Starter
Start here₹4,999one payment · 30 days
Good for a small engineering team running its first debugging assessment.
- 5 candidate starts per 30-day period
- 1 active assessment at a time
- 1 hiring-team member
Buglyst Hiring
Debugging work samples for engineering teams.
Give backend candidates an unfamiliar codebase with a failing check. Review recorded verification, candidate-written explanations and change scope before the technical interview.
From ₹4,999 for 30 days of access. One payment, no automatic renewal. Drafting is free.
Refund issued twice after a webhook retryILLUSTRATIVE SAMPLE
- 00:00Opened README.md
- 01:12Ran visible checksDid not pass
- 03:40Opened src/webhooks/refund-handler.ts
- 10:02Recorded an explanationCandidate-provided
- 16:58Ran visible checksPassed
- 31:07Submitted a solutionPassed private checks
Visible check runs
3
Private checks
Passed
Patch
1 changed file
Start with the work you need to discuss.
Choose a focused bug that reflects the role, keep the conditions consistent, and use the report to plan the next conversation.
Set before inviting
- A failure that resembles the role’s work
- Candidate instructions and allowed resources
- Consistent time, checks and AI policy
- A reviewer who will use the recorded evidence
Review after completion
- Recorded checks and submitted change scope
- The candidate’s exact written explanation
- Private-check results and missing verification
- Questions to discuss in a human interview
The report is based on recorded actions, not personality inference.
Investigation evidence comes from recorded workspace actions or the candidate’s own words. Buglyst does not score intelligence, personality, or an overall “engineering ability.”
ILLUSTRATIVE SAMPLE The data below is illustrative and does not describe a real candidate.
- 00:00Opened README.md
- 01:12Ran visible checksDid not pass
- 03:40Opened src/webhooks/refund-handler.ts
- 05:05Opened src/refunds/service.ts
- 10:02Recorded an explanationCandidate-provided
- 14:55Recorded an explanationCandidate-provided
- 16:58Ran visible checksPassed
- 19:21Opened src/idempotency/keys.ts
- 25:02Ran visible checksPassed
- 25:30Recorded an explanationCandidate-provided
- 31:07Submitted a solutionPassed private checks
Recorded file opens
README.mdopenedsrc/webhooks/refund-handler.tsopenedsrc/refunds/service.tsopenedsrc/idempotency/keys.tsopened
Change scope
- Recorded change
- Changed 1 file.
The report records the number of changed files. It does not display a raw patch diff or infer why a file was opened.
File navigation is recorded in standard assessments. Company-project reports explicitly say when navigation was not captured.
Visible check runs
3
One failed run, then two passing runs
Private checks
Passed
Recorded result for the submitted solution
Submissions
1
One submitted solution passed
Checks run the same way for every candidate. Private checks execute server-side in an isolated sandbox; candidates and reviewers see pass or fail, never their contents.
Candidate-provided context
- 10:02The queue can deliver the webhook more than once. I will check how the handler identifies a refund across deliveries.
- 14:55The idempotency key includes delivery time, so a retry gets a different key. The refund ID stays the same across deliveries.
- 25:30I used the refund ID as the stable key, then reran the visible checks. I would also ask about concurrent deliveries before changing the idempotency store.
Written by the candidate during the session. Buglyst does not infer what a candidate was thinking.
- Policy
- Buglyst assistant allowed
- Recorded assistant requests
- 2
Candidate-provided context
Asked about retry behavior and where to look for the idempotency key.
How suggestions were checked: Checked the handler and reran visible checks after changing the key.
Recorded requests and candidate-disclosed context are separate. AI use is not scored; requests do not establish whether AI caused a solution to pass.
From an open role to a report you can discuss.
Your team sets the assessment up once. Candidates work in the browser. Verification is automatic and identical for everyone.
Create an assessment
Name the role and pick up to three incidents from the verified library.
Your teamChoose debugging incidents
Focused Bugs for a short screen, Codebase Incidents for a deeper one.
Your teamSet instructions, time, and AI policy
Write candidate instructions, set a time limit, and decide whether the Buglyst assistant is on.
Your teamInvite the candidate
Send a private, expiring link. Candidates need no account and nothing to install.
