported by a mishmash of passive productivity tools that fundamentally rely on human initiative to be effective (spoiler: doesn't work). ## Solution & Product Jan is a personal intelligence that is private, context-aware, and proactively helps you get things done. Beneath the hood, Jan is an agent with 5 core components: - Your Profile (i.e. you, your background, your preferences, your goals) - High-level goals - Todo list - Knowledgebase (e.g. notes, articles, videos) - Connectors to your systems (e.g. calendar, email, notes, files) Jan's agent has two key capabilities: - Jan knows about you via its long-term memory - Jan can do tasks for you via its action engine ## Market Opportunity We believe that personal intelligence will grow to be a TAM of approx $38-48b annually. - ~400m users who already pay for, or are realistic future payers for self-paying productivity tools[^1] - Blended ARPU of ~$8-20/mth (estimate driven by task offload value) - Implies $38-48b TAM annually This TAM does not take into account potential longer-term upside from "agent platform" revenue (e.g. marketplace, third-party tools, API usage), nor Bitwarden-style self-hosted and small-team deployments that can layer on SMB/departmental contracts that nudge ARPU higher. ## Why Now? We believe a personal intelligence is now possible: - Agents and connectors: e.g. [MCP ecosystem](https:/modelcontextprotocol.io) - Long-term memory and task execution: e.g. [MemGPT](https://arxiv.org/abs/2310.08560) - Personal knowledgebases & Retrieval: e.g. [Obsidian](https://obsidian.md/), [Pocket](https://getpocket.com/home), [Context Retrieval](https://www.anthropic.com/news/contextual-retrieval) - World model and agentic design patterns: e.g. [JEPA](https://arxiv.org/abs/2403.00504), [Voyager](https://arxiv.org/abs/2305.16291) Furthermore, a personal intelligence can be viable from a price-performance perspective: LLM inference costs have dropped significantly, and [quantization-aware training](https://pytorch.org/docs/stable/quantization.html#quantization-aware-training) enables us to optimize gross margins in the future with Jan's own agentic models. ## Business Model Jan's business model is a consumer subscription model that starts at $5-20/month, and extends up to >$200/month tiers for power users. On unit economics, Jan targets 45%+ gross margins from launch, that can expand to 70%: 1. Data compaction and compression 2. Migrate higher % inference to Jan's own optimized models 3. Optimize our in-house infrastructure spend rather than pay hyperscaler markups. ## Traction & Metrics > TODO: fill in with real numbers after pilot - Traction - Y DAU - Y 7-d Retention, Z 30-d Retention - X waitlist signups - Y active weekly pilot users completing ~N tasks each. - Agent Success Rate - A% of submitted tasks successfully completed - B% fully automously completed - C% partially autonomously completed (requires human-in-the-loop) - Connectors live: - Gmail, Outlook, Notion, Google Drive, Slack - 15 additional connectors in development. ## Go-to-Market Strategy > TODO Jan's near-term TAM is to focus on people who already pay for personal productivity tools, who would pay for Jan to free up their time.  ## Competitive Landscape Jan competes with a variety of AI chatbots, but is differentiated by its clear focus on a personal intelligence (vs. horizontal LLM or Agent products). - Poke, by Interaction Company - ChatGPT, by OpenAI - Claude, by Anthropic - Gemini, by Google ## Technology & Differentiation Jan's core technology is are proprietary long-horizon agentic models, agent infra that is optimized for existing personal productivity systems (e.g. email, messengers, calendars, notes), and data engine, and in-house infrastructure. - Jan v1, Jan v2 agentic models with long-horizon capability, with proprietary eval and data feedback loops - Jan Agent orchestration layer (e.g. long-term goals, planning, workflow, state management) - Jan Agent Infra infrastructure layer (e.g. MCPs, browsers, VMs, Telephony) We believe that [The Model is the Product](https://vintagedata.org/blog/posts/model-is-the-product), and have taken the contrarian strategy of training our own Jan's own agentic models that we believe will be the differentiator in capability and cost competition in the long term. ## Team Jan team previously built: - Jan, a desktop-based local AI app (4 million downloads) - Jan models, an agentic model capable of long-horizon tasks ## Financials & Forecast ### Revenue Projections | Year | Revenue | Paying Customers | Key Assumptions | | ---- | ------- | ---------------- | --------------------- | | 2026 | $XM | 2,000 | Paid beta, $Y/mo ARPU | | 2027 | $10M | 20,000 | $50/mo ARPU | | 2028 | $100M | 84,000 | $A/mo ARPU | ### Gross Margin Path - 55% → 70% as Jan's in-house models get optimized and cheaper to run - Jan is run from our own datacenter, and we do not pay hyperscaler markups ## Funding Ask  ## Risks & Mitigation  # Appendix  [^1]: https://vintagedata.org/blog/posts/model-is-the-product [^2]: https://x.com/MatternJustus/status/1876829656483086601