How to hire people who thrive on ambiguity

1 The bottom line

Hire people who act without being told, show taste through what they built and chose not to ship, and learn and unlearn fast, then test it with hands-on work and ruthless references. When building is cheap, judgment and ownership decide who ships.

2 Key takeaways

• Screen for agency first. Skills are now at everyone's fingertips, so action is what separates people.
  • At Notion, Eric Lou went from strategy docs to Figma to prototypes after asking what it would take to be hired in a founder's first five Max Schoening, Notion11:01.
  • A GrubHub QA hire with an amazing resume waited to be told what to do and was let go after a week Casey Winters, SuperMe37:30.
• Screen for taste, because judgment is now the bottleneck. Engineering capacity no longer caps how many good products get made.
  • Five years ago you could give a founder $20 million for engineers and still burn it all without product taste Andrew Maguire, Volo Ventures4:30.
  • Anthropic's Claude Code team will hire almost anyone who has strongly shown product taste Cat Wu, Anthropic16:40.
• Measure learning speed as the speed of unlearning too. Tools change fast, so what someone can drop matters as much as what they can pick up.
  • Mosseri wants people willing to put themselves out there and sound foolish, like when learning a language Adam Mosseri, Instagram14:17.
• Replace talk with hands-on tests. Polished interviewing hides weak thinking.
  • 31,832 people applied to be a PM at Whatnot in two years and it hired one, using a case study they had to defend Tom Verrilli, Whatnot49:11.
• Check references until one is negative. A negative answer shows you have exhausted the pool.
  • DoorDash's Tony does 20 references on every senior hire Keith Rabois, Khosla Ventures10:50.

Hey, let me show you this thing I just built. Instant credibility.

Josh Woodward, Google Labs

2.1 Decide what you are screening for

Four traits predict who thrives when nothing is defined: agency, taste, fast unlearning and the ability to carry an idea alone.

• Screen for agency, the sense that the world around you is malleable. Max Schoening, Notion11:01
  • Before, people could say they would never do something because of a skill gap. Now what matters is whether you act.
  • Expect staff who keep asking what a PM, designer or engineer is supposed to do to struggle.
  • The most valuable person takes an idea to done and has the taste to know it is great. Everyone now manages something, whether agents or teams Andrew Ambrosino, OpenAI31:31.
• Define taste as more than visual polish. Andrew Ambrosino, OpenAI11:00
  • Taste also covers how something fits the system, which theme it belongs to, and how to present it.
  • It includes whether an animation's feel matches its meaning. The real question is what a thing should be when anything can be built.
  • Paul Graham wears cargo shorts and clearly has taste Andrew Ambrosino, OpenAI11:00.
  • Taste is the ability to predict whether a chosen in-group will like something. It is built through reps of ideas and feedback and cannot be speedrun Max Schoening, Notion57:50.
• Count unlearning as part of learning speed. Josh Woodward, Google Labs17:00
  • Ask how fast someone can drop what they learned once testing shows it is wrong.
  • Also look for people who sprint hard but pace themselves over years, and who can build trust with people as well as agents.
  • Mosseri's baseline for almost any role: fire in the belly, quick learning, and enough self-awareness to take feedback Adam Mosseri, Instagram14:17.
  • Anthropic's loop has not changed in three years. It looks for people who work out the user value for this user group and this technology, rather than pattern matching on consumer or B2B SaaS Dianne Penn, Anthropic40:05.
• Hire barrels before you hire more ammunition. Keith Rabois, Khosla Ventures17:10
  • A barrel takes an idea from inception to success alone, motivating people, gathering resources and measuring progress.
  • Adding people without adding barrels stacks them behind the same initiatives and raises coordination cost.
  • PayPal had 12 to 17 barrels among about 254 people in Mountain View. A large company like Lotus had two Keith Rabois, Khosla Ventures17:10.

