Announcing "Call Readiness Coach", a skill developed over many hours to be used with Claude, Codex, or Grok, designed to pressure test your preparation for the conversations that actually decide things.
Originally published on Substack.
There is a particular kind of confidence that gets people into trouble, the kind that feels solid until it suddenly collapses beneath you. It is the confidence of having practiced. Either you prepare for an interview, a discovery call, or a client kickoff. You run the story in your head until it sounds smooth, chasing that emotional lift and we have only one chance to make a first good impression. You walk in feeling ready. Then the second question arrives, the one that asks why, and the smooth story comes apart in your hands.
Here is the uncomfortable truth. The preparation felt like work, you spent some time and effort, but most of it was practice of a story you already believed. You were not testing the answer. You were memorising it. And a memorised answer has exactly one good move and no defence, a pattern consistently flagged by experienced interviewers and communication experts.
I kept watching this happen, to people I was helping and to myself, and I did the thing I always end up doing. I stopped treating it as a willpower problem and started treating it as a systems problem.
The conversation is the test, and you might be preparing for the wrong one
When the stakes are real, the conversation is the test. Not the resume, not the deck, not the proposal. The live exchange, where someone who is paying attention pushes on what you said and watches what happens next.
Most preparation practises for a friendlier test than the one you will sit. It runs through the opening, not the question after it. It builds the hero version of the story, the one where everything worked, instead of the honest version, the one with a real failure and a real recovery that a sharp listener will actually believe. It hides behind "we" when the interviewer needs to know what you did. It reaches for words like alignment and transformation when the other side wants a mechanism.
You cannot fix that by trying harder to be honest in the moment. The moment is too late. You fix it by building something that tests the answer before the conversation does.
So I built the test, and made it free
That something is Call Readiness Coach, and today it is public and Open Source. It is a simple, focused prompt system: small, sharp, evidence-disciplined, designed to do one repeatable job well and stay free to use and adapt.
It is one prompt you paste into any AI chat. It doesn't hand you a list of questions to memorise, because that would give you a longer story to train on. It asks one question, reads your answer, names the weak spot out loud, and either advances you or makes you answer the same question more clearly. It runs the same engine in 3 different modes: a job interview, a prospect discovery call, and a client onboarding after the agreement is signed.
The human discipline underneath it is the part I care about. The coach refuses the 4 habits that quietly sink high-stakes conversations. It catches oversharing, the detail that damages you and helps no one. It catches weak ownership, the "we" and "the team" that bury your own work. It catches buzzwords, the polished language that explains nothing. And it catches rambling, the point lost under 200 made up words.
It also refuses the hero story. If you tell it that nothing ever failed, it doesn't believe you, and it makes you find the real correction, the rejected option, the wrong assumption, the near miss. Because a person who has never been wrong is not impressive. They are unprepared for the one question that matters.
Why a coach that argues beats a coach that flatters
There is a principle here that I keep coming back to. You don't get better by talking to something that agrees with you. A single helpful assistant will happily polish your hero story until it gleams, because polishing is the helpful thing to do. That is exactly how you walk into the room overconfident and underprepared. AI will do the same, leaning into sycophancy until reality slaps you in the face.
So this special coach is built to be unimpressed. Its incentive is not to make you feel ready. It is to find the place where your answer breaks, before someone with a real decision finds it for you. Evidence over narrative. Ownership over the collective. The honest unknown over the confident guess. When it doesn't have a fact, it doesn't invent one. It marks it "unknown" and tells you to ask on the call, because a coach that fills your gaps with BS invention is just helping you lie more smoothly.
That last point is the whole philosophy in one line. The goal is not a better performance. It is a defensible explanation of what you actually did, with the gaps acknowledged instead of smoothed over. Like a lawyer presenting a case before a jury, even when the facts are on their side, it is still about how the story is told. You can practise a defensible explanation. You cannot be ambushed out of the truth.
What you actually walk away with
Run the drill, and you don't get a script. What you get is a sourced brief on the other side of the table, a single real claim matched to their actual pressure, your story or plan in plain order, 3 to 5 levels of the real "why", your real failures, the one place you would stall first, and a final readiness pack with the questions it expects you to face. Then you practice against the questions, not against the delusional fantasy.
It is open, it is free, and everything that runs it is public. The prompt, a plain-language guide, complete example transcripts, and even the evals (adversarial evaluations) I used to pressure-test the coach itself are all in my Github repository, and the case study on Systems Detective walks through what it proves and where its limits are. Paste the prompt into a chat, or install it as a skill for Claude or Codex so it is always on hand.
This is not my first open skill, not the last. It is one of a series I have been building and sharing, often live coding late into the night, for the community. As more of them become public they will gather into a proper directory. For now this one stands on its own, and it does one thing I believe in: it makes you earn your confidence before the conversation, instead of discovering in the room that you never had it.
The intent goes further than preparation. It is about helping people find new opportunities, get jobs, win clients, reduce churn, and show up as themselves instead of hiding behind polished outputs. It is about reducing AI slop, which is often just amplified human slop, and strengthening something we are at risk of losing: our own memory of decisions, events, and why we made them. This is the philosophy of extreme ownership applied to conversations. AI can automate repetitive tasks, but judgment is still human work. When reality hits, the illusion breaks. You can prepare perfect briefs and emails with AI, but there is always a moment where you have to show your cards. You cannot bluff your way through a real exchange.
That is why the drill matters. It forces you to understand what you are saying at depth, not just repeat it. There is a reason interviewers like Elon Musk are known for pushing candidates through repeated layers of “why, why, why...” questions, digging until they reach first principles or expose gaps in understanding. The more someone leans on buzzwords, the deeper the questioning goes. It is not about being harsh for the sake of it. It is about separating real understanding from reading a script like a parrot.
AI can make many people sound like experts. This is where that theatre stops. This is where you have the opportunity to regain trust, because you can actually explain, defend, and own what you say. When others cannot go deeper, you can. That is the difference between sounding prepared and being trustworthy.
The conversation is the test. Stop acting. Start drilling.
Otman
Sources
- Google re:Work, "A guide to structured interviewing for better hiring practices."
- Amazon, "Amazon job interview tips from a recruiter."
- Inc., "How to Hire Like Elon Musk."
- Cheng et al., "Sycophantic AI decreases prosocial intentions and promotes dependence," Science, 2026.
- OpenAI, "Why language models hallucinate."
- Anthropic Claude Docs, "Reduce hallucinations."