Questions and answers

Everything we’d expect you to ask.

Grounded in our latest product overview. If we’ve missed something, the founder’s inbox is open — see the bottom of this page.

The product

What Basanite is, and what it deliberately isn’t.

What is Basanite, in one sentence?
Basanite is the technical layer of the interview, rebuilt for the AI era. We run a two-round assessment that measures whether a candidate can actually do the work — both in conversation and at a keyboard, alongside an AI agent.
Why do you say the technical interview is broken?
Two breakdowns crystallised over the past 24 months. Coding tests have collapsed into a cheating arms race: capable AI agents and “interview-coder” overlays make take-homes and live-coding screens trivial to pass without exercising the underlying skill. And the capability that doesmatter — engineering effectiveness in an AI-augmented workflow — isn’t measured anywhere. Banning AI from the interview selects for unaided coding while leaving the AI-orchestration skill entirely untested.
What is Basanite *not*?
Basanite addresses the technical layer of the interview only. Psychometric assessment (personality structure, motivational profile) and culture-fit / values-alignment evaluation are explicitly out of scope. They’re different problems with different evidentiary bases, regulatory surfaces, and commercial dynamics — bundling them in would dilute the rigor of each. We may revisit them as separate product lines once the technical layer is at production quality.
How is this different from coding tests like HackerRank or Codility?
Conventional coding tests measure how well a candidate solves isolated puzzles under artificial constraints. Basanite measures how a candidate ships calibrated, complete work in a real codebase, alongside an AI agent — the way the actual job is done. Round 2 deliberately inverts the standard anti-cheating posture: rather than preventing AI use, we require and instrument it.
How is this different from AI interview platforms like Maki or HireVue?
Existing AI interview tools deliver pre-configured question sequences and score the transcript. They don’t adapt follow-ups based on what the candidate actually says, and they have no mechanism for distinguishing genuine capability from interview preparedness. Basanite uses Construct-Templated Adaptive Interviewing: different questions per candidate, identical underlying constructs and scoring rubrics — plus a second round in a real coding environment that no transcript-based tool can replicate.

The two-round assessment

How the interview itself works.

How long does the assessment take?
Round 1 (conversational) typically runs 20–30 minutes. Round 2 (AI Collaboration Workbench) is time-boxed by seniority: 35 minutes for junior, 60 for mid, 90 for senior, with an optional 120-minute extension for architecture-heavy senior roles. Both rounds terminate on signal saturation, not question or task count.
What is Round 1?
A structured conversational assessment. Basanite asks adaptive, CV-grounded questions and follows up on vagueness, gaps, and unsupported claims. It generates signal across the cognitive, judgmental, and tacit-knowledge dimensions — the things that surface through narrative.
What is Round 2?
The AI Collaboration Workbench. We provision the candidate with a sandboxed VS Code environment, a multi-thousand-line role-matched codebase, a real ticket calibrated to their seniority, and their choice of AI coding agent. We instrument keystrokes, agent prompts, git state, and verification behaviour. After the timed session, a 10-minute reflection conversation cross-references what the candidate did against what they understood themselves to be doing.
Which AI coding agent can a candidate use?
The candidate’s choice — Claude Code, Cursor, Copilot, Aider, or a local CLI agent. Basanite is tooling-agnostic. Forcing candidates into a custom UI distorts the signal; we let them work the way they actually work.
What does the codebase look like? Is it a toy?
It’s not a toy. It’s a multi-thousand-line synthetic project calibrated to the target vertical and seniority — a SaaS codebase for backend SaaS roles, an agentic-systems codebase with retrieval and evaluation harnesses for applied-AI roles, a security-engineering codebase with seeded vulnerabilities for security roles. Tickets are written in the style and granularity the candidate would receive on day one, with deliberate under-specification at senior bands so the candidate has to scope and (sometimes) negotiate with a simulated requester.
If AI use is required in Round 2, how do you prevent cheating?
We invert the standard posture. The “did the candidate use AI” cheating vector is gone — we require it and we instrument it. The risks that remain (a third party operating the candidate’s machine, someone else completing the session) are addressed through identity verification at session start, behavioural biometrics sampled across the session and compared against a Round 1 baseline, and a randomised in-session check-in where the candidate is asked mid-session to explain a specific decision they just made. Genuine candidates explain fluently from working memory; substituted operators don’t.
What does Round 2 deliberately *not* measure?
Round 2 is not an algorithmic-puzzle test in disguise. The codebase contains no LeetCode-style problems. Tickets are routine engineering tasks — the kind of work the candidate would do every day in the role. The point isn’t whether the candidate can solve a hard, isolated problem under artificial constraints; it’s whether they can ship calibrated, complete work the way the actual job is done. We also don’t test whether the candidate uses AI “more” or “less” — the target is judicioususe, calibrated to where the agent helps and where it doesn’t.