Your teamThe candidate investigates and submits
They read the incident, open files, run checks, patch the code, and submit.
CandidateBuglyst verifies deterministically
Visible and private checks run in an isolated sandbox. Same checks for every candidate.
BuglystReview the evidence-backed report
Timeline, files, patch scope, checks, candidate notes, and AI use, ready for your debrief.
Your team
Choose a supported assessment path.
Library items are authored by Buglyst and ship with visible checks, private checks, and a verified reference fix. 34 items meet that bar today.
Focused Bugs
28 ready
A single failure in a small repository: a pagination boundary, a config lookup, a retry that never gives up. Good for a first-round screen of 20 to 45 minutes.
Codebase Incidents
6 ready
A multi-file service with an incident brief, logs, and several plausible causes, such as a duplicate charge after a timeout or a worker processing a job twice. For deeper backend and platform roles.
Your company project — controlled pilot
Draft a small sanitized Node.js/TypeScript project with your instructions, checks and reference fix. Production company execution is not yet verified. Runner readiness and a reviewed frozen release are required before company candidate runs. Discuss a company-project pilot.
Repository-informed selection — approved pilot
The private GitHub pilot uses repository structure to select library exercises. Candidates do not execute that connected repository. Review the execution and access boundaries.
Where items come from
Items are synthetic and production-shaped, not copied from real companies. Some are also published in Buglyst’s public practice catalogue; the builder marks which, and a private variant is used whenever one exists.
You decide how AI fits the assessment.
Buglyst assistant requests and candidate-disclosed external tools provide context for the interview. Tool use is not scored as ability.
AI allowed Per assessment
When the assistant is configured for an allowed attempt, its requests appear on the report timeline. Before submitting, candidates can describe any other AI tools they used and how they checked the output.
AI not allowed Per assessment
The Buglyst assistant is switched off server-side for the attempt, and the instructions tell the candidate AI tools are not permitted. Buglyst cannot see tools outside the browser tab and does not claim to detect them.
ILLUSTRATIVE SAMPLE
Read the whole report before you talk to anyone.
Recorded steps, changed-file counts, verification, candidate-written explanations, AI-use disclosure and questions for your debrief. It prints cleanly to PDF for the hiring panel.
What Buglyst does with the code and the candidate.
Plain statements you can check against the trust page. Buglyst is in beta and does not claim SOC 2, ISO 27001, or any other external security certification.
Isolated execution
Candidate code runs in a managed sandbox, never on the Buglyst application host. Misconfiguration fails closed.
Hidden checks stay hidden
Private checks and reference fixes stay server-side. Reports show pass or fail, not check contents.
Tenant isolation
Every workspace request is authorised against your organization membership and role on the server. Invite links are random, expiring, and stored only as hashes.
Integrity context
Reports note whether a private variant was used, how often the shared task version has been used, and whether a similar or matching submitted fix exists, so a copied answer is easier to spot. These are signals for a reviewer, never accusations.
In transit
All traffic to buglyst.com is served over HTTPS.
Illustrative values from the sample report.
One payment. 30 days of Hiring access.
No automatic renewal. A new payment starts a fresh period and allowance. Prices in INR.
Growth
₹9,999one payment · 30 days
For a team hiring for a couple of roles with reviewers sharing the work.
- 20 candidate starts per 30-day period
- 5 active assessments at a time
- 3 hiring-team members
- 1 custom generation per 30-day period
- CSV export of results
Scale
₹19,999one payment · 30 days
For an engineering org running assessments across many roles in the same month.
- 60 candidate starts per 30-day period
- 25 active assessments at a time
- 10 hiring-team members
- 5 custom generations per 30-day period
- CSV export of results
Business
₹49,999one payment · 30 days
For high-volume hiring or private-repository requirements that need a review first.
- 200 candidate starts per 30-day period
- 100 active assessments at a time
- 25 hiring-team members
- 15 custom generations per 30-day period
- CSV export of results
- Private GitHub assessment pilot, approval required
- Priority email support
Make the first work sample useful.
Practical guides for selecting backend work, reading recorded evidence and asking a focused interview follow-up.
Ready to see how a candidate actually debugs?
Build the assessment for free. Pay when you are ready to invite someone.