The ux.md file we use for Sputnik

Who we build for and why they would pay: the personas, the three layers of what customers want to see, and the UX principles behind Sputnik Intelligence.

Since people asked, I am sharing here the ux.md file we use for Sputnik. It is the file that answers, for everyone working on the product: who we build for and why they'd pay. The original is on GitHub.

The problem we solve

Newsletters and podcasts are where humans talk to and influence humans. Named authors and hosts with real audiences, high trust, and skin in the game, while the open web fills up with machine-written content.

But this conversation is invisible to the tools people use:

  • Monitoring tools can't see it. Meltwater, Cision, Brand24, Muck Rack only watch a newsletter or podcast if you already know its URL, so you only find mentions where you already knew to look.
  • Your agents can't reach it (the technical problem). Podcasts are audio; newsletters live in inboxes. Neither has a clean way for an AI agent to search or read it. We turn both into one indexed, structured corpus your agent can query via MCP/API.

Personas

PR agency account lead

Manages 10–50 clients.

  • Priorities: prove coverage ROI to each client; know fast when a pitch lands.
  • Pains: mentions in newsletters/podcasts are invisible to their current tools; assembling coverage reports is manual.
  • Wants: per-client weekly coverage reports with audience size, fast alerts, "who covered the competitor but not my client" pitch lists.

Brand / comms manager (secondary, higher ACV)

Mid-market company, owns the brand's reputation. Slower to close than agencies.

  • Priorities: never be surprised; show share of voice vs. competitors.
  • Pains: finds out about critical coverage from a customer email days later.
  • Wants: crisis alerts within the hour, morning digest of all mentions, quarterly share-of-voice reports.

Sales team at a tech startup

  • Priorities: reach out warm, at the right moment. An alert when a prospect's exec is interviewed on a podcast means outreach with a specific, timely reference.
  • Pains: buying signals are buried in conversations no tool surfaces.
  • Wants: alerts on target-account execs and funding/hiring news, "which of our 200 target accounts got mentioned this week?"

Product team

  • Priorities: know what the trends in their space are, in real time.
  • Pains: by the time a trend shows up in mainstream coverage or analyst reports, it's old. The early, honest signal is in niche newsletters and podcast conversations.
  • Wants: "what are people in [category] talking about this month?" answered from real voices with quotes, theme summaries over their space, alerts on emerging topics, competitor feature and pricing chatter.

Founder / strategy team

Wants competitive intelligence from unguarded, opinionated coverage: competitor product and pricing mentions, category sentiment, competitor execs appearing as podcast guests.

Investor

Wants portfolio-mention alerts and early signal: which companies niche newsletters get excited about before mainstream coverage.

Newsletter writer / podcast host

The people in the corpus: "who cited me this month?", find shows on the same beat for cross-promotion.

What users want to see, in order

Almost every customer's needs come in the same three layers, widening outward:

  1. Themselves. What's being said about us: our brand, our CEO, our products. This is the first thing everyone sets up and the reason they show up.
  2. Direct competitors. What's being said about the companies we're compared to: their products, pricing, execs, coverage we're not getting.
  3. The space. The wide-angle view: what's going on in our market, which topics are rising, who's driving the conversation.

On top of the layers, specific questions the corpus can answer:

  • Placement: "Which podcasts and newsletters influence our market, and where can I place my team as guests?"
  • Ad intelligence: "Where is my competition advertising or sponsoring?"

UX principles

  • Evidence with every claim. A mention is only useful with its source, date, quote/snippet, and audience size. Reports must be forwardable to a skeptical client as-is.
  • Fresh data. The value of a mention decays fast. Alerts should reach the customer within minutes of publication; every surface should show how fresh its data is.
  • Plain language everywhere. Short sentences, no coined jargon, in UI copy, reports, and API field names alike. Every claim should survive a skeptical PR professional reading it.
  • Explain the data. Every surface should say what the user is looking at: what sources are covered, where a number comes from, what it means. Users can only trust and act on data they can contextualize.

Interaction patterns

Agent-first. Typically, the customer connects their AI agent (Claude, an OpenAI agent) to our data via MCP/API. The agent watches, digests, and reports. The dashboard exists but is not the primary interface.