Undermind

What do they actually do

Undermind is a web app that acts like an AI research assistant for scientific literature. You enter a detailed research question, and it runs an adaptive, multi‑step search across large indexes (e.g., Semantic Scholar metadata) to produce a citation‑backed report with a ranked paper list, short multi‑document summary, categories, a timeline/citation map, and exportable references (CSV/BibTeX/RIS) homepage example report whitepaper.

It uses a successive, agent‑style process that issues multiple searches, follows citation trails, classifies relevance, and estimates how exhaustive the search is. Runs typically take a few minutes and, for much of the corpus, operate on titles/abstracts rather than full publisher PDFs (full text is analyzed where available) whitepaper Katina Magazine review homepage.

They sell a free tier (limited), Pro and Team plans, plus Enterprise with an API for programmatic queries, batch processing, and custom integrations. Current pricing lists Free ($0), Pro ($16/month billed annually), Team ($15/person/month billed annually), and Enterprise by quote pricing enterprise.

Who are their target customer(s)

• Academic researchers (professors, postdocs, grad students): They spend hours or days assembling related work and citations, and need traceable summaries to support claims. They want faster literature sweeps and clean exports for papers and reviews example report homepage testimonials.
• Biotech and medical R&D scientists: They must keep up with fast‑changing, high‑stakes literature; missing a key paper risks wasted experiments or regulatory issues. They need citation‑aware searches and team features for group workflows YC listing enterprise.
• Machine‑learning and data‑science researchers: They need to map methods and citation chains quickly across fast‑moving subfields. They want tools that follow citations, rank relevance, and visualize lineage beyond single keyword queries whitepaper review.
• Corporate R&D teams and product groups: They need recurring, auditable monitoring and the ability to combine internal docs with external literature via batch/API so multiple teammates get consistent updates enterprise.
• Librarians, research managers, and consultants: They produce systematic reviews and track topics across many sources; manual alerts and curation are error‑prone. They want centralized tracking, exports, and notifications, knowing coverage is often abstract‑level today homepage whitepaper.

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

• First 10: Personal outreach to known professors, postdocs, and lab heads; offer extended trials and run one bespoke research sweep per customer to deliver an immediate, usable report and testimonial YC listing example report.
• First 50: Targeted academic and research outreach with live demos/workshops, plus publishing the whitepaper and short case studies (time saved, missed‑paper recovery) to drive inbound trials via departmental lists and researcher communities whitepaper review.
• First 100: Offer department/library pilots and an API/batch pilot to test recurring sweeps and internal+external search; convert successful pilots to Pro/Team or enterprise contracts with a standardized onboarding and ROI playbook, prioritizing medicine, biotech, and ML enterprise whitepaper.

What is the rough total addressable market

Top-down context:

UNESCO estimated ~8.85M full‑time‑equivalent researchers globally in 2018, and researcher density rose to about 1,420 per million by 2022, implying on the order of 10–11M research‑active professionals worldwide UNESCO 2021 Science Report UNESCO/UIS data UIS post.

Bottom-up calculation:

If Undermind sold one seat to each of ~11M research‑active professionals at ~$180/year (Team price), that implies ≈$2.0B annual seat revenue TAM; this excludes additional enterprise/API premiums pricing.

Assumptions:

• Researchers who perform literature reviews are the primary buyers; we proxy with ~11M global researchers based on UNESCO/UIS trends.
• Average annual price per user approximated at Team pricing of ~$180/year; Pro is $192/year pricing.
• Enterprise/API revenue and non‑researcher knowledge workers are excluded for conservatism.

Who are some of their notable competitors

• Elicit (Ought): AI assistant that retrieves and extracts findings from papers for Q&A and data tables. Strong for quick answers and extraction; less focused on multi‑pass, citation‑trail searches and long, exportable reports.
• ResearchRabbit: Discovery and tracking tool that builds visual citation networks and alerts. Emphasizes exploration and staying up‑to‑date over end‑to‑end evidence synthesis into a single report.
• Connected Papers: Generates co‑citation/similarity graphs to find related work and lineage. Useful for graph exploration; does not run adaptive, multi‑step searches or produce long‑form synthesis reports.
• scite.ai: Provides smart citations showing supporting/contradicting contexts. Complements verification and evidence‑checking rather than automating literature synthesis end‑to‑end.
• Consensus: AI answer engine that synthesizes literature into short, cited answers. Optimized for fast responses rather than deeper multi‑pass searches with citation maps and report exports.