Preparing rightly
2026-06-02
Preparing rightly.
There is a particular kind of failure that doesn't get talked about honestly.
A candidate comes in well-prepared. They've done the LeetCode. They know the system design principles. They've read about the company. And then the interviewer asks — not a hard question, just a clarifying one: "Can you walk me through your reasoning?" — and something happens. The answer that was perfectly assembled in the candidate's head becomes noise on the way out. Circular. Imprecise. Missing the structure it had thirty seconds ago in silence.
They don't get the offer. They go home and can't figure out why, because they knew the answer.
I've seen this more times than I can count. I've been interviewing candidates for three years as part of my work, on top of the nine years I've spent as an engineer. The gap that costs people the most isn't technical knowledge. It's the translation from knowing to saying — out loud, to a voice, in real time, under mild pressure.
That gap is a voice problem. And the only way to fix a voice problem is to practice with your voice.
The tools I could find
When I started looking for something to use for practice myself, I found three categories.
The first: live copilots. Tools that listen to your real interview through a hidden overlay and feed you AI-generated answers as the interviewer speaks. Several of these have raised tens of millions of dollars. One launched with a campaign that said "cheat on everything" and a Black Mirror–style video. Another pivoted from the phrase entirely after the backlash, but the product is the same underneath.
I understand the appeal. Interviews are high-stakes and often arbitrary. The incentive to level the playing field is real. But I noticed something: the candidates who use live copilots still have to do the job after they get it. The standup, the code review, the design critique, the first time their manager asks them to walk the team through their thinking. There's no overlay for any of that. And the gap — the gap between what they can do with an AI whispering and what they can do on their own — shows up fast.
The second category: flashcard apps. Question banks, recorded video answers, multiple-choice assessments. These are useful the way a textbook is useful. But a textbook doesn't talk back, and neither do these. You can rehearse the right answer silently twenty times and still freeze when someone asks a follow-up.
The third category: peer-to-peer matching. Platforms where you find a stranger online and practice together. This is genuinely good when it works. But "when it works" is the constraint — partner quality is inconsistent, scheduling is hard, and the stranger across the call doesn't have the context of the role you're applying for.
I wanted a fourth thing.
What I'm building
Intervues is a members' club for people who prepare.
Not a copilot. Not a flashcard app. A quiet room — voice-led, tailored to the specific role you want — that you can come back to as many times as you need until the interview feels ordinary rather than frightening.
Here is what happens in a session:
You paste a LinkedIn URL, drop a job description, or pick from curated banks — roles at Indian startups and global companies, built to match what actually gets asked in campus and off-campus interviews for software engineers, product managers, data scientists, designers, and consultants.
Then you talk. A patient voice interviewer asks ten questions — behavioural, role-specific, culture-fit — and follows up where a real interviewer would. The voice is calibrated for Indian English: slower pause budget, lower false-interrupt rate. If you need a moment to think, you get one. If you need the question rephrased, say so.
When you finish, you get a fifty-page report. Per-answer feedback — not "good answer" but exactly where your structure broke and why. A rewritten version of your weakest answer, so you can see concretely what better sounds like. A drill plan for the two answers that need the most work. A log of every point in the conversation where you contradicted something you'd said earlier — because the interviewer would have noticed, and you should know.
One session. Fifty pages. Credits, not a subscription. Buy a pack when you need one; nothing auto-charges.
Why "a club"
The word is deliberate.
A club is a small group of people who agreed on something and held to it. The thing we've agreed on is this: an interview is the first day of the job, not a test to be beaten. The preparation is the work. The interview is the demonstration.
We will not build tools that whisper in your ear during the real thing. We will not help you pass a bar you cannot, on your own, clear. Not because we can't — because we won't.
I chose the word "club" because the internet has plenty of products. It has fewer places where people who care about doing things the long way around can find each other. I'm trying to build one of those places alongside the software.
The founding cohort is five hundred seats. When they're filled, we close the doors until the next cohort. The first five hundred members get the credit rate locked forever — never raised — ten welcome credits, a numbered membership card, and a direct line to me. Not a ticketing system. An email address that reaches one human, read within twenty-four hours, replied to within seventy-two.
I'm asking them, in return, to be honest with me about what works and what doesn't. That's the deal.
Who it's for first
The first cohort is students.
Specifically: students preparing for campus and off-campus placements in India. SDE, product, data, design, consulting roles at Indian startups and global companies that visit campuses. The roles you're actually being screened for this placement season.
India produces approximately 1.5 million engineering graduates a year. By most estimates, roughly half fail interviews primarily on communication rather than technical knowledge. They know the answer. They can't say it.
The second cohort is mid-career engineers. The switch you've been putting off. The onsite next Tuesday. The senior loop where one conversation changes the offer.
Eventually: anyone with an interview. Fellowship panels, MBA admissions, visa interviews, civil services. If a stranger is going to make a decision about you based on how you answer questions in a room, we're building for it. That's a much larger problem than job interviews, and we'll get there — but we'll earn the right to extend by getting the job interview room right first.
Building in public
I'm sharing the real numbers as they come. Waitlist signups. Session quality. What's working in the product. What isn't. I'll post a build log every week until the founding cohort fills and beyond.
Not because transparency is a growth tactic — though it is — but because I think the most honest thing a solo founder can do is show people what it actually looks like to build something carefully, in the open, without pretending that the numbers are better than they are.
If you're preparing for an interview and this sounds like the kind of room you'd sit in, the waitlist is at intervues.club.
And if you know someone who's in a placement season right now, or about to start one — I'd be glad if you told them.
— Ayoush
Indore · June 2026
Intervues is a members' club for honest, voice-led interview practice. Founding cohort: 500 seats, opening July 2026 by invitation. Take a seat →
Questions
Is voice practice better than a mock interview with a friend?
A friend can't run the same role-calibrated interview twice, won't remember what you said at minute four when you contradict yourself at minute forty, and won't hand you a written report afterwards. The voice interviewer does all three, patiently, in Indian English, without anyone needing to schedule a call.
Why is interviewing a voice problem, not just a knowledge problem?
Most candidates who lose offers know the material. The failure happens in the translation — knowing something precisely, then saying it out loud, in real time, while being watched. That translation is a skill, separate from the underlying knowledge, and it only improves with practice on the output side: saying things, not just thinking them.
Does this work for Indian English and non-native English speakers?
Yes — that's specifically what it's built for. The voice interviewer waits longer between your words, doesn't misfire on filler sounds, and doesn't cut you off mid-sentence. Standard ASR systems have 30–50% word-error rates on Indian-accented English. Intervues is calibrated around that gap.
How do I join?
Join the waitlist at intervues.club. The first 500 founding seats open by invitation order in July 2026. Founding members lock in the founder credit rate for life and receive a numbered membership card.
Earned, not gamed.
Join the founding cohort