The May 2026 HN Hiring Gate: 60 Minutes in Lovable, Or You're Out
The May 2026 HN "Who is Hiring" thread added a 60-minute live build gate. Here is how to source candidates who can actually pass it.
The May 2026 Hacker News "Who is Hiring?" thread (item 47975571, posted May 23) quietly rewrote what "qualified" means at early-stage startups. The new gate is a 60-minute live build inside Lovable or Replit, watched, with a hard requirement that you show AI-built work shipped in the last 60 days. If you are sourcing for these roles with LinkedIn boolean and a "Cursor" keyword, you are filtering on the wrong axis.
What actually changed in the May 2026 thread
The canonical post in this month's thread spells the format out without ambiguity:
Required: top 1% with Lovable, Replit, or Cursor. Public GitHub or portfolio with shipped projects we can play with. Ability to ship a complex tool (multiple data sources, APIs, server hosting, LLM integration) in under a day.
The interview loop attached to that posting reads:
Process: 30-min intro, 45-min founder conversation, 60-min live build in Lovable or Replit (we watch you ship). Offer within 2 weeks.
This is not one weird startup. The same thread contains Waypoint Learning running a parallel variant: a one-week build challenge with a $500 prize and an automatic final-round slot for the top five submissions, deadline May 11. Featurebase, also in the thread, is the counter-example. It is still hiring a traditional NextJS/Node/Mongo product engineer with no live-build gate. The pattern is concentrated at AI-product startups and YC-stage companies, not universal, but it is the fastest-growing interview format in the thread.
The reason it spread this fast is that Y Combinator is now reporting 95% of its latest batch ships code that is AI-generated. Earlier reporting put a quarter of the W25 cohort at codebases that were 95%+ AI-written. The companies running this gate are not testing whether you can use an AI coding tool. They are testing whether you can ship a working product faster than the person next to you with the same tools.
Why your LinkedIn boolean is now a negative signal
Here is the part most sourcers have not internalized yet. "Cursor" in a LinkedIn bio used to be a useful filter. In May 2026 it is noise. Everyone has Cursor. Everyone has a Lovable account. The signal the May 2026 HN gate is asking for is not tool familiarity. It is shipped artifacts in the last 60 days. That is a recency-of-activity filter, and LinkedIn does not expose it.
We ran an internal query for professional-network headlines combining "Lovable", "Bolt", or "Replit" with "shipped" or "deployed" builder language. The match count was effectively zero. Candidates who can actually pass this gate are not self-tagging on LinkedIn. They are on GitHub pushing to a repo that imports the Lovable SDK, or they have a personal site that points at a lovable.app subdomain.
The same shift is happening in PM hiring. Vibe-coding interview rounds are now confirmed at v0, Bolt, Figma, and Perplexity, with design-led variants at Stripe and Netflix. None of those processes care what your LinkedIn headline says.
The X-ray queries that actually work
Stop searching titles. Start searching deployed artifacts. The four queries that surface real candidates for this gate:
site:lovable.appplus a domain or stack keywordsite:bolt.newplus a recency operatorsite:replit.apppaired with a personal-site hostsite:*.vercel.apppluslovableorv0in the repo readme
On GitHub the equivalent is filtering pushed repos from March through May 2026 that carry the lovable, v0, or bolt topic, or that import the Lovable or Bolt SDKs. The commit graph from the last 60 days is the qualification, not the resume.
This is exactly the friction we built Refolk for. You describe the candidate in plain English ("engineers who shipped a public Lovable or Bolt app in the last 60 days and have a GitHub profile with recent commits") and Refolk runs that query across GitHub, LinkedIn, and the open web at the same time, returning a ranked shortlist instead of three separate browser tabs of half-matches.
The candidates who pass this gate do not look like senior engineers
Andrej Karpathy retired the term "vibe coding" in February 2026 and replaced it with "agentic engineering": "programming via LLM agents is increasingly becoming a default workflow for professionals, except with more oversight and scrutiny." That reframing matters for sourcing. The HN gate is not testing algorithmic depth. It is testing orchestration, product taste, and ship velocity under observation.
The people who clear it are often:
- Ex-founders who shipped a failed product but have six deployed prototypes
- Designers who picked up Lovable and Bolt in late 2025
- PMs who started shipping their own internal tools instead of writing PRDs
- Junior engineers with strong GitHub graphs but no FAANG line on the resume
They are rarely the staff engineer with a clean LinkedIn and a Google badge. The big players are going the other direction on purpose. As one practitioner put it: "The big players are committed to the LeetCode style interview. Why? Because they are terrified of AI impostors." Startups are doing the opposite. They have decided the impostor risk is lower than the velocity risk.
The HN gate is not testing whether you can use AI tools. It is testing whether you can outship the person next to you using the same ones.
If you are a recruiter sourcing for a founder running this format, your job has shifted. You are not screening for credentials. You are screening for an artifact trail. That is a different muscle.
