Hacker News
The CEO of Desktop Commander shares his method for detecting candidates using real-time AI assistants during technical interviews, highlighting the tell-tale signs and why this approach is ultimately detrimental to the candidate.
AI 生成摘要
Desktop Commander的執行長分享了他在技術面試中偵測候選人使用即時AI助手的技巧，並指出其明顯跡象以及為何這種做法最終對候選人不利。
Eduard Ruzga is CEO of Desktop Commander, building tools that close the gap between what AI can describe and what it can actually do. His open-source MCP server (5K+ GitHub stars) enables anyone—technical or not—to accomplish system-level tasks through AI. Previously a Staff Engineer at Prezi, Eduard builds tools at the intersection of AI, visual communication, and knowledge management.
And why “I don’t know” is now the best answer you can give me.
At Desktop Commander, we’ve been hiring at our own pace. No rush. No pressure. Just looking for the right people.
And somewhere along the way, I developed a new rule. One I now state upfront at the beginning of every interview:
I give more points to ‘I don’t know’ than to someone clearly reading from a screen.
This usually gets a nervous laugh. But I’m serious.
It’s easy to spot now. And not because their eyes are moving like they’re reading.
Candidates whose answers come a beat too late. Words that sound technically correct but feel… disconnected from the situation. A strange emotional flatness, like the person on the other end isn’t fully there.
They’re not googling. They’re not typing.
They have some kind of AI assistant listening to our conversation and generating answers in real-time.
Here’s the thing most people don’t understand about AI:
Humans have a superpower AI can’t yet match — instant, deep grasp of context from minimal input.
In a real-time conversation, we know what’s actually being asked. We have theory of mind — the ability to step into someone else’s shoes, grasp the gravity of the situation, understand what and why something is being asked. We have skin in the game.
AI doesn’t. Not yet.
In theory, given enough context, AI can simulate this. But it requires feeding it everything: What’s the company? Who are you speaking with? What’s the emotion in the voice? What happened 30 seconds ago in the conversation? Dozens of signals.
That can be done — but not in real-time. Not with today’s tools.
So what happens when someone uses a listening AI to generate interview answers on the fly?
The answers come out generic. Like asking someone to answer a question they didn’t fully hear. Shallow question in, shallow answer out.
Asking AI to listen to a live conversation and provide deeply relevant, personal, contextual answers? I haven’t seen any tool do that well. Not yet.
When I start to feel something is off, I shift my questions.
I stop asking about skills or frameworks. Instead, I ask deeply contextual, personal questions about their work:
These questions require lived experience. Emotional memory. Personal context that no AI assistant has access to.
The AI-assisted answer? It comes back vague. Textbook. Zero emotion.
And here’s the second tell: the person reading the AI’s answer in real-time has no idea what emotion to convey. So there’s none. Their delivery is flat. Disconnected.
It feels like I’m not speaking with a living person.
I asked a question. I got read back a bunch of words that technically form an answer — but aren’t really an answer. As if my question wasn’t truly heard.
Deeply disconnected.
I’ve ended multiple interviews early because there was simply no point continuing.
Let me be clear: I love AI at work.
At Desktop Commander, we’re building tools that let AI do real work on your actual computer — touching files, running commands, automating workflows. I spend my days thinking about how to make AI more useful.
So this isn’t about being anti-AI. It’s about signal vs. noise in hiring.
If you can show me you’re good at using AI — and I can see you doing it, making decisions, steering it, knowing when to trust it and when not to — that’s a skill. A valuable one. Show me that, and I’m impressed.
But if all I see is AI talking through you? Then I have no idea who I’m hiring.
I’m not interviewing the AI. I’m interviewing you. Your judgment. Your experience. Your ability to think on your feet. Your personality.
If you hide behind the assistant, I can’t see any of that.
Here’s what’s ironic: by trying to seem more competent, these candidates reveal less.
A confident “I don’t know, but here’s how I’d figure it out” tells me everything.
A perfect-sounding answer with no soul tells me nothing.
If you’re interviewing soon, here’s my advice:
Show yourself. Not the assistant.
By the way — we’re hiring at Desktop Commander. If this resonated with you (in a good way), check out our careers page.
We're building AI tools that do real work on your computer — files, commands, workflows. See what it can do.
The powerful MCP to manage files, deploy servers, and automate workflows with ease.
© 2026 Desktop Commander MCP. Open-source software under MIT license.
Subscribe now to keep reading and get access to the full archive.
Type your email…
Subscribe
Continue reading