What's new in tl, v0.7.7–v0.7.10: deeper discovery, fairer pricing, cleaner reports

Three things got better across last four releases: how deeply your AI agent can search for creators, how much custom data questions cost, and how reports get built and saved. Nothing here asks you to learn anything new — your agent (Claude Code, Gemini, Codex, or OpenCode) reaches for all of it on your behalf.

Deeper discovery: every creator who fits a topic, not just the top few

When you ask for “every travel channel” or “creators who talk about personal finance,” your agent used to hit an invisible ceiling — it could hand back a few thousand of the strongest matches, but the long tail got quietly cut off. Now it can pull the complete ranked list, all the way down to the faintest match, so a target list or a look-alike search goes as deep as our catalogue does.

A few things worth knowing, in plain terms:

• The bigger the list, the more it costs — you’re charged for each creator returned. So it’s worth telling your agent how strong a match you actually care about (“only the solid ones”) before running it wide.
• It searches the creators who are genuinely sponsorable — English-language YouTube channels, music aside, with enough projected views and recent uploads. That’s roughly 46,000 creators today.
• It’s part of the Intelligence plan. If this sounds like your workflow and you’re not on it yet, ask your account manager.

What it unlocks: “Give me every channel with any signal on ‘home renovation’ above a decent match, to seed our spring outreach.” “Start from our ten best fitness sponsorships, find the topic they share, and pull everyone else who fits it.” “How many creators in total touch ‘crypto education’ at all?” The kind of “show me everyone” questions that used to stop short now run all the way to the bottom of the list.

When using the tl tool directly, use these examples:

  tl recommender channels-with-tag "travel" --limit 750 --offset 0    --json
  tl recommender channels-with-tag "cooking" --limit 10000 --json

Note that:

• The “recommender” feature operates on a subset of our channel list we consider “sponsorable” at this time:
  • Non-music YouTube channels
  • With at least 20k projected views per video
  • English language

Fairer, clearer pricing for custom questions

Sometimes a question is too specific for a ready-made answer, and your agent works it out with a custom query against the data. Two of these releases (v0.7.8 and v0.7.9) made what those cost simple and predictable:

• You pay for what you ask for. The price is a straightforward rate per row of data returned — no setup fee, no sliding scale. Tight, focused questions get tight, focused charges.
• Most queries cost 1 credit per record returned. Joining multiple tables increases the rate. Some data, like your sponsorship history, is 10x cheaper.
• You can see the price before you spend a credit. Ask your agent to preview the cost first and it’ll tell you the total before committing anything — handy for setting a guardrail, or for choosing the cheaper of two ways to answer the same question.

The upshot: you can inquire about the budget. “Before you run that, show me what it’ll cost.” “Answer this the cheapest way you can.” Your agent weighs it and picks the lighter route.

Using the tl tool directly:

tl db pg "SELECT COUNT(*) FROM thoughtleaders_channel" --pricing

A cleaner path to building and saving reports

v0.7.10 itself is a tidy-up. Reports are central to how brand teams and account managers use tl, and over time a couple of older, experimental ways of building them had piled up. We retired those, leaving three clear routes — so your agent now has one obvious choice for each kind of ask:

• “Build me a report” — hand over a brief and it assembles the whole thing for you.
• “Save what we just found” — spent a few minutes narrowing down a list of creators, brands, or sponsorships in chat? One line persists it as a report you can reopen later.
• “Add these specific ones to a report” — drop in a list and they land in the report you point at.

Nothing you could do before has gone away, and every report you’ve already saved is untouched. You stay in plain English; your agent picks the route.

Updating

Run tl update to pick all of this up — or just keep working, and the auto-updater will catch you up on your next command. tl changelog always shows the running log of what’s landed.

Happy hunting.