A Modest Proposal for LinkedIn

Observation #1: LinkedIn sucks

LinkedIn sucks. At this point, I doubt anyone needs to be convinced of this obvious fact.

• AI slop content has largely taken over the site. The one-sentence paragraphs, the breathless "insights", and the supposed personal growth from minor setbacks dressed up as Shakespearean tragedies. All of it relentlessly positive, performative, and empty.
• Automated engagement farming is endemic, with tools like Lempod and LinkedIn Helper turning the feed into a wasteland of automated posts and responses.
• Even in its best form the typical LinkedIn post is someone humble-bragging about their recent job promotion, more suited to a two-line message in a group chat than to a public forum.

Indeed, LinkedIn sucks.

So why does it persist in this horribly broken state? In part it's because its active participants continue to feel that being present there is less professionally risky than disappearing. This creates adverse selection in the content base even more, since the people with good job prospects won't bother with the whole farce, leaving the precarious hoi polloi to populate the wasteland.

But LinkedIn doesn't care about this. The company cares about revenue, and that's the real reason things don't change.

The corporate ad economy

Unless you've been living under a rock for the last 20 years, you already know deep in your bones the reason LinkedIn sucks and keeps sucking year after year. It's ads. Of course it's ads.

Microsoft doesn't publish LinkedIn financials as a standalone reporting segment. So we're forced to read between the lines on what they say and what appears on LinkedIn itself. As best I can determine, LinkedIn made $18b in revenue in 2025.

• Of that, a big chunk (maybe 40%) is Talent Solutions, which basically means the enhanced accounts they sell to recruiters so that the latter can run outbound campaigns to search for and message potential candidates. You may think it's unfair to call this advertising but I disagree. In the modern Internet economy, the line between targeted advertising and (paid) messaging has disappeared. It's all part of the dollars-for-eyeballs economy.
• Another big chunk (~40%) is Marketing Solutions and Sales Solutions. This is ads, but for B2B sales. Premium accounts like Sales Navigator, but also InMail credits, Sponsored InMail, and regular LinkedIn ads likely fall in this bucket.
• Finally there is the rest, a mix of LinkedIn Learning, and other premium subscriptions.

Again, I'm forced to guess here so I could be wrong but the general shape is clear: like every other social media site that monetizes their users, LinkedIn lives off ads.

And while ~$19 billion may seem like a lot of money, it's worth noting that Meta and Google both make roughly $240 billion a year in ads, so roughly an order of magnitude more. This is despite:

• LinkedIn has roughly a third the monthly active users as Meta has (1B vs 3B).
• Those users are significantly higher value (generally professionals with high disposable income).
• LinkedIn has the most valuable knowledge graph in the world. They have X-ray vision into every company, who works where, and with whom. Keep this in mind as you start to see where we're going.

So I think we're safe in concluding:

Not only is the experience of using LinkedIn awful, they're also massively under-monetizing the data they have AND screwing up on the ads business they're running.

Observation #2: Hiring sucks

Hiring sucks. At this point, I doubt anyone needs to be convinced of this obvious fact.

The process of applying to jobs (typically online) is relentlessly dehumanizing:

• Building the brag sheet (cough resume, cough shitty GitHub repos).
• Translating normal work into epic high impact achievement theater (turn "talked to three dudes about what to do" into "collaborated cross-functionally to deliver stakeholder-aligned outcomes").
• Jumping through the arbitrary application steps (show us something you're proud of, and also re-enter your resume in this compliant format).
• Writing the cover letter you know no human will ever read.
• Waiting for the response which never comes (by some estimates, over half of job applications are ghosted by the employer).
• When the response comes, it's an email so generic that it barely hints at the metaphysical concept of work at all.
• If you're lucky, a different response comes. One which asks you to complete a "short exercise" that ends up consuming your whole weekend. Unpaid of course.
• If you're even luckier, you're subjected to the interview gauntlet, where the average interviewer has received no training, doesn't even want to be there, and so phones it in by asking the same question they were asked 10 years ago. Questions which in turn were asked by whoever hired them, all the way back to when someone in Victorian England wondered why manhole covers are round.
• All of this, of course, while simultaneously trying to make sure you're a cultural fit. Reading the tea leaves during lunch, trying to discern how hard to laugh at a joke, or whether to make one yourself. Because your whole life, all you've dreamed about is working on CRM software or whatever they do.
• All the time, you're competing with professional interviewees whose every-two-years job hopping for marginally better compensation (a) doesn't seem to hurt them, (b) hides all the other interviewing they're always doing, and which (c) has made them savant-like in their ability to play the hiring game.
• And finally at the end of it all, the offer stage. Where you discover the stated salary range evidently included the possibility they might hire Jesus Christ come again, and your experience puts you at the bottom of the range.
• And please make your decision within the next 48 hours.