2.2 Find people who have already shown the work

Public projects and built apps are the cheapest early filter for agency and taste.

• Screen public GitHub profiles for side projects. Jack McDermott and Tanmay Jain, ElevenLabs24:00
  • Look for engineers who build side projects and obsess over a user pain point. Also look at companies with strong product-building cultures.
  • This self-selects people who want end-to-end ownership, and their referrals compound.
• Welcome a built app instead of a resume. Lazar Jovanovic, Lovable1:33:20
  • He became Lovable's first official vibe coding engineer by sharing projects, failures and knowledge on YouTube and LinkedIn, and by doing the job before being hired.
  • Elena Verna picked him out of many vibe coders because he was building in public. Several other hires stood out by sending apps, and Lovable employees always open a lovable.app link.
• Build on undiscovered talent, not what every recruiter wants. Keith Rabois, Khosla Ventures25:00
  • Startups have a fraction of the salary cap of big competitors. Ask what Meta, Google or Coinbase would miss about a candidate and why.
  • Younger people have fewer data points, so rigid screening processes score them poorly. That creates your edge.
  • Peter Thiel taught this to Rabois in his first week at PayPal, and Rabois has followed it for 25 years.
• For designers, hire from the portfolio and ignore pedigree. Evan Spiegel, Snap40:50
  • Look for wide range. A uniform style signals an artist rather than a designer who empathizes with different needs.
  • Ask the candidate to pick one piece they feel strongly about and tell the story of why they made it and what they learned.
  • Spiegel interviews every designer who joins, and the team includes people from 3D animation and electrical engineering.

2.3 Test hands-on in the interview

Make candidates do, show and defend real work, then watch how they handle the unknown.

• Give a hands-on case study and make them defend it. Tom Verrilli, Whatnot49:11
  • Whatnot gives a prompt and data, asks for a PRD, and has the candidate defend it verbally.
  • Look for macro system thinking and micro impatience about how to validate quickly.
  • Many candidates present well, but their thinking decays fast when asked for specifics.
• Ask what they built but chose not to ship. Peter Sellis, Snap and Discord35:27
  • Ask about things they prototyped, held in their hands and decided the world was not ready for. Do not ask about things that tested badly.
  • Restraint is one of the few muscles of taste you can exercise.
  • Sellis saw Evan Spiegel say no to better ideas than he has seen at almost any other consumer startup.
  • Also let candidates show something they just built. It earns instant credibility Josh Woodward, Google Labs51:45.
• Hand over a narrow outcome and watch. Keith Rabois, Khosla Ventures19:50
  • See whether they deliver it come what may, or come back early with the issues, what they tried and the root causes.
  • Then expand the scope and difficulty you entrust to them. A barrel lets you fire and forget.
  • An intern solved the 9 p.m. cold smoothie delivery on his second day, after the office team had failed at it. He was later given almost everything to do.
• Listen for what they spend the interview on. Tom Verrilli, Whatnot18:06
  • Candidates who dwell on driving alignment and stakeholder management trend down. Look for people who made decisions and drove change, not those who polished the edges of an existing product.
  • Ask about unknown, unstructured situations and look for people energized by them Josh Woodward, Google Labs51:45.
  • Test paired opposites with stories Peter Sellis, Snap and Discord1:16:33.
  • The pairs: confidence with humility, process with comfort throwing the checklist out, and a clear picture of the world 6 to 24 months out with a borderline annoying bias for action.
  • Ask senior candidates what they would have done differently as CEO, then why they could not persuade the real CEO. It shows how much of the business model they absorbed and whether they understand tradeoffs. The point is not to hear them criticize the company Keith Rabois, Khosla Ventures13:25.
  • Tom Verrilli says he used to be one of the alignment-first PMs himself.

2.4 Check references ruthlessly

References work when you do many of them and ask about the real job.

