Altercat

Why we built this

It started as a prompt. A long, detailed prompt that turned a language model into a strategic communication coach. It presented scenarios, waited for a response, then broke down exactly what you protected and what you exposed. What the other party now believes. What could be tighter. A letter grade with no padding.

It worked better than it should have. The scenarios felt real. The feedback was specific. But the real breakthrough was responding by voice. Speaking forces you to commit. You can't revise mid-sentence. You can't pause for three minutes to compose the perfect deflection. You just have to go.

That's the actual skill. Not composing a careful email. Not editing a Slack message before sending. The real test is the room — the investor follow-up you didn't expect, the reporter who asks the question you hoped they wouldn't, the board member who probes a sore spot over cocktails. Those moments don't give you a backspace key.

Every AI communication tool we found trains the wrong thing. They count your filler words. They measure your pace. They coach you on vocal delivery. All useful, none of it strategic. Nobody was training what to say under pressure. What to reveal, what to withhold, how to deflect without looking evasive, how to read a probe for what it actually is. The cognitive game behind the words.

We decided to build the tool that does.

How it works

You pick a training mode. Generic scenarios across difficulty levels if you want to build general reflexes. Preparation for a specific situation if you have something real coming up. Targeted drills if you know your weak spots — maybe you over-explain under pressure, or you fill silence when you should let it sit.

Altercat sets the scene. The players, the stakes, what you know, what you're protecting. Then it puts you in a specific moment and waits.

You respond by speaking. That's the point. Voice is the primary interface because voice is where the skill lives. You can type if you need to, but the training is in the speaking. The discomfort of committing to words in real time, without editing, is the whole exercise.

After each response, Altercat breaks it down. What you protected. What you gave away. What the other party now believes based on your answer. What could be tighter. Alternative phrasings when they'd help. A letter grade. Then it continues the scenario with follow-up pressure or moves to the next one, based on how you handled it.

Difficulty scales with performance. Social navigation at the low end — workplace small talk, casual probes from acquaintances. Professional pressure in the middle — investor questions, client negotiations, internal politics. Public exposure at the top — press conferences, panel Q&As, leaked documents. Compound traps at the hardest level, where every answer has costs and there's no clean move.

What makes it different

The communication coaching space is growing fast. None of it trains what Altercat trains.

Existing tools sit on one side of a divide. They either analyze your delivery mechanics — filler words, pace, tone, eye contact — or they put you through scenario practice focused on familiarity. Have you rehearsed this type of conversation before? Can you handle an objection? Useful skills, but not strategic communication.

Strategic communication is the layer between thinking and speaking. It's information management — controlling what you reveal through both your statements and your questions. It's reading a probe for what it is and responding appropriately without showing you noticed. It's calibrating your openness to match the trust level and stakes of the relationship. It's staying short when the pressure mounts, because length is where people hide when they don't have a clean answer.

That's a trainable skill, but only if the training creates real pressure. Text input doesn't do it. Typing gives you time to think, revise, and polish. The gap between first instinct and final answer stays hidden. Voice closes that gap. It exposes the reflex, which is exactly what needs coaching.

The grading reflects this. Altercat doesn't count syllables per minute or flag hedging language. It evaluates what you said strategically. Did you take the bait? Did you expose a position you meant to protect? Did you create a headline? That's a different scoring system, and it teaches a different lesson with every round.

Where it's headed

Voice as text is the starting point. You speak, Altercat transcribes, and the coaching evaluates your words. But the audio itself carries information that transcription strips out.

Cadence. How long you paused before responding. Whether your pace changed when the question got harder. How many filler words crept in on a topic you're less confident about. Whether your tone shifted when you were deflecting versus when you were being direct. All of that is signal.

Multimodal analysis means Altercat can process the audio as its own scoring artifact. What you said, how you said it, and the gap between the two. Someone who gives a strategically clean answer but takes eight seconds to start and fills the first clause with "um" has told the room something different than someone who delivers the same words cleanly and immediately. Both get coaching, but different coaching.

The goal is a practice environment that's closer to the real room than anything else out there — where the stakes are low enough to fail, the feedback is specific enough to learn, and the skill transfers to the moments that actually matter.