The hidden filters nobody is naming
Two non-obvious things are happening inside this format that founders running it should think harder about.
The format selects on wallet, not skill
The startups posting these roles are not always providing enterprise seats for the AI tools during the interview. So a candidate on ChatGPT's free tier is racing against a candidate paying $200 a month for Claude Max and $20 for Cursor Pro. The speed delta in a 60-minute window is real. If you are running this loop and you care about pipeline diversity at all, supply the seats. Otherwise you are filtering on subscription tier and calling it skill.
A live demo is easy to game with a pre-built scaffold
A single 60-minute session rewards rehearsed templates. A candidate who has run this same build five times in practice will look like a genius. The signal that is much harder to fake is a deployed portfolio across 60 days, with commits spread across multiple projects, multiple stacks, and multiple failure modes. Weight the artifact trail over the live demo. The artifact trail is what proves the candidate can do this on a Tuesday at 3pm with a real bug, not just at the interview.
A sourcing playbook for the next 30 days
If you are filling one of these roles between now and the June HN thread, here is the order of operations that actually works.
1. Source the artifact, not the title
Run the four X-ray queries above. Pull every deployed Lovable, Bolt, and Replit URL you can find. Cross-reference the deployer back to a GitHub handle and a personal site. The candidate's name on the LinkedIn profile is the last step, not the first.
2. Filter by 60-day freshness
The HN posters are explicit: shipped work from the last 60 days. Anything older does not count, because the tools have changed underneath the candidate. Lovable in March 2026 is not Lovable in May. Pin your commit-graph filter to March 23 through May 23.
3. Use the project itself in your outreach
This is the one place the numbers are unambiguous. Codility's technical recruiting team co-authored outreach emails where the hiring manager referenced a specific GitHub project. Those emails generated a 30% response rate from passive candidates, 5x higher than generic outreach. If your first message does not name the candidate's deployed Bolt URL, you are throwing away most of your pipeline.
4. Pre-qualify with the deploy URL before the loop
Before you spend a founder's 60 minutes watching someone build, ask the candidate to send three deployed URLs from the last 60 days. The candidates who cannot produce them are not going to clear the live build. This saves the founder six hours a week and stops you from burning your best slots on rehearsed-template candidates.
Refolk handles steps one and two in a single prompt. Step three is still your craft. We are not going to write the outreach for you, but we will hand you the project URL to reference and the GitHub handle to attribute it to, so the message writes itself.
What this means for the June thread
The June 2026 HN "Who is Hiring" thread will almost certainly contain more live-build gates, not fewer. The format has a viral quality among founders: it is fast, it is concrete, it selects for the people they actually want, and it shortcuts the resume theater. Expect to see it appear at slightly larger Series A companies by the end of summer, and expect at least one of the named AI-product companies (v0, Bolt, Lovable itself) to formalize it as their default loop.
The sourcing implication does not change. Stop searching for "AI engineer" on LinkedIn. Start indexing the open web for deployed artifacts and recent commit graphs. The candidates who can pass this gate are findable, just not in the place you have been looking. That is the entire job now. If your stack cannot do plain-English search across GitHub, LinkedIn, and the open web in one query, you are going to lose these candidates to the founder who DMs them on a Saturday with a link to their own Bolt project.
FAQ
Is the 60-minute live build gate legal and fair?
It is legal in most jurisdictions if applied consistently. The fairness question is real, though. The format leans on the candidate's personal subscription stack (Cursor Pro, Claude Max, ChatGPT Plus) unless the company provides seats during the interview. If you are running this loop, supply the tools for the session, or you are selecting on disposable income alongside skill. That is also the cleanest way to defend the process if a candidate later raises a discrimination concern.
How do I find candidates who shipped a Lovable or Bolt project in the last 60 days?
Google X-ray for site:lovable.app, site:bolt.new, and site:replit.app, then cross-reference deployers to GitHub handles with recent commits. On GitHub, filter for repos with the lovable, bolt, or v0 topic pushed since March 2026. This is a multi-source query, which is why we built Refolk: you describe the candidate in plain English and get matches across GitHub, LinkedIn, and the open web in one shortlist instead of stitching three searches together.
Does this format actually surface better engineers than LeetCode loops?
For startup product roles, in most cases yes, because it tests the thing the job actually requires: shipping a working tool with an LLM in the loop. For infrastructure roles at scale, it is a worse signal than a systems-design interview. Match the format to the work. A founding engineer at a YC AI-product company and a senior platform engineer at a payments company are not the same job, and one interview format cannot screen for both.
Should we drop LinkedIn entirely for these roles?
No, but invert the order. Start with the deployed artifact, find the GitHub handle, then use LinkedIn at the end to confirm location, work authorization, and recent employer. LinkedIn is verification, not discovery, for this candidate pool. The discovery happens on the open web and on GitHub commit graphs from the last 60 days.