• Keep going until you hit a negative reference. Keith Rabois, Khosla Ventures10:50
  • Treat referencing as a trainable skill and do many references on senior hires.
  • A negative answer shows you have exhausted the pool.
  • Ask what would make the person most successful and what the root cause would be if it did not work out.
  • Rabois applied the same rule to founders: no stopping until you hit a negative reference.
• Frame the question around the role, not generic employee quality. Keith Rabois, Khosla Ventures12:55
  • Ask whether the person can be a world-class entrepreneur, or whatever the real job requires.
  • Many VCs passed on Fair after asking whether Max Rhodes was a good employee, and got a mixed answer. The answer to whether he could be a world-class entrepreneur was yes.

2.5 Check the hire after they start

Close the loop fast, so you learn from every hire and cut passive ones early.

• Run a 30-day check on every hire. Keith Rabois, Khosla Ventures9:10
  • Have the whole hiring team ask whether they would make the same decision.
  • Research he read says this is about as accurate as judging a year or two out, so you get a tight feedback loop.
• Throw early hires in the deep end and drop passive ones fast. Casey Winters, SuperMe37:30
  • At early stage, direct reports to founders get no 90-day plan. Let go of people who wait to be told what to do.
  • SuperMe's ops hire laid out in the interview how he would approach the work and proposed next steps, so nothing was uncertain.
• Pair high agency with high accountability. Fiona Fung, Anthropic39:00
  • Give everyone freedom to propose and build, then ask what hypothesis they are trying to prove and what the accountability is for it.
  • Agency alone produces lots of activity with no clear result. The Claude Code and Cowork team holds the pair as a core value.
• Judge a PM against running the product with no PM. Peter Sellis, Snap and Discord16:40
  • A PM has to add value over having nobody, and the cost of a weak PM is often not counted.
  • Sellis argues PMs skew bad because the best leave to become founders or executives, while weak ones stay in a lucrative role. Great engineers rarely complain about PMs because they bend them to their will or avoid bad ones.

2.6 Where they disagree

Some tactics fit only certain stages, teams or products.

• Flexible generalists versus deep experts.
  • Coconote's four-engineer team ships a lot because each engineer can work beyond their main strength, such as an iOS engineer also doing back-end work Brett Bauman and Zack Hargett, Coconote27:27.
  • Anthropic's Claude Code team and growth team hire product-minded engineers who can act as mini PMs Cat Wu, Anthropic16:40, Amol Avasare, Anthropic43:13.
  • Fiona Fung found early Claude Code had good product generalists but lacked distributed systems expertise. Where models still need verification, she keeps investing in deep expertise Fiona Fung, Anthropic17:35.
  • Ian Silber says no single designer ticks every box, so build a team of complements. A startup might hire the best generalists Ian Silber, OpenAI28:15.
  • OpenAI does not expect AI experience. It wants curiosity and the ability to learn AI in the design process Ian Silber, OpenAI29:10.
• Former founders and juniors fit only some cultures.
  • OpenAI has more than 120 ex-YC founders on staff, hired for taste and the drive to build Tibo Sottiaux, OpenAI23:45.
  • ElevenLabs pairs product-minded engineers with former founders, but only inside a culture that supports speed and autonomy Jack McDermott and Tanmay Jain, ElevenLabs24:00.
  • Casey Winters says to keep hiring juniors with AI-native onboarding on day one. They have no bias against AI coding tools, and Shopify still loves junior talent. Mix them with experienced people Casey Winters, SuperMe36:15.
• Some screens fit only some products.
  • At Solid Starts, each team member messaged the founder to join because of their passion for baby feeding. Mission obsession suits a product with a cause Yuliya Lennox, Solid Starts56:14.
  • For a channel-led role like TikTok, a skilled marketing generalist can translate the skill. A TikTok Shop background adds selling skill and current formats, and avoids recycling year-old trends Kieran, StudyFetch45:15.
  • Kieran had never made a TikTok video before joining StudyFetch 2.5 years earlier.