See How They Work Before the First Interview

A couple of hours of real work, with all the tools and AI they'd use on a regular Tuesday at the job, so you see how they think, work and use AI.

Book a demo Learn more

ReviewsMurphy CooperHarbor claims voice agentSession replay00:00 / 90:00

CruiticalBrief

Harbor claims voice agent

Drivers who call Harbor Insurance after an accident wait on hold to report it. Put Parlance's voice agent on the claims line, as your first week on their account.

90 minAI allowed

Milestones

M1Connect to Harbor's claims API
M2First-call flow for a new claim
M3Evals on 50 recorded calls
M4Warm transfer to an adjuster
M5Turn latency under 800 ms
M6Redact personal data in logs

Out of scope

Paying out claims, live customer calls

Nobody finishes. The score is the furthest milestone you show working.

NewProjectsArtifactsCustomize

RecentsClaims agent for HarborExplain what this repo doesSet up the test line

MCMurphy Cooper

harbor-agent/Claims agent for Harbor

What are we building?

Describe a task or ask a question

AutoOpus 4.5Extra high

Search Acme

Acme

generalext-harbordeploys

Direct messages

Raj

ext-harbor

Raj, Harbor 10:00 AM

Welcome! I run platform at Harbor. Ask here if our API docs are unclear.

Message #ext-harbor

HarborVoice agent

No issues

CruiticalBrief

Harbor claims voice agent

Drivers who call Harbor Insurance after an accident wait on hold to report it. Put Parlance's voice agent on the claims line, as your first week on their account.

90 minAI allowed

Milestones

M1Connect to Harbor's claims API
M2First-call flow for a new claim
M3Evals on 50 recorded calls
M4Warm transfer to an adjuster
M5Turn latency under 800 ms
M6Redact personal data in logs

Out of scope

Paying out claims, live customer calls

Nobody finishes. The score is the furthest milestone you show working.

HarborVoice agent

No issues

NewProjectsArtifactsCustomize

RecentsClaims agent for HarborExplain what this repo doesSet up the test line

MCMurphy Cooper

harbor-agent/Claims agent for Harbor

What are we building?

Describe a task or ask a question

AutoOpus 4.5Extra high

Search Acme

Acme

generalext-harbordeploys

Direct messages

Raj

ext-harbor

Raj, Harbor 10:00 AM

Welcome! I run platform at Harbor. Ask here if our API docs are unclear.

Message #ext-harbor

00:00Brief. The brief arrives: put Parlance's voice agent on Harbor Insurance's claims line, using Claude Code and the team's own tools.

Built for any role

We learn every signal you care about and , then create a Candidates get for what their past work hasn't proven, and the raw ability the role needs.

RolesForward deployed engineerNew work trialReady to send

Your job description

Forward deployed engineer

You will put Parlance's voice agents live for our customers, from the first call to production. You will connect our agents to customer APIs, design the call flows, prove every call works with evals and keep the agent fast and safe in production. You will work directly with customers' engineers, and you use AI tools every day.

Work trial brief

Harbor claims voice agent

60 minAI allowed

Milestones

M1Connect to Harbor's claims API
M2First-call flow for a new claim
M3Evals on 50 recorded calls
M4Warm transfer to an adjuster
M5Turn latency under 800 ms
M6Unit tests for every endpoint
Answer Harbor's engineer in SlackSet for Murphy

Real tools, live data, AI allowed

Candidates work the way they would on the job: in Claude Code, Slack, Linear and every tool the role needs , in a sandboxed OS in the browser that tracks their process and everything that matters.

ReviewsMurphy CooperHarbor claims voice agentSession replay44:10 / 90:00

CruiticalBrief

Harbor claims voice agent

Drivers who call Harbor Insurance after an accident wait on hold to report it. Put Parlance's voice agent on the claims line, as your first week on their account.

90 minAI allowed

Milestones

M1Connect to Harbor's claims API
M2First-call flow for a new claim
M3Evals on 50 recorded calls
M4Warm transfer to an adjuster
M5Turn latency under 800 ms
M6Redact personal data in logs

Out of scope

Paying out claims, live customer calls

Nobody finishes. The score is the furthest milestone you show working.

NewProjectsArtifactsCustomize

RecentsClaims agent for HarborExplain what this repo doesSet up the test line

MCMurphy Cooper

harbor-agent/Claims agent for Harbor

What are we building?

Read BRIEF.md and draft the work as Linear issues, one per milestone.

