Base, High, Ultra: The Thinking Each Lead Deserves

#You decide how hard Twain works

Until this release, Twain thought equally hard about every lead. The two hundred names pulled off a conference page and the twelve accounts your team spent a week qualifying got the same research and the same writer, because there was one research pass and one writer.

That was wrong in both directions. The scraped list got more thinking than it earned. The qualified list got less than it deserved, because the ceiling was set by what made sense to run two hundred times.

Generation now has three modes: Base, High, and Ultra. Same pipeline, three depths: the mode changes what the research finds, how much reasoning runs on top of it, and how much freedom the writer gets. You choose per lead. The deepest mode goes as far as a written strategy for what to say and why.

#What each mode gets you

Base is real-time research and fast writing. It researches the account and the contact live, collects signals, and reads the relationship: do you have history with this company, or are you walking in cold? That read matters most for campaigns out of HubSpot or Salesforce, where the answer is usually "we've talked before" and the opener needs to know it. You get two tabs, Sequence and Research, and a writer that follows your instructions strictly.

What Base skips: LinkedIn and YouTube activity, absent signals, the thesis, and the strategy. The Absent signals block (what the deeper modes went looking for and couldn't find) still appears on the Research tab, greyed out and marked Available in High and Ultra, so you see what they'd have added instead of wondering whether Twain looked.

High is more signals, plus a thesis. It adds LinkedIn and YouTube activity to the research, and it fills in Absent signals, so you know what not to assume about the lead. It also adds the Thesis tab: why now, why us, the qualification, and the narrative for this specific prospect.

High is where Twain starts making calls you didn't ask for. In the walkthrough, with zero instructions, it reframed the entire angle for a Director of Partnerships at Apollo. It worked out that a native Apollo integration already exists. Pitching an integration to someone who already has one wastes the first line.

Ultra is everything above, plus judgment. It runs the most searches, produces the longest research and the richest thesis, adds the Strategy tab, and puts the strongest writer on the message. It notices what a shallower pass skims: in the walkthrough it caught that a contact holds three part-time advisor roles alongside the title on their profile. The kind of detail that changes what's worth mentioning. It's also the mode you instruct the least. Ultra is built to make the calls itself, and detailed writing instructions mostly get in its way.

The trade-off is time. Research, qualification, reasoning, then writing is a long chain, and Ultra runs all of it. A batch is slow next to the same batch in Base. That's the deal.

#Strategy: deciding what to say

Deciding what to talk about is the hardest problem in outbound. Harder than phrasing, harder than any tone setting. Ultra's Strategy tab is that decision, written down before the message is written, so you can read it, argue with it, and keep it.

For each lead, the strategy covers:

• Level alignment: what the company cares about and is incentivized to do, then the same read at the level of the individual contact.
• Risks: an honest difficulty read. "High-difficulty cold inbox. This person judges outreach professionally." The sort of thing you want to know before the first line.
• Channel preference: where to reach this person, and why there instead of somewhere else.
• Language and tone alignment: how they write, so the message can meet them.
• Social proof: which of your proof points is relevant to this prospect. The strongest logo is often the wrong one.
• Timing and date math: a signal from last week and the same signal from last quarter mean different things.
• What not to write about: relevant context that shouldn't appear in the message, kept for your team instead.
• What to write about, in priority order: what leads, what supports, and what's real but least useful.

Better messages are where you feel it first, but the strategy is an asset on its own. It exports along with the thesis and the research. Into your CRM, where the account carries context it didn't have. Or to the BDR picking up the manual touches, the follow-up call, and the demo. The hardest thinking about a lead stops living inside one generation run.

#What it costs

Per lead, per generation:

• Base: 1 credit.
• High: 2 credits.
• Ultra: 8 credits, plus 2 per message step. A one-step sequence runs 10 credits a lead; three steps, 14.

Base and High are flat at any step count. Only Ultra prices the steps, because every extra message there is another full round of its writer working against the same research.

