Everyone lists "AI"
on their résumé.
Few actually work AI-native.
The gap between "has used Copilot" and "gets 3× more done with an agent" is enormous — and invisible on paper. Thought Needle analyzes coding-agent trajectories to show you how a candidate actually collaborates with AI, so you can hire for the skill that now matters most.
Talk to usSignal a take-home
can't fake.
A finished diff looks the same whether it took ten minutes or two days. The trajectory — the actual sequence of moves — shows the difference.
How precisely they direct an agent — the quality of instructions, context, and course-corrections that get to a good result fast.
Whether they check the agent's work — testing, reading diffs, catching the confident-but-wrong before it ships.
What they do when the agent goes sideways — do they spiral, or diagnose and redirect with intent?
How much they actually offload versus micromanage — the instinct for what AI should and shouldn't own.
Consented,
structured,
comparable.
Candidates opt in and complete a real coding task with an agent. Harnest captures the trajectory — with PII stripped client-side — and we turn it into a clear read on how they work with AI, comparable across your pipeline.
- 01Candidate opts in and does a realistic agent-assisted task
- 02Harnest captures the trajectory; PII is masked on-device
- 03We surface how they steer, verify, and recover
- 04You get a comparable read — not a black-box score
Candidates consent to being evaluated and know what's captured. We assess how someone works with AI on a real task — not their private data, and not a personality profile. The goal is to give people who are genuinely good with these tools a way to show it.
Hiring engineers for
an AI-native team?
We're partnering with a few teams who are rethinking how they screen for AI fluency. Come build it with us.
hello@thoughtneedle.com