David Arnoux, Helping GTM Leaders & Founders Grow With GTM x AI | Fractional CxO | Building Linkedin Tools @ humanoidz.ai, recently posted something that made me stop scrolling: "This headline isn’t actually about jobs. It’s about a Nobel prize you haven’t heard of… It’s a story about what happens when signals become free."

He then unpacked a problem that sits at the center of AI, hiring, and the future of work: what happens when all of our signals – resumes, cover letters, even LinkedIn messages – become so easy to generate that they cost almost nothing?

"In 1973, economist Michael Spence published a paper on signaling that won him a Nobel Prize. The core idea is simple. When two parties don’t know each other, they look for costly signals to establish trust."

In this article, I want to build on what David Arnoux explained and explore why his post resonated so widely – and what it means for anyone trying to get hired, build trust online, or use AI responsibly.

The Nobel Prize Idea Behind the Post

Michael Spence’s signaling theory starts from a basic problem: when two parties do not know each other, they do not share the same information. An employer does not know how capable a candidate really is. A buyer does not know how reliable a seller really is. So they look for signals.

As David pointed out, those signals only work if they are expensive in some way. The peacock’s tail is costly to grow and maintain. A university degree costs years of time, tuition, and effort. The effort is not a side effect; it is the mechanism. The difficulty is the signal.

If almost anyone can acquire a signal, that signal stops meaning much. This is why economists draw a sharp line between signaling and "cheap talk." Signals cost something. Cheap talk does not. And once something becomes cheap talk, bad actors can flood the system with it.

From Costly Signals to Cheap Talk in the Age of AI

This is where AI enters the story. As David Arnoux wrote, AI has driven the cost of many traditional signals close to zero.

Resumes used to require time, self-reflection, and editing. Today, a large language model can generate a polished resume in seconds. Cover letters once revealed how someone thought and communicated; now they often reveal which model and prompt template they used.

"Here’s what AI did. It drove the cost of nearly every signal to zero. Resumes used to cost time and thought. Now they cost a prompt. Cover letters used to reveal how someone thinks. Now they reveal which model they used."

On the surface, this seems efficient and inclusive. But the dark side is that when everyone can produce near-perfect written artifacts instantly, those artifacts stop being strong evidence of anything. They start to move from signals toward cheap talk.

AI vs. AI: When No One Is Really Signaling

David also pointed out that it is not just candidates who leaned on AI. Companies did too. Instead of carefully reading applications, many organizations installed AI-powered screeners and filters.

Now you have one AI generating the signals and another AI evaluating them.

"Now you have two AIs talking to each other. One generating signals. One evaluating them. Neither connected to anything real."

In that loop, almost no human judgment touches the process. The models optimize for the patterns they are trained on: keyword matches, formatting, expected phrases. But they do not know whether you can actually do the work, collaborate well, or think clearly under pressure.

The Atlantic comparison to Tinder that David cited is instructive. Dating apps optimized for engagement metrics: swipes, matches, time in app. The system rewarded quick, low-cost gestures instead of deep, high-cost commitments. The result was more activity but less trust, and ultimately less satisfaction.

Job platforms and AI-powered hiring tools risk the same fate: a lot of motion, very little connection.

Why Friction Is a Feature, Not a Bug

The surprising conclusion in David Arnoux’s post is that the cure for this is not more efficiency, but more friction.

"The strange implication: the solution isn’t efficiency. It’s friction. A signal is only credible if it’s expensive enough that bad actors can’t afford to fake it. The friction was never a bug. It was the mechanism."

Friction, in this context, means meaningful cost: time, effort, or personal risk. When we strip friction out of every interaction, we also strip out the very thing that made those interactions trustworthy.

Think about security: CAPTCHAs, two-factor authentication, and identity checks are all forms of friction. They slow us down, but they make it harder for bots and attackers to impersonate real users. In reputational systems, friction plays the same role. It is harder to fake what is expensive to produce.

Signals That Still Work When AI Is Everywhere

So what still works as a signal in a world where AI can write, summarize, and rephrase on demand? David highlighted three examples worth unpacking:

"Referrals work because someone risks their reputation. Handwritten notes work because they take time. Long-form work samples work because they’re hard to fake."

Each of these carries real cost and real risk.

Referrals require someone to put their name on the line. If you fail, they take a reputational hit.
Handwritten notes take focused, non-automatable time. You cannot yet one-click your way to a thoughtful physical message.
Long-form work samples – in-depth case studies, deep articles, code repositories, product demos – take hours or weeks to create. Doing them well means you actually have the underlying skill.

They share something David emphasized: commitment, skin in the game, and the willingness to spend something you cannot get back.

How to Apply This as a Candidate

If you are looking for a job in an AI-saturated market, the implication is clear: do not compete on cheap talk.

Use AI to polish and speed up routine tasks, but invest your real energy in costly signals:

Build a portfolio of real work: articles, analyses, code, designs, or product experiments that show how you think.
Offer to solve a small, relevant problem for a company and share your process.
Ask for referrals from people who have actually seen you work, and make it easy for them to vouch for you honestly.
When appropriate, follow up with a short, thoughtful note that clearly could not have been copy-pasted.

These things do not scale like blasting out 200 AI-written applications. That is exactly why they stand out.

How to Apply This as a Hiring Manager or Founder

On the other side of the table, if you run a team or a company, David Arnoux’s post is a warning against over-automating trust.

AI screening tools can help with volume, but they should not be the primary gatekeepers. Instead, redesign your hiring process around costly, meaningful signals:

Ask for specific work samples or small paid projects that mirror the real work.
Spend time on referrals and backchannel references from people whose judgment you trust.
Use interviews to probe how candidates made past decisions, not just what they claim on their resumes.

This introduces friction into your hiring funnel, but it improves the precision of your decisions and helps serious candidates differentiate themselves from the noise.

The Paradox of Efficiency and Trust

Toward the end of his post, David Arnoux captured the paradox of our current moment:

"As AI makes communication free, the only credible signals will be the ones that remain expensive. Human judgment. Personal risk. Time that can’t be recovered. The paradox of this moment is that efficiency has become the enemy of trust. The things that can’t scale are the things that still mean something."

In our rush to make everything easier, faster, and cheaper, we accidentally erode the mechanisms that let us trust each other. AI accelerates that trend by turning high-effort signals into low-effort outputs.

The opportunity is not to reject AI, but to be intentional about where we accept efficiency and where we protect friction. Use AI to draft, organize, and explore. But reserve the crucial signals – the ones you want others to trust – for work that genuinely costs you something.

In a world flooded with infinite, AI-generated words, the most valuable signals are the ones that still hurt a little to send.

This blog post expands on a viral LinkedIn post by David Arnoux, Helping GTM Leaders & Founders Grow With GTM x AI | Fractional CxO | Building Linkedin Tools @ humanoidz.ai. View the original LinkedIn post →