But unless you've had a lot of experience on the other side of the table, it may not be obvious how much it sucks to be the employer:

• If you're a recruiter (whether in-house or retained), dealing with the hiring manager who writes the worst, vaguest job description, then complains about the quality of the applicants, and ghosts you when you ask what they mean by "team player".
• If you're a hiring manager, dealing with the recruiter whose incentives aren't remotely aligned with yours. They're getting paid (if retained) or incentivized (if in-house) for filling reqs. Not for finding you a candidate who can actually do the job and will stay with you for years.
• Receiving literally a thousand applications for every one of your postings. What hope do you have of actually reading and responding to all of them?
• Dealing with the manifest unsuitability of 99% of the applicants.
• And the lying, my god the lying. So much lying.
• Dealing with the professional interviewees, always one step ahead by reading your questions on Glassdoor and Levels.fyi.
• And when you're finally excited to make an offer, you find yourself in a battle to the death with all the other companies from whom they've received offers.
• If you even bother, you contact the candidate's references and find out that, as expected, they are indeed the greatest person the referee has ever met. With their only flaws being that they care too much about work, and oh boy would you ever be lucky to get them.
• And finally, most gallingly, paying the retained recruiter 3-4 months' salary for what amounts to convincing someone to send you a resume. And praying that they don't "maintain the relationship" with your new hire too well, so that in 3 years they come calling to pry them out of your company and into another one for that sweet sweet commission.

The problems we need to solve

I think you can see where we're going with this, but it's important to identify precisely which problems we want to fix.

1. Applying only for the few jobs you're most interested and qualified is a dominated strategy. Given that there is basically zero cost in applying to jobs, everyone's incentive is to apply to as many of them as possible. This leads to every job receiving hundreds or thousands of applications, which causes many of the dehumanizing characteristics we described above for both sides.

2. It is very difficult to signal and evaluate skills, experience, and suitability for a role. Words on a page are easily fabricated, and evaluating skills for a cognitively demanding job is very difficult to do in the artificial setting of an interview. Unpaid job tests suffer their own problems: if (a) they're take-home tests then they're easy to cheat, and if (b) they're in-person work trials then there is little incentive for good candidates who value their time to subject themselves to this. Taken together, unless the employer is literally the most high-status, sought-after employer in the field, unpaid job trials adversely select the candidate pool.

3. Having to "re-qualify" one's basic skillset at every employer is a waste of time for the candidate and a waste of resources for the employer. The first round (or first two rounds) interview for every employer covers basically the same basic skillset ("can you write code" for a programmer, for example). This is a huge waste of resources. It's the problem that TripleByte tried to solve. Unfortunately, their incentives were still largely those of a retained recruiter, so the model started to fall apart. Then at the end, they destroyed their reputation by changing their candidate data policies in an effort to find revenue.

4. Retained recruiters often extract rents far in excess of their economic value. It's true that recruiters do something: sourcing, screening, some amount of persuasion, logistics. Maybe even a little judgment here and there. But extracting a third of an annual salary for shuttling information between candidate and employer, just because they can't find each other, is a massive tax on market opacity. They also have all the incentive to reach out to the new hire after a contractually acceptable time to poach them for another job. Basically, recruiters have the perverse incentive that the worse the hiring market becomes, the more valuable they become!

5. Resumes and job ads are not incentive-compatible. That means that neither side is incentivized to reveal their true preferences and true information. Candidates exaggerate, embellish and yes lie on their resumes because it serves the purpose. Getting that foot in the door, then figuring out how to deal with the effects afterwards. Employers don't want to call the job "Data Cleaning" so they'll tweak the description here and there and retitle it "AI Engineer".

