AI-native candidate assessments

Assess how candidates work with AI.
Not without it.

ScreenStack gives candidates a real codebase, AI access, and a production-shaped problem—then shows you how they direct, verify, and improve the work.

Launch a candidate assessment See how it works

candidate-assessment · 45 min

Two checkouts hit at once and one oversells. Fix the race in the state transition and show me the diff.

The read–check–write sequence isn't atomic. Holding a per-order lock across the whole transition:

async function checkout(order) {
+  const release = await lock(order.id)
  const state = await load(order.id)
+  if (state.stock < order.qty) throw new ConflictError()
+  await apply(order.id, state)
-  await charge(order.id)
+  release()
  return state
}

apply() re-reads state, so that's still racy. Keep read, check, and write in one critical section — then run the tests.

8 tests passed · diff +14 −6

ScreenStack Report Assessment complete — candidate screen-c4a9 45 min

Scorecard91

Functional pass40%

Prompt & steering20%

Verification25%

Diff quality15%

2 prompts · 2 test runs · 1 steer · gemini-2.5-flash

View full scorecard

Why hiring teams move to ScreenStack

Hiring signal that matches how software is actually built.

The task looks like the job

Every assessment is a focused, production-shaped task in a working codebase. Candidates use AI to complete it, because that's how software is built today, and the way they use AI becomes part of the signal.

See how they work with AI

The shipped code is only half the story. You also see how carefully they framed the problem, how they steered the model when it drifted, and whether they checked its output before trusting it.

It fits in 45 minutes

The whole assessment takes one focused sitting. No weekend take-homes, and no reason for strong candidates to walk away from your loop.

Evidence over opinion

The scorecard is generated automatically from the work itself: functional results, AI direction, verification habits, and code quality. No senior engineer needed to grade it.

See how it works

Modernize your hiring signal

Evaluate AI-assisted engineering work before you make the hire.

Buy a candidate assessment