# Jan's Roadmap

I need help to figure out how to organize the team to address business opportunity:
- How to "pathfind" roadmap items
- Clear areas of responsibility and individual ownership (ie who is responsible)

## Business Opportunity

- Jan should build a feature set that allows enterprises and prosumers to "customize" their AI - Open Superintelligence
- Jan Server
- Jan works with cloud and local models, ensuring you are never vendor locked
- Jan allows you to manage a team of users, aggregate and track costs to OpenAI, Claude and other providers
- Smart Routing: route between Cloud/Local models, eg based on PII information
- Running Local Models scalably ("Ollama doesn't scale")
- Running MCPs Servers
- Jan Studio
- Jan allows you to create custom AI Agents (tools to create and eval custom Agents, continually improve them)
- Jan allows enterprises to align open source model (e.g. guardrails, AI safety, model switching)
- Jan allows you to post-train open source models e.g. finetuning, RL (especially for agentic)
- Jan allows you to improve your AI by building evals into the consumer interface
- Jan should focus on being a distribution method for model companies
- i.e. Jan Hub
- Partnerships (e.g. Qwen, Mistral, etc)
- We provide a "shopping experience"
- We quantize and finetune models for Jan compatibility (e.g. LaTeX rendering)
- i.e. Jan Evals
- Evals should focus on providing objective information to enterprises and consumers picking models
- Focus on providing evals for things people actually care about
- e.g. Psychosis, AI Safety, Political Alignment etc
- Competitors
- Short-term: OpenWebUI, Dify, AnythingLLM, Ollama
- Longer-term: ChatGPT Enterprise, Claude Enterprise
- We focus on Agentic AI, vs. AI Workflows
- AI Workflows: n8n, Camunda, Dify
- Agentic AI is the blue ocean
- Over time, we build up Jan as a full self-hostable, open-source Product Suite built over 10 years that replaces multiple AI products
- Similar to Posthog's strategy (and Posthog)
- Provide affordable SaaS version, that can also be self-hosted
- Have better UX and product quality than Microsoft (ultimate last-mover)

| Jan | Alternative |
| ------------ | --------------------- |
| Jan | ChatGPT, Claude |
| Jan Voice | Elevenlabs, Whisprflow |
| Jan Meetings | Otterai |
| Jan Code | Windsurf, Claude Code |

## Key Objectives

- Jan builds Jan Server, Studio, Hub by dogfooding Jan Cloud ourselves
- Build it to meet our own needs (e.g. internal ChatGPT)
- Replace API Keys within Menlo (i.e. wrap OpenRouter, ChatGPT, Claude)
- Collect evals
- MAU target is the Menlo Team
- Build open-source in public, Jan Server should always be in deployable state
- Security, ACL, Team organization
- Reach OpenAI Platform equivalence in 6 months
- Jan Studio: ``/evals`, `/fine_tuning`, etc
- Drop-in replacement for OpenAI as an API
- Scalably run local models (or route to cloud ones)
- Jan Server can be deployed by an Enterprise in 6 months
- "Jan is the best option to deploy AI, and customize it for your enterprise or organization"
- Beat OpenWebUI, Dify
- 1-click deploy on AWS, GCP etc
- Jan has a BD Pipeline
- Start spinning up Enterprise Sales pipeline
- Build Forward Deployment Engineer program (i.e. services business)
- Be able to spin up Agent within 8 hours
- Jan has 70k stars
- Github
- Active community of Connectors, Agents, Extensions?

## Needs Ownership

Jan's Roadmap has a few unclaimed roadmap items that need to be figured out:

### Jiaqi, Norapat, Thinh
- Jan Studio -> Custom Agents
- Prompt
- Model
- Connectors (i.e. MCP)
- Voice Agents
- Eval and Monitoring suite
- Deploy: Slack, Discord, Whatsapp, etc
- Integrations: n8n, etc
- Be able to accomodate Agentic Workflow and Agents
- https://platform.openai.com/docs/guides/agents (see APIs)
- https://platform.openai.com/docs/guides/retrieval#vector-stores
- https://platform.openai.com/docs/guides/embeddings
- "Post-training"
- https://platform.openai.com/docs/api-reference/fine_tuning/create
- Take open source model, and customize it to your needs
- e.g. Agentic (i.e. GRPO, RL training environments)
- e.g. RLHF
- What can be taken from the Jan v1 model pipeline?
- "Alignment"
- Done together with evals?
- https://platform.openai.com/docs/api-reference/moderations/create
- Evals
- Evals-as-a-Service
- Political eval
- Mostly done thru research collaborations with universities
- https://platform.openai.com/docs/api-reference/evals/create
- [ ] Jan Models
- [ ] Small, Agentic Models
- [ ] Voice Models
- [ ] Jan has Voice & Realtime
- Voice models (local and cloud)
- Finetune, customize, add custom words
- Find appropriate API to align with (e.g. Elevenlabs?)
- https://platform.openai.com/docs/api-reference/audio/createSpeech
- https://platform.openai.com/docs/api-reference/realtime

### Emre, Rach

- [ ] Jan User Interviews
- Talking to users
- Emre, Rach?
- [ ] Jan Community Call
- Emre, Rach
- [ ] Jan Website
- Emre
- More or less done
- [ ] Jan Pitch Deck
- Emre and Rach?

### Rach

- [ ] Overall Design of Jan w/ Faisal
- [ ] Jan Browser Extension
- [ ] Jan Model Hub
- Models
- [ ] Jan ?? Hub
- Connectors
- "Agents"?
- Should lead to SEO on Jan Hub
- https://lifearchitect.ai/
- https://artificialanalysis.ai/

### Team (coordinated by Yuuki)

- [ ] Jan User Support
- Team-based concept?
- Duty roster?
- [ ] Jan Docs
- Should be a team sport and driven by Yuuki
- Engineers should document as they go
- Ideally autogenerated in code
- API Reference is very critical
- [ ] BD Pipeline
- Enterprise sales pipeline
- Forward Deployed Engineers -> System Integrator Partnerships
- Services + deploy Jan Server