To fit three prices, the credit split in two: prices are quoted in the smaller credit, and every existing balance doubled at the same time. The same money buys the same work. The practical result: High costs exactly what generating a lead cost before this release, and Base is half that. The depth your campaigns already ran at didn't get more expensive; the shallow option got cheaper.

The price covers the whole generation. Research and sequence used to be two separate runs; under a mode, one run produces the research, the thesis and strategy where the mode includes them, and the messages, at one price. The one exception is sales research: the legacy research-only call still charges 2 credits a lead, and a workflow that adds sales research to a run charges the same 2 on top of the mode's price.

The regeneration rules are short. Re-running a single lead at its current mode is free, Ultra included. Upgrading charges the new mode's full price, and there are no downgrades: you already own the deeper version. Leads generated before modes count as Base here: they predate the thesis and the new signals, so upgrading one re-runs the research fresh at the deeper depth. Bulk regeneration follows the same rules with one exception: leads already on Ultra pay the 2 per step, priced for the message work the batch re-runs.

#The campaign log

Modes make spend a decision, and on most teams it isn't one person making it. Every campaign now keeps a log, so nobody reconstructs what happened from a balance drop. Credits at the top, the full history underneath, and every row naming who did it: you, a teammate, the API, or a workflow acting on its own.

The donut shows where credits actually went: Base, High, and Ultra generations, verified-email lookups, and credits that went to leads someone later deleted, each in its own slice. The average per lead is beside it. Hover a slice or its legend row and the rest dims. If a teammate ran a batch in Ultra, you see who, when, and what it cost.

The timeline records the whole history, not just spend:

• The campaign itself: created, renamed, duplicated, moved to a folder.
• Leads in: imported from a named source (HubSpot, Salesforce, Apollo, Amplemarket, Outreach, a CSV or XLSX file) with duplicates-skipped counts, added manually, through the API, or by a named workflow.
• Leads out: removed manually or dropped for missing emails; exported, naming the destination and the campaign or list on the other side.
• Generation: generated or regenerated, with the mode, whoever ran it, and what it cost. Partial runs name both numbers ("90 of 125 leads generated") and say so when credits ran out. Failed runs are recorded too, and charge nothing.
• Verified emails: how many were found and what the lookup cost.
• Sequence and targeting: personas added, removed, or changed; a framework changed on a step; steps added or removed.
• Workflows: attached, detached, paused, or resumed, by name.

Same-day repeats collapse into one row, so deleting four leads one at a time reads as one line rather than four. A filled dot marks the rows that moved leads or spent credits.

#Picking a mode for a list

The mode is chosen per lead, so the fastest way to calibrate is to stop guessing: take one representative lead and walk it up the ladder. Generate it in Base, upgrade it to High, then to Ultra, and read what each pass added. On a one-step sequence the whole ladder is 13 credits. The gap between Base and High on your data, with your agent, tells you more than any description of the modes can.

As a starting point:

• Base for volume, and for lists where you're still testing whether the segment responds at all.
• High as the default. It's the depth your campaigns already ran at, now with LinkedIn and YouTube activity and a thesis attached.
• Ultra for lists that earned it: the short, qualified, high-value ones where a single reply pays for the whole batch, and where you'd rather hand Twain the decisions than write instructions.

The rule of thumb: spend where a reply is worth the most, and let Base carry the rest.

#The rest of the update

A few other things moved in this release:

• The scorecard that used to live in the top-right corner is now the Research tab, one of four in the lead's report: Sequence, Research, Thesis, and Strategy, unlocking as the mode includes them.
• Research gained a dedicated relationship read, the field CRM-sourced campaigns needed most.
• The sidebar collapses, so a lead's report gets the full width.
• The log is on every campaign, as described above.
• Exporting the finished campaign works the same as before; see Export campaigns into HubSpot for the shape of it.

If you've read the Campaign Brief post, modes are the other half of the same idea: the brief decides what the campaign is trying to do, and the mode decides how much work goes into each lead inside it.

Try one list in all three modes and tell us what you find. Where High stops being enough is the most useful thing you can send us right now.