A playbook for founders
Hire your first remote engineer in 2026 - the playbook founders learn the hard way.
A hiring funnel playbook for founders making their first few engineering hires.
Aim for volume, efficient filters, and AI age tests.
With hundreds of applicants you have to be smart about your time.
01 — Volume
You can't filter what you don't have.
Aim for 1,000 candidates a month. Not because you'll interview them all, because most applicants are completely unqualified or bots.
Post everywhere remote-first.+
The best engineering candidates for a remote startup won't be in the well known tech cities you know. They're going to be in secondary cities, with different time zones, and some have non-CS backgrounds. Get away from competing with the crowd.
Post in multiple cities. The algorithms favor specific city posts than country-wide "Remote" posts. Aim for 5 to 20 cities.
Build a career page worth visiting.+
The first thing a strong candidate does is Google your company. Day-in-the-life. A vision they can connect with. Team photos. Values. A careers page does two jobs: it tells the right people what a week here feels like, and it tells the wrong people to move on.
Launch a $1,000 referral bonus.+
Use Sequoia's strategy called "memory palace": sit each teammate down and walk them backward through high school, college, internships, and past jobs, asking questions like "who were the three smartest people in your class?" or "who did you look up to at your last job?". Budget a 30-minute session per person, aim for a 90% response rate on the intros, and when a candidate passes use these same questions and process with them. Full Sequoia playbook →
02 — Vetting
Spend zero founder time on unqualified candidates.
Add knockout questions to the application process.+
Knockout questions are a filter at the top of your application form: if a candidate picks the wrong answer on a hard requirement, the form auto-rejects before a human ever sees the submission. Wire the form through Tally, Youform, Typeform, or your recruiting software with conditional logic, use only multiple-choice, yes/no, or number-threshold answers, and surface only the candidates who clear every question. Keep it to three to five questions.
Examples on the right rotate through our recommended set.
Filter unqualified candidates, VA-led, AI-assisted.+
A VA (virtual assistant), using Claude or ChatGPT, reads every application that cleared knockouts against the job spec, verifies LinkedIn and GitHub match the resume, and produces a ranked shortlist. You get 10–15 names with rationales. The rest get a polite rejection email from the VA. Pro tip: The VA should flag if the candidate has a repo with more than 20 stars.
Run a 15-minute structured screen, VA-led.+
A VA (virtual assistant). The screen tests communication skills, English fluency, and one behavioral question like: "Tell me about a time you disagreed with a peer and you were wrong. How did you handle it?" 10x cheaper than a founder, and the signal is just as good at this stage.
Pro tip: skip the screen for warm intros.+
Candidates who reach out directly on LinkedIn or via a warm intro have selected into your company. They've done the research. Send them straight to Verification.
Years of production experience
Pass: 1+ years.
Years shipping on your stack
Pass: 1+ years in TS + Node + React.
Compensation alignment
Pass: Yes to max $X.
Starting a new project, you prefer…
Pass: rough goal + freedom, or define it yourself.
Comfortable doing work "below" your title?
Pass: Yes.
You disagree with the founder on a big call
Pass: raise it once, then commit — or keep pushing.
First or only responsible for something?
Pass: Yes. Evidence of unscaffolded ownership.
Something breaks, unclear whose job it is
Pass: fix it if you can, flag it if you can't.
OK not knowing your work 3 months out?
Pass: Yes. Roadmaps at small companies shift constantly.
03 — Verification
Volume gives you candidates. Your test decides who you hire.
Select the top candidates from the list your VA provided to be tested. Your test needs to filter for people that work like you do, that should include using AI as a productivity multiplier. A well defined concise problem is not a good fit anymore in the age of AI. The filter here can be a test for critical thinking skills, agency and judgement. Leave the raw technical skills assessment for a live interview exercise, where the candidate can't use AI.
Introduce deliberate ambiguity.+
Include one requirement that can't be completed without a clarifying question. Candidates who ask filter in as proactive thinkers. Don't forget to tell them: "If you have any questions about the project, let me know."
Force architectural tradeoffs.+
Make sure the test has no single right answer, just tradeoffs. The candidate shouldn't be spoon-fed the choice.
"Design the product catalog for a marketplace. Each seller defines their own product fields: cameras have focal length and sensor size, furniture has dimensions and materials. Buyers need to run queries like 'all orders over $500 in the last 30 days, grouped by seller category.'"
Hidden tradeoff: Open schemas pull toward document stores; analytical joins pull toward normalized SQL. A strong candidate will be able to identify and articulate the tradeoffs involved when asked.
"Design the pricing engine for an e-commerce platform. Customers need prices to update instantly as they add items, apply promo codes, or change quantities. The same prices must be the ones charged at checkout, and must match exactly on invoices six months later."
Hidden tradeoff: Instant UI feedback wants pricing logic on the frontend; auditability and a single source of truth want it on the backend. A strong candidate names the duplicated-logic problem and proposes where to put the seam (shared package, backend-only with optimistic UI, etc.).
Have them defend their submission live.+
Walk the candidate through their take-home. At three or four decision points, ask "why did you do X here?" Then hand them a small new requirement and have them extend the code live, on screen-share. Capable engineers can explain and extend their own submission in real time, here's where you test for actual technical skills. Pseudo-code, diagrams and explanations are OK here.
04 — Velocity
Good engineers run three to five processes at once. If you're slow, you lose them.
Hold a 24-hour SLA.+
Measure it. Track it. At VA-managed stages, aim for under an hour. The urgency is quantified: an MIT study of 15,000 leads found that the odds of contacting a lead drop 100x when you wait 30 minutes instead of 5, and the odds of qualifying that lead drop 21x. Hiring funnels behave the same way. A candidate who waits 72 hours has already signed somewhere else. MIT Lead Response study (PDF) →
Use a paid take home test.+
A 4–16 hour project is enough. Pay for the work — not just to be a nice person, but because you're creating goodwill that will help close the great candidates.