Skip to content
Probe

Your recruiting deserves
better evaluation

Click anywhere to start
CofaProbe

AI-era talent
evaluated in a new way

Don't measure tool fluency — measure the thinking: how someone explains a problem to AI, validates the output, and chooses the right tool. Beyond the deliverable, “how they solved it” is captured automatically, so reviewers can see the whole process at a glance.

Request a free consultTour all of Cofa
Probe candidate workspace — task start, web IDE solving, submit preview, completion

What are you looking at to hire AI talent today?

What makes a good AI hire, what 'using AI well' actually means, how much AI literacy the people evaluating themselves should have — none of it is clear.

Recruiter A

“No AI hiring profile”

Because each company and role needs different AI abilities, there's no agreed-upon, universal definition of an AI hire.

Recruiter B

“No yardstick for 'good with AI'”

Shipping code fast isn't the same as using AI well. You need a measure of what to look for, and how far it should go.

Interviewer C

“I'm a reviewer, but I barely know AI”

If the people evaluating aren't comfortable with AI tools and environments, they can't read the real gap between candidates.

Defining AI talent, setting the bar, automating assessment — all in one module

Probe gives recruiters an AI hiring standard, co-authors the problems with them, and runs an automated evaluation.

Usecase

KRAFTONCofa

The future of hiring evaluation,
built with an AI-first company

Probe is in production as Krafton's standard hiring assessment. From researchers and engineers to non-developer and back-office roles, Probe delivers trusted tests and assessment reports aligned to Krafton's Vision & Value.

AI-native hiring know-how

Krafton's criteria and methods for hiring at the AI frontier — packaged in.

In production, expanding

Krafton itself is rolling AI capability assessment across more and more roles.

Our Goal

Define AI literacy, rewrite what evaluation means

Define AI literacy clearly, then provide differentiated standards and methods to measure it. Make 'good with AI' something you can actually compare.

What is AI literacy?

  1. 01

    Model literacy

    Which models and tools to call, and how to prompt them.

    tool · prompt
  2. 02

    Environment literacy

    Understanding the layer around the model — Memory, MCP — and configuring it for the task.

    memory · MCP · harness
  3. 03

    System literacy

    Beyond one-off calls — wiring AI into the full workflow.

    end-to-end · automation

How it works

Candidates solve; reviewers see the process

Candidates get a smooth task environment; reviewers get the deliverable plus a timeline on one screen. Neither side has to write extra narrative.

  1. Step 1

    Task starts

    Open the browser IDE and start solving. Built-in chat and terminal, plus top models provided free.

    Web IDEBuilt-in AI modelsConnect external tools via MCP
  2. Step 2

    Process auto-captured

    Prompts, tool switches, test runs, and git diffs are all recorded on the timeline. No write-up required from the candidate.

    Prompt logTool switchingTest resultsgit diff
  3. Step 3

    Reviewer

    Deliverable + per-area evaluation and report + reviewer's qualitative notes — provided together on one screen.

    Timeline viewPer-area scoringTags & notes

Evaluation areas

We evaluate, area by area

We map signals an evaluator can read from automatic capture vs. signals that need a human reviewer — separately, per area.

Problem framing

Track decomposition of requirements in the initial prompt and how sub-questions are split.

  • # of decomposition steps
  • Sub-question split pattern
  • Whether constraints/assumptions are stated

Prompt design

Whether constraints and output format are specified, context attached, and system prompt used.

  • Output format specified
  • Context attached
  • System prompt usage

Validation · critical thinking

Follow-up rate, edits to AI output, edits before paste, test re-runs.

  • Follow-up rate
  • AI output edit count
  • Edits before paste
  • Test re-runs

Tool use efficiency

Model choice patterns, tool switch frequency, completeness vs. prompt count.

  • Model choice patterns
  • Tool switch frequency
  • Completeness vs. prompts

Final output

git diff, test pass rate, and the result file — read directly. No automatic scoring.

  • git diff
  • Test pass rate
  • Result file

Timeline integration

Per-area logs in chronological order on one screen. Click an event → jump to the original prompt/diff/log.

  • Event → original jump
  • 1–5 scoring per area
  • Tags & notes

Same starting line for everyone
and we focus on the problem-solving process

During the assessment we provide top Claude, OpenAI, and Gemini models for free. We also place no caps on model power or scope — candidates get the same freedom they'd have in real work.

  1. A

    Candidate A

    · No paid subscription
    ClaudeGPT-5Gemini Pro
    Same environment
  2. B

    Candidate B

    · Cursor Pro user
    ClaudeGPT-5Gemini Pro
    Same environment
  3. C

    Candidate C

    · First time with AI tools
    ClaudeGPT-5Gemini Pro
    Same environment
Same models, same token quota for everyone. Personal subscriptions don't affect evaluation.