About Spint
Honest signal about what you know beats another credential
Spint is a skill check for developers: you answer real questions about a topic, and you find out where your understanding holds, where it thins out, and what to do about it.
- Hand-authored rubrics, facet by facet
- A strong answer, a weak answer and known misconceptions behind every rubric
- Difficulty calibrated on real performance
- 5 scoring dimensions · 3 levels
Origin
Why Spint exists
From the engineer who started it.
I am a software engineer with more than a decade in web development. Since 2025 I have also run over a hundred technical interviews — not as a recruiter, but as an engineer assessing other engineers, session after session. Spint came out of those sessions.
What struck me was what happened when I did write real feedback. It opened people's eyes. They would hear where they were genuinely strong and where they were genuinely thin — often for the first time — and you could watch it land. But getting there took an hour or more per candidate: thinking it through, being fair about it, and constructing it well enough to be worth reading.
And that is exactly the work nobody is paid to do. Once a candidate is rejected, no one has any reason to sit down and explain why — so the explanation never gets written. Almost everyone who has been turned down has had to guess at how they did, and guess at what to fix.
You find out that you did not pass. You almost never find out why.
That gap is expensive. If you cannot tell a clarity problem from a knowledge problem, you practise the wrong thing — and you plateau while feeling like you are working hard.
So I built the thing I kept wishing I could hand people at the end of a session: a way to find out where you stand, judged at the level you are actually aiming at, broken into the dimensions that show what is actually missing — honestly, fairly, and specifically enough to act on.
The rubrics are written by hand, facet by facet. Each one carries a strong answer, a weak answer, and the misconceptions that actually turn up when you ask real people real questions. That is what lets the feedback name what went wrong instead of gesturing at it — and it is the part of Spint that could not be generated.
One thing we want to be straight about: Spint does not make you a better engineer. That part is your work, and nobody can do it for you. What Spint gives you is a fair signal — your real gaps, your real strengths, named — so the effort you put into upskilling lands where it counts instead of where it feels productive. Growth plans are coming to point the direction; walking it stays yours.
We would rather be narrow and right than broad and vague. That is why we are launching with a small number of topics we can genuinely stand behind, and adding more only once the rubrics behind them are worth your time.
Mechanics
What Spint actually does
Four steps, start to finish. No preparation, and nothing to install.
Pick a topic and a level
Choose what you want checked and the level you are aiming at — Junior, Mid or Senior. The level sets the depth expected of you, not just the difficulty of the wording.
Answer an adaptive skill check
Each question is written fresh for you, against a hand-authored rubric for one specific facet of a subtopic, and aimed at where your previous answers left off. A focused sitting, scaled to your level.
Get dimensional scoring and feedback
Every answer is scored 0–10 on five dimensions — correctness, completeness, clarity, depth and practical relevance — with written feedback naming what was strong and what was missing.
See where to work next
Your weakest subtopics and dimensions are named explicitly, so you finish knowing what to go and fix rather than just how you did.
Your team's production build minifies and bundles a Node.js service, and source maps are generated but only kept internally (not shipped to clients). After deploying, an error report shows a stack trace pointing to bundle.js:1:284913 inside a promise chain, with no reference to the original TypeScript source or async call site. Walk through how you would use the available source maps to recover a meaningful trace, and explain why async operations make this reconstruction harder than for purely synchronous code.
Methodology
Inside the engine
If something is going to put a number on your understanding, you are entitled to know how. This is the whole mechanism — how a question is chosen, what each dimension measures, the rules the evaluator is held to, and the arithmetic that turns answers into a score.
How a question gets chosen
A topic breaks into subtopics, and each subtopic into facets — the specific things that can actually be asked about it. A session walks several subtopics, a couple of questions each.
Every facet carries an empirical difficulty, learned from how real users actually performed on it rather than from an author's guess. Selection weights facets by how close their measured difficulty sits to a target for your level, skips facets you have recently seen, and compares question embeddings so you never get a near-duplicate — not within a session, and not across your own history.
Once you demonstrate mastery of a subtopic — a high score held across several answers and more than one facet — it stops being eligible. A session can also end early on that basis, rather than making you grind through what you have already proven.