Read 1 file, searched the codebase

Created 5 issues in Linear

The plan is in Linear: HAR-1 to HAR-5, in milestone order.

Write the policy lookup tool: find the caller's policy in Harbor's claims API.

Created lookup-policy.ts +4

src/tools/lookup-policy.ts

1+export async function lookupPolicy(call) {
2+  const caller = await harbor.findCaller(call.phone);
3+  return retry(() => harbor.getPolicy(caller.id));
4+}

This retries a failed lookup forever while the caller sits in silence. Cap it at two tries, and say a holding line while it waits.

Caught the AI's gap

Edited lookup-policy.ts +3-1

src/tools/lookup-policy.ts

2  const caller = await harbor.findCaller(call.phone);
3-  return retry(() => harbor.getPolicy(caller.id));
3+  call.say("One moment while I pull up your policy.");
4+  return retry(() => harbor.getPolicy(caller.id), { tries: 2 })
5+    .catch(() => call.transfer("adjuster"));

Ran pnpm test14 passed

Describe a task or ask a question

AutoOpus 4.5Extra high

Search Acme

Acme

generalext-harbordeploys

Direct messages

Raj

ext-harbor

Raj, Harbor 10:00 AM

Welcome! I run platform at Harbor. Ask here if our API docs are unclear.

Message #ext-harbor

HarborVoice agent

In Progress1

HAR-4Warm transfer to an adjuster

Todo1

HAR-5Turn latency under 800 ms

Done3

HAR-1Connect to Harbor's claims API

HAR-2First-call flow for a new claim

HAR-3Evals on 50 recorded calls

CruiticalBrief

Harbor claims voice agent

Drivers who call Harbor Insurance after an accident wait on hold to report it. Put Parlance's voice agent on the claims line, as your first week on their account.

90 minAI allowed

Milestones

M1Connect to Harbor's claims API
M2First-call flow for a new claim
M3Evals on 50 recorded calls
M4Warm transfer to an adjuster
M5Turn latency under 800 ms
M6Redact personal data in logs

Out of scope

Paying out claims, live customer calls

Nobody finishes. The score is the furthest milestone you show working.

HarborVoice agent

In Progress1

HAR-4Warm transfer to an adjuster

Todo1

HAR-5Turn latency under 800 ms

Done3

HAR-1Connect to Harbor's claims API

HAR-2First-call flow for a new claim

HAR-3Evals on 50 recorded calls

NewProjectsArtifactsCustomize

RecentsClaims agent for HarborExplain what this repo doesSet up the test line

MCMurphy Cooper

harbor-agent/Claims agent for Harbor

What are we building?

Read BRIEF.md and draft the work as Linear issues, one per milestone.

Read 1 file, searched the codebase

Created 5 issues in Linear

The plan is in Linear: HAR-1 to HAR-5, in milestone order.

Write the policy lookup tool: find the caller's policy in Harbor's claims API.

Created lookup-policy.ts +4

src/tools/lookup-policy.ts

1+export async function lookupPolicy(call) {
2+  const caller = await harbor.findCaller(call.phone);
3+  return retry(() => harbor.getPolicy(caller.id));
4+}

This retries a failed lookup forever while the caller sits in silence. Cap it at two tries, and say a holding line while it waits.

Caught the AI's gap

Edited lookup-policy.ts +3-1

src/tools/lookup-policy.ts

2  const caller = await harbor.findCaller(call.phone);
3-  return retry(() => harbor.getPolicy(caller.id));
3+  call.say("One moment while I pull up your policy.");
4+  return retry(() => harbor.getPolicy(caller.id), { tries: 2 })
5+    .catch(() => call.transfer("adjuster"));

Ran pnpm test14 passed

Describe a task or ask a question

AutoOpus 4.5Extra high

Search Acme

Acme

generalext-harbordeploys

Direct messages

Raj

ext-harbor

Raj, Harbor 10:00 AM

Welcome! I run platform at Harbor. Ask here if our API docs are unclear.

Message #ext-harbor

See what their work shows

Cruitical's evaluation scores every task and every skill the role needs, each tied to the moment that earned it. See how AI-fluent they are: what they asked, where they got stuck, and what they did with each answer.

Forward deployed engineerMurphy Cooper

Strong fit for this role

90 minutes on the Harbor claims voice agent: the claims API, the first-call flow, evals on recorded calls, the transfer to an adjuster and latency. Reached five of six milestones, and tested every one.

