Last September, a plain white billboard went up on Larkin Street in San Francisco's Nob Hill neighborhood. And for the vast majority of people who passed by it, it was nearly incomprehensible (even for a city that’s been memed for its increasingly bizarre AI billboards). 

It was simply one line of black text on a plain white background:

https:// {64659, 123310, 75584, 8138, 38271}

Alfred Wahlforss, co-founder and CEO of Listen Labs, needed to hire more than 100 engineers and was going up against companies with name recognition he had no way to match (at the time, Listen Labs was a Series A company). 

"Recruiting had been hell," he admitted at a Notable fireside chat.

So he tossed the typical recruiting playbook out the window completely and stopped competing on the pitch.

To most people passing the billboard, the numbers meant nothing, which was the point. To the audience Wahlforss wanted, they were obviously tokens, the units large language models break text into. Decode them, and you landed on a website with a puzzle challenge that engineers were eager to sink their teeth into.

First, a quick detour into one of the most famous techno clubs in Berlin. Housed in a massive former power plant, Berghain is regarded as “the world’s most exclusive nightclub” with a strict (and ever-shifting) door policy that’s part science, part vibes. 

The challenge for the billboard-decoders: Build an algorithmic bouncer for Berghain. Fill the club with exactly 1,000 people, at least 40% Berlin locals, and at least 80% wearing all black. Reject as few people as possible getting there. Anyone who solved the puzzle got an interview with Listen Labs, and the winner got an all-expenses-paid trip to Berlin.

Underneath the edgier premise, it's a constrained optimization problem dressed up in club attire. It is also a different spin on the problem Listen Labs solves every day.

Listen Labs runs AI-moderated customer interviews at scale. When Sweetgreen wanted to know whether to rebuild its menu around protein, Listen wrote the interview guide, went out to that audience, paid people to talk, ran hundreds of interviews in parallel, and came back with an answer. The hard part is the door. As Wahlforss describes it, the work is selecting the right people from a huge pool, under constraints, with imperfect data. Most of Listen's engineering effort goes into building and verifying that audience.

The billboard traveled, quickly racking up roughly 5 million impressions across X and Reddit, got a reply from Sam Altman, and coverage from local and national outlets. 

But the success of the experiment hinged very little on volume. A careers page tries to appeal to everyone who reads it. The billboard was built to lose almost everyone who saw it, on purpose, and the people it didn't lose had already self-identified as the right hires. "Candidates already understood what kind of company we are," said Wahlforss.

Around 10,000 people engaged with the challenge in some form. 430 solved it. 60 got interviews. Some got hired. And the winner was indeed flown out to Berlin, all expenses paid.

For a founder and CEO who says he spends about 80% of his time on hiring, you might be surprised that he doesn’t worry all that much about picking the right candidate.

Perhaps that’s why earlier this year, after closing its $69M Series B round valuing the company at $500M, Listen Labs stayed on theme and threw its own Berghain-inspired party, complete with bouncers at the door, and invited prospective candidates to join. It might have cost $25,000, Wahlforss admits, but the math works out when a recruiter placing one engineer charges at least $50,000. Make a single hire off the party, and it has paid for itself twice over.

"If you're competing with companies offering insane compensation, job ads won't do the job," Wahlforss wrote. "Design a filter so good that the right people opt in on their own."

“Back in 2014, I met a 24-year-old who wanted to spend an hour talking about the parts of building software that most irritated him.”

Notable’s Glenn Solomon reflects on over a decade backing Mitchell Hashimoto, and leading the seed round for his next ambitious company, Superlogical.