One consequence worth being upfront about: there is no bank of pre-written questions to pull from. What is hand-authored is the rubric behind each facet; the question itself is written for you at the moment you need it, against that rubric and against how you have been answering. That is why there is a short pause — usually a few seconds — before each question appears. We think a question aimed at you is worth the wait over a canned one served instantly.
The five dimensions, one at a time
Every answer is scored 0–10 on each of these separately. That separation is the point: a single number cannot tell you whether you were wrong, incomplete, or simply hard to follow.
Correctness
- What it measures
- Factual accuracy — outright errors and known misconceptions.
- Why it is scored separately
- It is the only dimension where being confident and wrong is worse than saying less.
- What a low score tells you
- You have a knowledge error, not a communication problem. Fix the fact first.
Completeness
- What it measures
- Coverage — which key aspects of the question you addressed, and which you left out.
- Why it is scored separately
- Separating it from correctness distinguishes wrong from partial. Those need very different fixes.
- What a low score tells you
- You knew the material but stopped early. Practise answering fully rather than studying more.
Clarity
- What it measures
- Structure and coherence of the explanation itself.
- Why it is scored separately
- Two people can know the same thing and only one can transfer it. That gap is career-relevant.
- What a low score tells you
- Your understanding may be fine and unreadable. This is the dimension that most affects how senior you appear.
Depth
- What it measures
- Level-appropriate depth, aligned to Bloom's taxonomy for your chosen level.
- Why it is scored separately
- Depth is relative: what is excellent for a junior is shallow for a senior.
- What a low score tells you
- You are answering below your target level's cognitive bar — recalling where analysis was expected.
Practical relevance
- What it measures
- Real-world applicability versus purely theoretical recall.
- Why it is scored separately
- Textbook-correct answers that never touch practice still fail real work.
- What a low score tells you
- You can recite it but not situate it. Build something with it.
The bars above show an example profile to illustrate the shape of a result — they are not averages.
Depth is measured against your level, not against the best answer in the bank
The depth dimension is calibrated to Bloom's taxonomy, and each level is held to a different cognitive bar.
Junior
Remember · Understand
Recall the concept and explain it in your own words.
Not marked down for missing advanced or edge-case detail.
Mid
Apply · Analyze
Use the concept in a concrete situation and break down why it behaves that way.
Not marked down for skipping architectural trade-off analysis.
Senior
Evaluate · Create
Judge between approaches, name the trade-offs, and design a solution around them.
Marked down for oversimplifying rather than for going deep.
The rules your answer is judged under
These are constraints on the evaluator, not aspirations. Each one is enforced on every answer.
Rubric-anchored, not vibes
Each facet carries a hand-written strong answer, a hand-written weak answer, and the misconceptions that actually show up. Your answer is placed against those anchors rather than judged in a vacuum — and every misconception the evaluation confirms costs a point and must be named in your feedback, so you always find out which one it was.
The senior bar is not applied to juniors
Every facet belongs to one difficulty and carries its own pair of hand-written answers — a strong one and a weak one — composed at that level. A junior answer is placed against what a junior should be able to do, never against a senior exemplar, because the senior exemplar belongs to a different facet entirely.
You are only judged on what was asked
The evaluator is explicitly instructed not to penalise you for related best practices the question never elicited. If it was not asked, its absence is not a deduction.
Behavioural signals never touch your score
We record the time you spend on an answer, how often you switch tabs, and how many times you paste. Those numbers are barred from the score and from all five dimensions — at most they add a note to your written feedback. Your score comes from your answers alone.
What we collect, and what we do with it, is set out in full in our Privacy Policy.
From scores to a plan
There is no hidden weighting. Your overall score is the plain mean of all five dimension scores across all of your answers, rounded to the nearest whole number. That is the entire formula.
The report is structured, not free-form prose: a short synthesis of the session, named strengths, named weaknesses, and a handful of prioritised improvement items each carrying its own severity and mapped back to a subtopic. The model is required to find patterns across your answers — “clarity is consistently low”, “strong depth but weak practical relevance” — rather than restate per-question notes.
Growth plans are not switched on yet. They are built and they work, but they stay off until plans and pricing go live — so what follows describes what is coming, not what you get today. Everything above this paragraph is live now.
