Don't Make Job Referrals Public | Mohamed Elashri

Mohamed Elashri

A collection of notes about physics, programming and personal thoughts.

🌙

posts

micro

links

عربي

github

rss

Don't Make Job Referrals Public

17 Sep 2026

3 min read

I am currently in the job market, so I receive some recruiter emails. Recently, I started seeing a slightly different kind of message.
I posted in Hacker News' "Who wants to be hired?" thread with a short description of my background: physics, C++, CUDA, GPU programming, machine learning, and some recent work around LLM inference.
A little while later, several different people contacted me about the same Senior Software Engineer position. Their messages mentioned specific parts of my background, but they all led to the same company. Each included a different referral code.
The job ad says a successful referral pays $1,500. The payment only happens if the candidate is hired through that referral, so this is not a complaint about people being paid for every application.
The problem is what that incentive does when the referral program is public.
I am not an obvious fit for the role. I am finishing a PhD and moving toward industry, while the position is explicitly senior. I have some relevant technical overlap, but I would not describe myself as the established senior engineer the title suggests.
It seems unlikely that several unrelated people independently studied my experience and all decided I was an unusually strong candidate for the same senior position. A more likely explanation is that my comment, along with the other posts in the thread, was searched for keywords. I looked vaguely plausible, and sending me a message was cheap compared with the chance of earning $1,500.
The messages are also easy to personalize at scale. An LLM can produce a note that mentions my background in physics, CUDA, or LLM inference without the sender knowing much more about me than those keywords.
I do not particularly blame the people who contacted me. They are responding rationally, from their perspective, to the incentives the company created. But from the candidate's perspective, the result is spam: several unsolicited messages from strangers, each trying to get credit if I eventually get hired.
A traditional referral usually carries some information. Someone knows the candidate, has worked with them, has seen their work, or at least knows enough about them to say, "I think this person is worth talking to."
Here, the referrer may know nothing about me beyond a public comment and a few matching keywords. The referral code mainly determines who gets the bonus if I am hired. It does not explain why that person is in a position to recommend me.
That is not necessarily dishonest. It is simply closer to candidate sourcing than to a personal referral.
There is a cost for the company, too. A public referral program can generate more outreach and more applications without producing better candidates. Recruiters have more noise to filter, candidates receive more quasi-personalized messages, and the referral itself carries less information.
A company may reasonably decide that $1,500 is worth paying for a successful hire. The issue is not the size of the payment or the fact that it is conditional. The issue is making the payment available to anyone willing to send cold emails to public profiles.
Referrals work better when the referrer is expected to know something about the person being recommended. Broad candidate sourcing can be handled through recruiters, job boards, communities, and advertising.
Once a referral program becomes a public competition for attribution, candidates stop experiencing it as a recommendation. They experience it as spam.
So please don't make job referrals public.

The author addresses the phenomenon of receiving unsolicited messages for job referrals when the referral program is made public, arguing that this structure fundamentally shifts the nature of the interaction from a personal recommendation to mass spam for the candidate. The author details an experience where multiple individuals contacted them regarding a senior software engineer position, each supplying a different referral code, yet the underlying connection was based solely on matching technical keywords from public profiles rather than genuine personal endorsement.

The core of the argument centers on the mechanism of attribution. Traditional referrals operate on the premise that the referrer possesses specific knowledge about the candidate, having worked with them, or at least recognizing their worthiness. In contrast, a public referral system, where compensation is conditional upon a successful hire, allows unrelated parties to leverage public keywords to initiate outreach. The author suggests that the likelihood of several disparate individuals independently studying a candidate's history and converging on the same conclusion regarding a senior role is improbable. Instead, the likely explanation is that the author's public posts served as searchable data points that matched the requirements of the job advertisement, making the prospect of earning the associated bonus attractive, even if the connection is superficial.

Furthermore, the author points out that this system is highly susceptible to automation. Large language models can generate personalized messages that cite technical background, such as physics, CUDA, or LLM inference, without the sender needing to possess deeper knowledge of the candidate. This capability allows for the scaling of outreach, turning the referral pool into a mechanism for mass, quasi-personalized messaging. From the candidate's perspective, this results in a barrage of unsolicited communications, which the author categorizes as spam rather than a meaningful recommendation.

The problem extends beyond the candidate experience to the entity running the referral program and the company itself. While a company may reasonably value the $1,500 incentive for a successful hire, making this incentive accessible to anyone capable of sending cold emails to public profiles introduces significant noise. Recruiters face an increased volume of irrelevant communication to filter, and candidates are subjected to messages that lack the informational density of a true referral. The referral code primarily serves as a means of tracking attribution rather than verifying the quality or context of the recommendation.

The author contends that referrals function optimally when the referrer is expected to have substantive knowledge of the candidate. When the system is open to broad candidate sourcing, it is more effective to utilize established, structured channels such as recruiters, job boards, and specialized communities for discovery. By turning the referral process into a public competition for attribution, the system devalues the recommendation itself, transforming it into an impersonal method of noise generation. Therefore, the author concludes by advocating against making job referrals public to preserve the integrity and value of the referral mechanism by ensuring that recommendations carry genuine contextual information.