6. The open hiring market doesn't value reputations. When a candidate spams applications, lies on their resume, ghosts interviews and aggressively pits employers against each other in matching/exceeding offers, little of that bad behavior follows them. There are lots of companies out there, and no reputation layer to keep track of everyone. When employers post fake or already-filled jobs, ghost candidates, do open-ended unpaid work trials and low-ball on salary ranges, again there is little cost. Yes Glassdoor and Levels.fyi try, but it's well-known that the business model for those sites is largely reputation-laundering at best, outright extortion at worst.

7. There is no coordination on timing. Good candidates have to do an incredibly difficult and dangerous dance of keeping potential options roughly in step with each other. The worst thing that can happen is to receive an exploding offer from one company when still in the process with a more desirable company. Rejecting the former and then not getting the offer from the latter is such a high risk that many sub-optimal matches happen as a result. Conversely, on the employer side, the narrow incentives to use exploding offers are obvious but the result is hardly Pareto-optimal. Manufactured urgency is a silent cost in the hiring market.

And finally, perhaps most importantly:

8. No one is accountable for match quality. Everyone has a different incentive. Job boards need postings and clicks. Retained recruiters need churn. In-house recruiters need placement metrics. Hiring managers usually want to avoid a truly bad hire more than they want to find a great hire. In an ideal world everyone finds a job they're happy with, and where they're valued, and stay for 5+ years. In the whole cast of characters and incentives we've described so far, who's incentivized for that outcome? No one, as far as I can tell.

This is what we need to fix; LinkedIn is the only one that can do it.

A modest proposal

In what follows, I will describe a complete reimagining of LinkedIn that fixes the problems I've described above. I think that if it's implemented well, it can unlock hundreds of billions of dollars of value around the world, and make LinkedIn incredibly valuable. But first I want to head off an obvious question: what about AI?

Here I will put my cards on the table. I don't believe we're headed for a world where AI replaces large swathes of the workforce. It will likely dramatically change it, but all of the issues with the hiring market remain in nearly all future worlds with AI. Companies aren't going anywhere, and neither are employees.

Finding the consumer surplus

In order to make money doing something better, we need to find where the money is being wasted. Here's what I see:

1. Recruiter/intermediation rents:

Recruit Holdings estimates the following annual spend:

• Job Advertising & Talent Sourcing: $34B
• Direct Hire: $71B
• Retained Search: $24B

Let's say we can eliminate 30% of the first and 50% of the second two. That gets us to $58b.

2. Internal hiring process waste:

The same Recruit document estimates the 2025 Internal Recruitment Automation market at $68B. Recruit defines this as historical employer spending on internal talent acquisition resources that can be automated and monetized, including estimated cost savings from automation. Again we assume we can fix 50% of this: $34b.

3. Vacancy/time-to-fill loss:

150M addressable hires × $45K avg salary × 10 days faster fill × 35% vacancy productivity value = $91b.

4. Mishires/early attrition:

150M hires × $45K × 80% replacement-cost proxy × 12% bad/fragile match rate × 15% reduction = $97b.

5. Candidate time waste:

~150 excess applications per hire × 20 min × $15/hr × 150M hires × 90% reduction = $101b.

This gives us a total of $381 billion dollars of waste we're going to try to eliminate. Most of this comes in the form of hiring faster, reducing bad hires, and eliminating candidate time waste. Interestingly, killing the recruiter industry isn't really a core requirement to get all this surplus. It's just a happy by-product.

Let's say that in 10 years we can capture 25% of these savings. That's almost $100b/year of revenue, roughly 5x what LinkedIn is making now.

Why LinkedIn specifically?

Because they're the ones who have a chance to do it:

1. They're already the closest we have to a global professional identity layer, which is critical to construct the durable reputations on both sides of the market.
2. Size. LinkedIn has ~1B candidates and 70M companies.
3. Existing relationships with employers. They already sell into companies with Recruiter, Jobs, etc.
4. They have the graph, and they're the only ones who have it. Mining this for information is critical to unlock the surplus.

Basically, they're the only ones who could plausibly make both sides of the hiring market accept a new and better protocol.

