Skip to main content
Straight With Me

How this works, and where it gets things wrong

Most products in this category will not tell you how they work. Here is exactly what produces your questions and your feedback — and the honest headline is that no artificial intelligence model is involved today.

Where the questions come from

Each of the eighteen roles has a hand-written set of six questions, chosen for what that job is actually assessed on. They are not generated, and they are not drawn from a generic bank shared across roles.

If you paste a job description, the tool looks for the words that posting repeats most and uses them to reword one of the six, so that question is anchored to the role you are going for. If you type something you are worried about being asked, that replaces another. Everything else is the same set every time.

So: adapted from your job description, not written from it. If you have read otherwise elsewhere on this site, that wording was wrong and is being corrected.

Where the feedback comes from

Pattern rules running in your browser. The tool reads your answer as text and looks for specific, checkable things:

What it checksHow it checks itWhere it goes wrong
Whether you gave a resultLooks for numbers, amounts, durations — or, for roles not measured in numbers, for words that describe how a situation endedIt can miss a result you described in unusual wording
Whether the answer trails offLooks for endings like “and eventually they came around” that stop before the outcomeOccasionally flags a deliberate ending as vague
Whether you hedgeLooks for repeated softeners — “sort of”, “I guess”, “kind of”Some roles talk this way normally and it can over-flag
Length and openingCounts words, and checks how long you take to reach the pointA crisp short answer can be marked too short
Link to the jobChecks whether the terms the posting repeats appear anywhere in your answerYou may have addressed the requirement without using its words

This method is blunt, and it is honest about being blunt. It cannot understand your answer. It notices things that are usually true of weak answers, which is useful for rehearsal and is not the same as judgment.

What is deliberately not here

  • No language model. Nothing you write is sent to OpenAI, Anthropic, Google or anyone else for processing.
  • No score. There is no number, so there is no false precision to defend.
  • No analysis of your voice, face, tone or confidence. Claims made in this area by other products are not well supported, and none are made here.
  • No invented facts. Where a stronger answer would need a number you did not say, you see [ADD YOUR REAL NUMBER] — a visible gap you can fill truthfully, rather than a plausible figure you never earned.

Who processes your data

CompanyWhat they handle
HostingerHosting. Serves these pages and keeps standard web server logs, including your IP address.
Your browser’s speech provider (Google, for Chrome)Only if you choose to speak. Your browser may send the audio to its own speech service to transcribe it. This is your browser’s processing under its own policy, not this site’s. Typing avoids it entirely.
Nobody elseTypefaces and every other asset are served from this domain. Loading a page here makes no request to any other company.

That is the complete list. There is no payment processor, no email platform, no analytics provider and no AI provider connected to this site today, because none of those features exist yet.

If a model is ever added

Saved progress and richer feedback may eventually justify using a language model. If that happens, this page changes before the feature ships, naming the provider and what exactly is sent to it — and the invention problem described above becomes a real risk that will be described honestly here rather than glossed over.

Scroll to Top