The eight dimensions

What we score, and why these.

What are the eight dimensions and where do they come from?

Each dimension has a formal construct definition, intellectual provenance, and an empirical reference list. They draw from cognitive science, philosophy of knowledge, behavioural decision theory, organisational psychology, and the emerging human–AI collaboration literature.

  1. Judgment Under Ambiguity — committing to a defensible course of action when information is incomplete (Knight; Tetlock).
  2. Tacit-Knowledge Articulation — surfacing knowledge that lives in practice rather than in text (Polanyi; Nonaka & Takeuchi; Collins).
  3. Intuition Under Data Scarcity — recognition-primed judgment that distinguishes real expertise from vocabulary (Klein; Dreyfus & Dreyfus; Kahneman & Klein).
  4. Psychological Safety & Collective Learning — the conditions under which errors surface early and dissent is voiced (Edmondson; Project Aristotle).
  5. Creative Problem Reframing — recognising when the team is solving the wrong problem (Schön; Dorst).
  6. Ethical Reasoning in Practice — feeling the weight of real tradeoffs and navigating them with integrity (Aristotle’s phronesis; Rest; AI-ethics applied work).
  7. Transformative Learning From Experience — updating prior beliefs in proportion to disconfirming evidence (Flavell; Kolb; Mezirow; Argyris & Schön).
  8. Human–AI Collaboration Intelligence — the calibrated orchestration of AI tooling (Mollick; Dell’Acqua et al.’s “jagged technological frontier”).
Why these dimensions, and not raw coding throughput?
These are the qualities that distinguish high performers in complex, AI-era engineering work — and the ones conventional technical-interview instruments structurally cannot detect. They cannot be retrieved from a knowledge base. They are forged through real experience and legible only to evaluators who know what to look for. As AI handles more execution-layer tasks, the residual human contribution shifts toward judgment, synthesis, and collaborative intelligence; raw throughput is the skill AI is replacing fastest.
How do scores get assigned?
Every dimension is scored against a behaviourally-anchored rating scale. No dimension may receive a score above 3 without a specific verbatim statement from the candidate cited as evidence, drawn either from the Round 1 transcript or the Round 2 reflection conversation, or from observable patterns in the Round 2 trace. Scores are grounded in what was actually said and done, not overall impression.

For hirers

Configuration, calibration, and what the report contains.

