跳过「末日漏斗」：Cursor 招顶尖 50 人的打法

Ward rates today's hiring market an 11 out of 10, comparing it to the scramble for mobile engineers when iOS skills barely existed. The difference is speed: the mobile moment played out over about two years, while this one moves in days and weeks. He describes a tale of two cities, with headlines about huge offers to new PhD graduates next to news of big companies laying off 10% of staff, and he believes talent and labor are going through a resettlement.

The role he sees most in demand is the forward-deployed engineer, a deeply technical person who can partner with sales and a customer to deploy a product and help executives move from maximizing token use toward optimizing it. He thinks the best candidates are often people who were full-stack engineers three to five years ago, and that the shift can unlock a new chapter in their careers.

On the softer side, Ward points to very narrow, deeply specialized people and to new graduates whose work tools like Cursor can now partly do. He sees engineering, product and design squeezing together, and a return to craft, taste, decision-making and curiosity, which he links to the rise of the power IC.

Ward credits his first manager at Trilogy, Jeff Daniel, with coining "the funnel of doom." Most companies treat recruiting like a sales funnel with stages and pass-through rates, drawing candidates from applications, sourcing or referrals. If you reach out to 100 people and 20 reply, those 20 are by definition not the best 20, only the ones who happened to be receptive that day.

Weeding people out at every stage and hiring whoever remains means the company regresses to the mean as volume grows, even with a good interview process. Ward adds that this approach is very expensive. He suggests the sales analogy is part of why the problem persists: a widget is static and rational, while recruiting involves a hiring manager, a recruiter and a candidate, all of whom are human.

His alternative is to know up front who the top 20% are and use the assessment process to validate that thesis. In other words, start with a pillar of excellence rather than a wide funnel. This is the way senior specialist and leadership searches are normally run, and he wants every hire to be treated that way.

The first step, scoping, means getting clear on what the top 1% means for this specific role. Ward warns that "I'll know it when I see it" does not work, because you cannot recognize great without having defined it. You list the skills and experiences you need, stack rank them, and settle why the role matters and what success looks like, all before talking to anyone.

Good scoping then drives everything downstream: the sourcing strategy, the pitch, what the assessment tests for, and the case for why this person will have impact. Ward calls it the most underinvested part of most hiring processes, even though it is a cornerstone of leadership searches.

He cautions against defining great by company logos or by copying another company's process, since most companies are doing recruiting wrong anyway and your product is different. A team at a particular company at a particular time may have done remarkable things, which can serve as shorthand, but he says most companies jump to it almost immediately.

For mapping, Ward starts with trusted people who have high taste in talent. The mistake is asking "who is the best product engineer you know?" Instead, ask who has been the most collaborative with designers, or who can turn a framework into a product better than anyone. Because the questions map back to the scoping work, the same names begin to appear from different sources, and you can add them to the target list.

He also suggests picking two or three desired characteristics, such as curiosity or systems thinking, and looking for other companies whose hiring emphasizes the same traits, which can be done with a quick Boolean or AI search of job openings. Asked about systems thinking, Ward describes it as identifying the fundamental problem and breaking it into small pieces, and says the best systems thinkers are extremely tactical.

Tools like LinkedIn and AI help, but Ward says the most valuable method is also the most time-intensive: sitting down with people, every other week for six weeks, to dig through their networks and follow up on names. At Cursor, some team members' whole job is digging into employees' networks. Another tactic comes from Coda, where a Slack channel lets anyone drop in a profile of someone impressive, and the company rallies around how to engage them.

Lenny points out that this approach is slow and laborious. Ward's reply is a question to founders: where does talent fall on the list of what matters? If the answer is number one, then time, cost and effort cannot be obstacles, and it cannot be outsourced to a recruiting organization. He adds that what gets measured gets done, so hiring managers and leaders should be held accountable for hiring like any other metric.

Ward says one common failure is that recruiting owns the hiring goal even though recruiters never make hiring decisions. In his view, the recruiter is a confidence engine, not a decision-making engine. Nobody feels the joy or pain of a good or bad hire more than the hiring manager and team, so accountability should sit with them.

Ward says pursuing the top 50 starts a slow burn, but over time you plant seeds that mature at different rates, in weeks, months or sometimes years. If someone says the timing is wrong, he suggests asking not for an interview but for a next conversation: a coffee, a visit to the office, or an introduction to a teammate who admires their work. Such touchpoints can only move things forward, whereas waiting until late fall risks someone else getting them.

For hard-to-reach people, he says almost as much effort goes into finding a warm introduction as into the message itself, since someone who has ignored one outreach finds it easier to ignore the next. His motto "caring is free" captures the idea that a company can care about a candidate more than anyone else, and that candidate satisfaction scores show that feeling wanted comes first. He tells his team they are hiring one person for a role, not ten.

He also argues that a funnel mindset makes hiring too robotic. Instead of letting interviewer availability decide who talks to a candidate, a thoughtful recruiter picks someone with a shared interest. He considers the first conversation with the hiring manager the most important in the process, because it is the person who will make the decision, and he suggests the atomic unit of a search is the person, not the job spec.