The core components

1. Proof of personhood

Good news, LinkedIn already has this. Verified profiles will be necessary to apply for jobs and enter the durable reputation economy.

2. Make the Open To Work badge meaningful

Good news, LinkedIn already has a notion of a person looking for a job. But as it's currently implemented, it's actually counterproductive. It's a pathetic signal that you can't get a job the normal way, and are reduced to begging on the streets of the Internet to see if anyone is interested in someone that desperate.

Instead, OTW should be a phase that you have to put in effort to enter:

• You can only enter OTW status once a year, unless a qualifying event (layoff, etc) occurs. This makes OTW scarce.
• In order to enter, you have to describe in detail what you're looking for:
  • Target roles
  • Acceptable levels
  • Minimum compensation
  • Target compensation
  • Location/remote constraints
  • Industries excluded
  • Company size preference
  • Start-date availability
  • Work authorization
  • Interview availability
  • Maximum interview burden
  • Willingness to do work trials
  • Possible referees
• The candidate gets four free tokens to apply for jobs. But they cannot be used for just any job. LinkedIn will block applications which don't match enough for the desired details above.
• After four free tokens, you have to pay to apply. Yes you heard that right. This is what prevents application spamming. And every additional marginal application costs you more: cost = max($25 × n, 0.02% × target comp × n) where n is the incremental application above 4.
• The application cost is refunded if the employer misbehaves (ghosts, role is canceled, changes terms of the job, etc).
• It's not refunded if the candidate ghosts, withdraws without reason, lied about credentials, etc.

Once a candidate enters OTW, it sets up a cadence.

Observe how we've synchronized applications and processes. Companies get a small number of candidates, so they can spend real effort vetting them. Candidates in turn feel like the company actually pays due attention to them instead of being a piece of meat in the grinder. And finally, because everything is synchronized, no more arbitrary offer deadlines and uncertainty about offers.

Of course, nothing prevents someone not in OTW from passively looking. There also could be two levels of OTW, one that's public and one that's private only to the employers the candidate applies to.

And the LinkedIn platform handles everything: scheduling, reminders, tracking, penalties, refunds, token allocation, etc, automatically.

3. Skills certification

LinkedIn can do what TripleByte tried to do, but make it stick. The current badge system is a joke, and should be blown up. Instead LinkedIn becomes the platform where credentials are issued by third parties, and LinkedIn is merely the system of record for those credentials. Of course, the credential providers themselves have reputations, which is what keeps incentives compatible. If a candidate goes for a credential from a sketchy provider, that's fine but the employer will see that credential for the sham that it is.

This is actually a way that universities and other institutions can continue to stay relevant and generate revenue in the future. By becoming credential providers, they can monetize their reputation in a way that actually helps the hiring market. Credential providers can also specify a half-life or decay time appropriate to that particular skill.

For example:

• Candidate has verified Senior Backend Level 4 credential from Stanford.
• Employer’s role spec says Backend L4 accepted from a set of providers (including Stanford).
• LinkedIn guarantees that this bypasses recruiter screen + first technical screen.

4. Employer bond, role spec, and reputation system

In the same way that identity verification allows candidates to enter the durable reputation system, so do employers stake their reputation in order to enter the system. This begins by paying a bond to "post" a job.

The bond is partially forfeited for:

• Ghosting after interview
• Materially changing the comp band
• Exceeding declared interview rounds
• Cancelling the role after meaningful candidate labor
• Exploding offers outside platform rules The bond is refunded or credited when the employer follows process, and repeated bad behavior increases future posting costs.

When they post a job, employers must specify in a canonical machine-readable format:

• Comp band
• Location/remote constraints
• Required skills
• Nice-to-have skills
• Acceptable substitution paths
• Expected interview stages
• Expected decision timeline
• Reporting line
• Team function
• Reason role exists
• Success criteria at 6/12/24 months

This allows the system to automatically match and restrict candidates from applying, again increasing the signal to noise ratio for the employer.