A growth plan is built from those improvement items. Overlapping areas are merged so nothing gets worked twice, and each plan carries an objective, a priority, an effort estimate with a suggested weekly pace, and typed steps that run learn → practice → reflect, each with a stated reason for being there. Alongside them: key concepts with context, resources from beginner to advanced, practice tasks, and milestones with evidence.
Two guarantees worth stating plainly:
- No invented links. If the generator is not confident a URL exists and is stable, it omits the link rather than fabricating one.
- Acceptance criteria you can actually check. “Understand X” is rejected; “can explain X to a colleague without notes” is what a practice task has to look like.
What a score means
The same five bands you see on your result page.
- 0–2RawLittle of the concept came through.
- 3–4RisingThe shape is there, the substance is thin.
- 5–6SolidDependable working knowledge with visible gaps.
- 7–8SharpStrong command, minor omissions.
- 9–10EliteAt or above the exemplar for the level.
Scope
What Spint is, and what it isn't
A boundary, not an argument — so you know what you are getting before you spend a session on it.
What Spint is
- A skill check: you answer questions about a topic at a chosen level, and get scored, specific feedback on where your understanding holds and where it thins out.
- A diagnostic: every answer is broken down by dimension and mapped back to the subtopic it came from, so the result points at something you can actually go and work on.
What Spint is not
- A real interview, a hiring recommendation or a certification. It is built to measure your skills and show you what to work on, not to serve as proof of employment readiness.
- A coding environment. Spint works through conversational questions about conceptual and applied knowledge, not live coding exercises.
Principles
What we hold ourselves to
Four commitments, each with the mechanism that makes it more than a slogan.
Clarity over hype
Plain language, and arithmetic you can check. If we cannot explain how a number was produced, we do not publish the number.
How it is enforced: Your overall score is the plain mean of the five dimensions across your answers. No hidden weighting.
Evidence-based growth
Feedback has to point at something specific in what you actually wrote, or it is just encouragement.
How it is enforced: A confirmed misconception has to be named in your feedback — the engine cannot deduct for it silently.
Fairness by design
The bar moves with the level you chose, and scope discipline is enforced on the evaluator rather than left to its judgement.
How it is enforced: Each facet belongs to one difficulty and carries its own strong and weak exemplar, written at that level — so a junior is never measured against a senior answer.
Privacy-first
No video, no audio, no biometrics, no third-party sharing of your results. What we do collect, we name.
How it is enforced: Tab switches and pastes are recorded and then barred from scoring — at most they add a note to your feedback.
Roadmap
Where we're going
Ordered by how close it is, and labelled honestly. Shipped means you can use it today.
- Shipped
Adaptive skill checks with per-dimension feedback
The core loop is live: adaptive questioning against hand-authored rubrics, a 0–10 score on each of the five dimensions, and a written report naming your strengths, your weaknesses and the misconceptions it found.
- Building
More topics
We are starting deliberately narrow — a small number of topics we can stand behind — and widening from there, adding both new areas and more depth inside the ones already live.
- Building
AI as an assessable topic
AI and LLM work becomes a subject you can be assessed on, sitting alongside JavaScript, React and the rest. To be unmistakable: this is a topic you get tested on, not an AI feature inside Spint.
- Next
Growth plans and progress analytics
Prioritised study paths generated from your weakest areas, plus a skill radar and per-dimension trends across sessions. Built, and switched off until plans go live.
- Next
Plans and pricing
Pricing is hidden while we are pre-beta. When it opens, the tiers and what each one includes get published here.
- Exploring
More robust retakes
Retakes that aim at the specific subtopics and dimensions you scored lowest on, which the per-subtopic mastery data already makes possible.
Company
Who you are dealing with
Registered details
- Operated by
- Gegham Hakobyan, IE
- Registered address
- Armenia, Yerevan, Margaryan 2nd lane, 2nd b., app. 22
- Serving
- Developers worldwide.
Get in touch
- Support
- support@spint.techWe aim to reply within two business days.
- Media and partnerships
- support@spint.techSame inbox for now — put the topic in the subject line and it gets routed.
Want to see where you actually stand?
Pick a topic, pick a level, and find out — with the reasoning attached.