Among his tactics are sharing a meal, inviting candidates to give a talk or work on a fun side project, and playing home games by bringing them to the office so they get a feel for the place, as with visiting a university campus. Lenny recalls Airbnb's version, getting people into the elevator.

Ward notes that research shows work samples are the best predictor of success, yet many companies still rely on one-on-one assessments across a table. Cursor is known for long on-sites for some roles, where candidates work side by side with the team on a project. That gives the company a strong signal and gives the candidate a real view of the job, which he values because good self-selection produces good data.

The recruiter acts as quarterback, but hiring managers contribute insight into who the candidate should meet, and recruiting coordinators handle much of the hosting and shepherding across a day or two. Meals are built into the schedule. Ward says that even candidates who do not end up joining can leave as promoters, and because they are influential, those promoters carry extra weight.

Cursor tried dropping work trials, but Ward says the team struggled to have confidence in its decisions, so they are bringing them back. Other functions have versions tailored to the role: go-to-market candidates work through a challenging customer scenario, and product candidates do a project with the hiring manager. Candidates choose from a set of problems, can have the trial resized, and can do it on a Saturday, while Cursor moves quickly once they are ready.

Cursor's decisions are led by the hiring manager rather than a committee, with team input and a debrief the same or next day, and the founders confirm the decision. Ward stresses that it is not a democracy. When a candidate holds a huge competing offer, he says the goal is not to sell at the close: recruiters should learn the candidate's motivations early and shape the whole experience around them, so that the offer feels like a natural next step.

Cursor tries to focus on intrinsic factors such as the role, the team and the people, because the extrinsic ones are hard to win against enormous salaries. He thinks people who chase the highest number often find six months later that it is just their new baseline, and he is comfortable when someone concludes that Cursor is not the right fit. As a recruiter he would rather learn that now than redo the work in six months.

Ward says he respects companies that refuse to negotiate but that this approach may not suit Cursor. His advice to candidates is to make requests follow a clear line of logic and to state the few things that matter up front rather than adding new asks later. He also says that with the pace of AI hiring, the most fixed and finite resource is time, and that researchers who once worked on the sidelines are having a well-deserved moment.

Asked what gets a fence-sitter over the line, Ward gives three answers. Closing happens throughout the process, by removing objections along the way. It is also a team sport: for high-caliber candidates, the team holds a short daily stand-up on what everyone is doing to get them to yes, and each candidate gets their own Slack channel. Former peers can help too.

The third is making it personal. He tells the story of a New York candidate who had trained as a classical violinist; a recruiter found a specialty instrument shop, bought a unusual Turkish string instrument for a couple of hundred dollars and presented it at the dinner where the offer was made. Ward says the point is the thoughtfulness: candidates remember an experience that felt tailored to them.

Ward admits this does not scale with a hundred candidates per role, but with a small group you have already been cultivating, the time is simply displaced to other parts of the process. His thesis at Growth by Design was that companies at the cutting edge of a technical revolution were still recruiting in a web-browser way, and that committing real time and effort could be an unfair advantage. He argues execution comes down to having the right talent at the right time.

Ward calls working with an executive team that values recruiting a gift and at times frustrating. The founders care and will join conversations, and the team hired roughly its first 80 people itself. The drawback is that roles blur and things happen without Ward knowing, which he says he would take any day. He also holds the hot take that recruiters should be among the highest paid on the team, given a thin supply of strategic recruiters.

On talent density, Ward warns that companies over-focus on individuals when density is a property of a team. He prefers thinking of a 10X team over a 10X person, saying you cannot have a 10X person on a bad team. Great managers are puzzle builders who assemble complementary people, and an excellent person gets throttled on an unsupportive team.

For early-stage companies, the common mistake is a silver-bullet first talent hire who must both make a backlog of hires and build a system, two competing jobs. He recommends decoupling them, for example pairing a systems builder with a contingent recruiter. In a first recruiter he looks for curiosity and business focus, then a high motor, noting that recruiters do best at 90 to 110% of capacity.

Ward says his team looks different at each company, and Cursor needed recruiters who understand leadership searches, avoid the funnel numbers game and bring their own networks. He looks for three attributes: excellent humans who are excellent at the work, people with a chip on their shoulder and an internal drive to be excellent, and people who put the team before themselves. He tries to balance being operational and in the weeds with painting the bigger picture.

One trend he adds is reneges on accepted offers, which he sees increasing in a hot market. Cursor treats the period between acceptance and start as a second motion: dinners with others who have accepted, a laptop sent in advance, and a sense of community, both to make sure people show up and to speed their ramp.

Ward also points to the rise of the talent engineer, a function that builds tools for recruiting. His warm take is that every recruiter should be one, since they can now build tools themselves in Cursor instead of begging engineering for time. Lenny agrees that every function will get its own version of an engineer.