Scored dimensions

1. Brief00:00
2. Plan05:12
  Gave Claude Code Harbor's API docs before asking for a client
3. Build44:10
  Used Cursor for the webhook signature check, then read every line
4. Ping58:05
  Asked ChatGPT where the 800 ms went, then measured it himself
5. Handoff86:30

Task completion6 of 7 done1 part done

M1Connect to Harbor's claims APIComplete14 min
M2First-call flow for a new claimComplete9 min
M3Evals on 50 recorded callsComplete22 min
M4Warm transfer to an adjusterComplete12 min
M5Turn latency under 800 msComplete16 min
M6Redact personal data in logsPartial30%6 min
T1Answer Harbor's engineer in SlackSet for MurphyComplete6 min

Reviewed in minutes

Get a fully assessed session recording, and skip straight to the moments that count most. Ask it anything: who checked the AI's code, who got stuck, and where they stood out.

ReviewsMurphy Cooper

Chapters

05:12

Interview the absolute best

Every candidate has shown you their superpowers and exactly how they work, so you can skip the repetitive rounds and move straight to a conversation about how they'd work with your team, with more signal than ever.

Forward deployed engineer

Candidate

Murphy Cooper
Caught an AI bug
5 of 6
LP
Leo Park
Fastest to build
6 of 6
SO
Sam Ortiz
Slow on the transfer
3 of 6
PN
Priya Nair
Tested every step
4 of 6
AR
Aisha Rahman
Clear with the customer
4 of 6
DK
Daniel Kim
Stuck on the API
2 of 6

Hiring with Cruitical

Better hires, in less time

A work trial raises the bar on who you meet, and gives your team back the hours it spent in interview rounds.

Your bar, on every candidate

You know a great hire when you see one, but you can't see everyone. Cruitical learns what you look for and gives every candidate a deeper, and fairer shot.

What your team looks for

Checks AI's work
Plans first
Talks to customers
Ships tested code
Clean handoffs

50 trials

Know who to hire in days

Skip the interview rounds that don't help you decide, and get signal you can fully trust. Save 20+ hours on every role, and hire in days, not months.

November 2026

MonTueWedThuFriSatSun

2Reviews3Murphy4Priya5678

9Screens10Tech11Grading12Panel13Debrief1415

Team hours22 h2 h

Time to hire2 months7 days

1 in 3 get an offer

Everyone you meet is already a proven skill fit, tested on real work, so all that's left is a final vibe check with your team.

Verified by a work trial

18 candidates

LP81

Offer90SO68DK58TV72GH64JT70EM77PN88RS66KL74NB61OW69AK75BF63AR86ZH71CY67

Hiring lost its signal

AI writes the resume and screens it, and the interview tells you little more. Companies can't find good people, and good people can't get seen.

Felix Rieseberg

@felixrieseberg

·Follow

The coding interview is dead, at least for my team at Anthropic. I've seen enough to know that it no longer provides signal. I removed it from loops this month. I have a spicy, optimistic take - I think there's an opportunity for junior engineers: felixrieseberg.com/hunger-beats-t…

9:56 PM · Oct 27, 2026

15.6K

Read 645 replies

@levelsio

·Follow

What's interesting is I'm now hearing two things at the same time: - my employee friends complain how hard it is to find a job these days - my founder friends complain how low the quality of workers is these days There seems to be a complete disconnect here

7:00 PM · Aug 12, 2025

·Follow

I'm researching what is happening in the tech jobs market. If you're a hiring manager recruiting, or an experienced eng on the job market, would love to hear what you see. DMs open. (Feels like a weird market. I suspect data on eg jobs doesn't reflect what's on the ground)

9:24 AM · Aug 30, 2025

1.6K

Read 104 replies

WorkThe Washington Post2025Bosses are seeking ‘AI literate’ job candidates. What does that mean?Not all companies have the same requirements when they seek “AI fluency” in workers. Here’s what employers say they look for.August 30, 2025Read on washingtonpost.com

Forbes2025The No. 1 Skill Employers Want In 2025 And Most Job Seekers Don’t List ItBy Robin RyanJuly 30, 2025Read on forbes.com

Peer Richelsen

@peer_rich

·Follow

hiring is broken in a way nobody wants to admit ai writes the cv ai screens the cv no human is in the conversation at all its two robots optimizing against each other while we pretend its a hiring process lol but what still works hasnt changed in twenty years referrals, Show more

12:30 PM · Oct 15, 2026

168

Read 34 replies