Finally, the system constructs a reputation score for the employer that tracks how they behave:

• Response rate
• Median response time
• Interview burden
• Comp accuracy
• Offer reliability
• Role-cancellation rate
• Ghosting rate
• Retention outcomes
• Candidate satisfaction
• Manager quality
• Post-hire success
• Expected candidate hours consumed per accepted offer

Again, since we have proof of identity, existing employees can anonymously provide feedback to the platform to generate these long-timeline, hard-to-obtain datapoints which are so valuable to understanding the real lived experience of working for a company.

5. Reference system

In order to have accurate durable reputations, we need a system to elicit truthful, accurate references from co-workers. Referees will be paid for their references, in order to make this a worthwhile system in which to participate. But this needs careful design. For example, candidates could theoretically bribe their former boss to provide a quality reference.

Instead, the referee gets paid based on their reputation. First-time referees will get paid only a nominal amount, but their reference will likely be given little weight. High-frequency referees (usually managers and directors) will get paid more for their reference, reflecting the value of their time, but also:

• They provide high quality references with concrete, falsifiable examples.
• Their references correlate with later outcomes.
• They are willing to put reputation at risk. The point is for the reference to be predictive of future performance. Of course, another natural failure mode is that no one says anything negative. This is why references need to stay private, accessible only to the referee and the employer. While this opacity will initially feel awkward to the candidate, the resulting system benefits will become clear over time.

6. Collusion detection

There is a set of natural failure modes when people coordinate to game the system. Positive reference rings, paid fake references, certification cheating, fake job postings to harvest candidates, side-channel applications to avoid fees, employers labeling layoffs as performance failures to avoid fees, and many other possible attacks. However none of these are strong attacks on a system with billions of participants and access to the full connection graph.

Cliques of bad actors can be quickly identified and either penalized or removed. After all, an employer getting blacklisted from the most important hiring platform in the world would be a virtual death sentence. This should serve as a strong deterrent to organized system gaming for employers on the margins.

7. LinkedIn revenue model

LinkedIn charges a placement fee, just like recruiters. But because of the efficiency and high signal, it can charge much less than the current market. Importantly, the placement fee gets escrowed, and vests with employee tenure at the company. For example, the employer agrees to pay 12% of first-year comp.

• 20% vests at start date.
• 20% at 90 days.
• 20% at 6 months.
• 20% at 12 months.
• 20% at 24 months.

This aligns the economic incentives for LinkedIn so that they actually care, for the first time, about match quality. Refunds and forfeitures depend on reason:

Odds and ends

There are a few things that I haven't fully specified yet.

• What's the matching algorithm from candidate desires and skills to approved postings?
• What do incentives look like for passive candidates?
• What do appeal rights look like in the determinations for refunds and payments?
• What are the data retention and privacy considerations of such a massive system?
• How does this all interact with the laws of hundreds of jurisdictions? The reference system in particular will need refinement and iteration to avoid employment law issues, especially as it relates to protected classes.
• How does the system work for lower-status candidates? Do they need a bigger helping hand?
• The same question applies to lower-status companies. How do they get surfaced and shown to good candidates?
• How do those employers, who are on the hook for the bond, receive guarantees that the candidate pool they receive is appropriate to the positions for which they're hiring?

There is also the broad question about whether the mechanisms described here are indeed incentive compatible. It's likely a lot of parameters will need to be tweaked and iterated.

So how do we get there?

What I've described above is the vision, the end state. But transitioning from where we are now to that world will make or break this idea. In the words of the old SNL sketch about driving directions in New England, maybe it's a case of "you can't get there from here".

That's why the system will transition gradually. It seems best to start with high-salary, high-signal, non-executive roles in:

• Software engineering
• Product management
• Sales leadership
• Finance/accounting
• Healthcare and related services

The goal is to find pockets where recruiter rents, candidate application costs, and mishire costs are all high. Over time, the system can expand downwards and outwards to encompass the whole graph. And as it expands, the network effects become more powerful. I expect that it will be hard going initially, but once the superlinear benefits start to kick in, the transition could be relatively quick.

Conclusion

In Swiftian tradition, this isn't a modest proposal. It's explicitly trying to blow up a broken hiring system by blowing up an even more broken social network. This project isn't for the faint of heart. But I want to see someone try it because the benefits to the world would be enormous, and it could make the people who figure it out rather rich.

So what do you say Satya? Want to give it a shot?