How does Basanite calibrate to seniority?
We use a three-band model — junior (≈ L3 / 0–3 yrs), mid (≈ L4–L5 / 3–7 yrs), senior (≈ L6+ / 7+ yrs). The same role at different bands weights the eight dimensions differently and uses a different sandbox library. Sub-band calibration (distinguishing L4 from L5, or Staff from Senior Staff) is deliberately out of scope — that’s a final-round human responsibility, and we don’t claim AI can do it well.
Does Basanite recommend hire / no-hire?
No. We produce evidence; the human interviewer makes the decision. The hirer report is designed as a briefing document for the final human-led interview: dimension-by-dimension scores grounded in candidate quotes, a technical capability map (areas of demonstrated depth vs surface fluency vs blind spots), and a cheating-risk assessment scored independently of capability.
What does the hirer report look like?
A composite document integrating both rounds. Where the rounds agree, the signal is reinforced. Where they disagree — a candidate who articulates strong principles in Round 1 but ships sloppily in Round 2, or vice versa — the disagreement is itself flagged for the human interviewer to probe. The report explicitly identifies cross-round discrepancies and recommends interview directions to resolve them.
Which roles and verticals does Basanite support?
We calibrate for 25+ representative roles across nine verticals: Consumer Internet & SaaS, Cloud Infrastructure & DevOps, AI / ML / Data, Cybersecurity, Fintech & Financial Services Tech, HealthTech & BioTech, Hardware & Semiconductors, Robotics & Autonomous Systems, Gaming & Interactive Media, and Developer Tools & Languages. Each role × seniority band is a distinct calibration profile. The map is a living artifact — new verticals (e.g. quantum software, BCI engineering) and new roles (e.g. agent-platform engineer) are added as their job markets reach the volume threshold at which dedicated calibration is justified.
Roles outside that map?
Sales, marketing, operations, legal, and executive hiring are out of scope at this stage. Bundling non-technical roles into a technical-evaluation product would dilute the rigor of both. Universities, graduate employers, and professional training programmes are in scope — their assessment-centre infrastructure (£500–2,000 per candidate in assessor time, venue, coordination) is the cost surface Basanite displaces most cleanly.
Does Basanite integrate with our ATS?
Yes. We connect to Greenhouse, Lever, Ashby, and 50+ other ATS providers via Merge.dev. Candidates flow into Basanite assessments automatically as they enter a mapped role, and results push back to the candidate’s ATS record as a structured note plus a link to the full report PDF. Recruiters never have to leave their ATS.
How is pricing structured?
Tier-based, with the commercial argument shaped to your pipeline. For high-volume technical recruiters running 30–40+ technical hires per year, the displacement maths is clean: one Basanite deployment can replace the screening and first-round assessment work of one to three full-time recruiters. For SMEs without dedicated talent functions, pricing is structured as infrastructure rather than labour substitution — you get the evaluation sophistication of a much larger company at a fraction of the cost of hiring a Head of Talent. Specific rates are agreed per engagement; reach out via the waitlist to start that conversation.
How rigorous is the underlying methodology?
Round 1 deploys a documented inventory of structural, questioning, depth, consistency, anti-cheating, and scoring techniques — 22 named methods including Narrative Anchoring, Boundary Condition Probing, Counterfactual Pressure, Progressive Excavation, Vagueness Targeting, Honest Failure Elicitation, Predict-Your-Own-Error, Narrative Consistency Tracking, Cognitive Priority Testing, Information Gap Injection, AI Output Signal Detection, Cognitive Load Escalation, Latency Awareness, and Tacit Knowledge Consistency Testing. Round 2 adds six observable sub-dimensions scored from the trace: Delegation Calibration, Prompt Quality and Decomposition, Verification Rigor, Override Judgment, Engineering Taste, and Solution Completeness. Validation work — construct, content, concurrent, and predictive — is ongoing and pre-registered.

For candidates

What the experience is like, and what we do with your data.

Will I be told what is being evaluated?
The methodology is openly documented. The eight dimensions, the design philosophy, this FAQ — all public. What we don’t disclose during the interview itself is which specific question maps to which dimension. That’s a structural opacity choice: if every question were tagged, candidates could optimise their performance toward the score rather than toward the underlying signal. What is being measured is transparent; which question is measuring what is concealed.
Can I use AI?
In Round 1, no — Round 1 is a conversation, not a coding task. In Round 2, yes — required. Bring your tool of choice (Claude Code, Cursor, Copilot, Aider, local agent). We’re testing whether you ship calibrated work with AI in the loop, not whether you can avoid AI.
What if I don’t know something?
Saying “I don’t know” with genuine awareness is treated differently from a confident but hollow answer. Basanite doesn’t penalise candidates for acknowledging uncertainty — the literature on calibrated expert judgment treats appropriate uncertainty as evidence of genuine expertise, not its absence.
Will I get feedback?
Every candidate receives a personal feedback report regardless of outcome. It’s a brief, neutral plain-language summary: what you demonstrated well, areas for development, constructive suggestions. It’s deliberately designed to be useful without being reverse-engineerable — you can’t use it to game a future Basanite assessment.
What about my privacy?
Before any recording starts you’ll see a consent screen explaining what we capture (voice and transcript in Round 1; keystrokes, agent dialogue, git state, and time-on-task in Round 2), where it goes (Anthropic, ElevenLabs, Supabase — listed in full on our sub-processors page), and how long it’s kept (recordings 6 months, transcripts and reports 12 months, then automatically deleted). You can access, export, or erase your data at any time at basanite.co.uk/data-rights. We don’t sell candidate data and we don’t use it to train AI models. Full details in our Privacy Notice.
Is the interview AI? Can I ask for a human to review my result?
Yes — the interview is conducted and scored by AI. Under UK GDPR Article 22 you have the right not to be subject to a decision based solely on automated processing. There’s a tickbox on the consent screen before the interview, and a self-serve form at basanite.co.uk/data-rights, that flags your assessment so the hirer must apply human review before acting on the score.

Still have questions?

Book a 20-minute call with the team.

Book a call