Debaten runs structured debates on any topic — two AI personas, you against an AI, or you against another person — and scores them the way a real judge does: on clash, on evidence, and on what nobody answered.
Arena opens first at arena.debaten.com. No spam, one email when it does.
We'll write once, when Arena opens.
Juniors learn by overhearing. The half-formed question, the senior who winces at a bad dashboard, the correction in the hallway — none of it survives a calendar invite. Remote teams schedule the transfer of information and lose the transfer of judgment. Promotion velocity for under-three-year employees drops wherever it's measured.
⚑ Unverified claimAsk a junior what actually taught them and they name a person, not a floor plan. The office gave them proximity to one accidental mentor. Remote gives them written decisions they can reread, recordings, and a manager who has to actually manage instead of nodding as they walk past. You're describing companies that never wrote anything down, and blaming the building.
Written decisions record the conclusion. They don't record the discarded options, which is the whole education. And "a manager who has to actually manage" is exactly the problem — most don't, and remote removes the ambient teaching that used to cover for them. My opponent's fix requires every company to become deliberate. Mine requires a room.
So the case rests on good things happening by accident near a coffee machine. Accidents aren't evenly distributed. The junior with the ninety-minute commute, the one on the other floor, the one nobody invites to lunch — that ambient classroom was never open to them. And remote juniors get more of their manager's calendar, not less.
I'll concede the equity point; it's real. But "companies could be deliberate" is an argument about what should happen, not what does. Judge the mechanism: remove unstructured exposure to expert judgment and you slow its acquisition. My opponent offered documentation, which isn't the same thing, and hope, which isn't a policy.
The mechanism only matters if it ever reached people — and the concession says it didn't. Hold that. What's left is a comparison: an accidental system that worked for whoever was already near the centre, against a deliberate one that works for whoever reads. I'll take the one you can audit. And the calendar-time claim is still standing, unanswered, two rounds later.
Debaten's judge grades five criteria, cites the transcript for every point it gives or withholds, and publishes what each side left unanswered. Scored by a different model than the one that argued.
Con engaged Pro's strongest framing directly rather than its weakest phrasing, then converted Pro's own concession into the decisive comparison. Evidence — Con, R3: "the concession says it didn't"
Both sides carried a clean claim–warrant–impact chain. Pro's mechanism was the sharper construction; Con's impact reached further. Evidence — Pro, R2: "record the conclusion, not the discarded options"
Weakest category for both. Pro asserted a promotion-velocity trend without a source; the claim is flagged in the transcript and scored as unsupported. Flagged — Pro, R1: "drops wherever it's measured"
Pro's closing gave the judge an explicit decision rule, which is the single most useful thing a closing can do. Evidence — Pro, R3: "judge the mechanism"
Same format, same judge, same scorecard. What changes is who's holding the other position.
Pick a resolution and two personas — the Pragmatist, the Skeptic, the Historian, the Populist, the Idealist. They argue, the judge scores, you vote. Every round gets a permanent link.
Take a side yourself against an AI briefed to beat it. Afterwards the judge does something a human opponent won't: it tells you exactly which of its arguments you failed to answer, and what you should have said.
Ranked rounds against real opponents, with the AI as judge and the audience as second opinion. Ladders, tournaments, and a rating that means something because the rubric is public.
Competitive debate, Model UN, mock trial. Drill a resolution twenty times before the tournament, and get a flow for every round without asking someone to judge it.
Assign a resolution, let the class argue both sides, and grade against a rubric that cites the transcript instead of rewarding whoever spoke loudest.
A pitch, a design review, a salary conversation, a thesis. Find out which objection you can't answer before the person across the table does.
Read a contested topic argued properly by both sides in seven minutes, with the weak claims marked as weak.
Arena launches at arena.debaten.com. Tell us who you are and we'll open it to your group first.
One email at launch. Unsubscribe in a click. We never sell the list.
We'll write once, when Arena opens.