The First Engineering Hire Playbook for Founders Who Are Still Writing Code

The first engineering hire for an AI startup technical founder should solve a validated shipping bottleneck. Hire for the work that is holding you back, look for proof the person has shipped alone, and keep the interview process short and built around real problems. Those decisions need to come in that order, because finding someone fast cannot make up for choosing the wrong role.

Why the moment to hire is harder to read than founders expect

Being overwhelmed feels like evidence that you need help. It is not evidence that you need a permanent engineer. The pressure can arrive before you know which problem customers will pay you to solve, so hiring may lock in work that still needs questioning. CRV’s founding engineer guide calls hiring before problem validation one of the most expensive mistakes a founder can make. While validation is underway, a contract engineer offers less risk and an engagement that can be wound down cleanly.

The timing becomes clearer when at least two concrete triggers appear together. A pre-seed or seed round has closed, founder bandwidth is gone, the problem is validated enough to define 90-day success, or the roadmap requires technical skills the founding team lacks. Funding alone does not settle the question. Carta’s State of Seed 2025 report describes leaner seed-stage teams than in 2021, making each early hire a larger part of the team’s ability to reach meaningful milestones.

What a founding engineer actually is, and why that matters before you write a job description

A senior engineer executes within an existing system. A founding engineer creates that system while building the first product. A CTO builds the engineering organization and sets long-term technical strategy, a job that comes later once there is a team to lead. The founding engineer remains a hands-on builder, but thinks like an owner and helps decide what gets built. Treating these roles as interchangeable produces the wrong job description and the wrong interview.

In 2026, AI coding tools do more of the code generation, so more of the founding engineer’s value comes from architectural judgment and checking AI output. Raw coding output is not enough on its own. The person must make decisions AI cannot, move forward from an incomplete spec, and explain why a design fits the product. Someone who needs every requirement settled before starting is mismatched to the job, even if their code is excellent.

The word “founding” also does not make someone a co-founder. CRV distinguishes the two on legal and structural grounds. A co-founder joins when the company is formed or close to that time, has founder-level common stock, and often has governance rights. A founding engineer is an early employee with options from the employee pool, higher salary and less equity than founders, and no board seat. That distinction belongs in the role definition before the offer conversation.

Matching the engineering archetype to your specific bottleneck, not to a job title

What is the single biggest thing preventing you from shipping? Start there, rather than with the title another startup advertised. If you cannot build the product end to end, the first hire should be a full-stack engineer who can ship solo. If the product exists but its AI/ML core does not work, hire an ML/GenAI engineer before adding app engineers. More interface work will not fix the core.

A different blocker needs a different engineer. A product that fails under real load needs backend or infrastructure skills, because the constraint is architecture. An unmaintainable codebase calls for a senior or staff engineer with architecture skills who can navigate it independently. If none of these blockers applies and you can still build the product yourself, keep building. A job title is not a diagnosis.

The kind of company helps narrow the search, but the search still needs to start with that diagnosis. SaaS and enterprise products generally need a full-stack generalist to ship features quickly, while financial technology and regulated products need backend strength for security and correctness. Consumer apps need someone who can do mobile and backend work from the start. Developer tools call for full-stack or platform engineers who can use the product themselves and whose software engineering skills meet a high bar. For an AI/LLM startup whose core constraint is model behavior, the priority is an ML/GenAI engineer who can ship the eval layer first.

That AI role is not automatically a classical ML research role. Much of the work has shifted from custom model training into retrieval, agents, evals, and orchestration. If the job is deploying production AI systems rather than publishing papers, hands-on LLM deployment experience outweighs a PhD without production experience. Hiring someone for their name in research when you need deployment skill holds up the product and spends equity on the wrong person.

Green means a history of shipping a product alone, comfort with incomplete requirements, and architecture opinions the person can explain. Yellow means the title looks right but there is too little proof of those skills. Red means they wait for a complete spec or cannot explain a design decision. Breadth usually beats two specialists who must coordinate, while a staff title is useful only when the person can build without an established team or structure.

What seniority and experience level actually mean for this hire

For the generalist version of this role, the target is a senior engineer with several years of production experience and comfort across the full stack. Prototypes alone are not enough evidence of fit. Startup experience helps, but a senior engineer from a large company can fit if they have been the technical lead on a shipped product. A junior engineer, narrow specialist, or career switcher introduces mentoring or learning demands the founder may not have time to support.

Seniority earns its cost through independent shipping. At early user scale, the need is often someone who can ship five features this week, watch what users do, and ship five more next week. Eight-plus years of production experience can offer an advantage over three years spent still learning that compounds over months. Yet years alone do not establish independence. Find someone whose experience removes work from the founder without creating a new supervision job.

The traits become concrete when you ask what the person does without prompting. Ownership means finding something broken and fixing it, while product thinking means questioning whether a feature should exist before choosing how to build it. Comfort with ambiguity means asking questions that help make sense of the work and getting started before everything is clear. Speed with judgment means knowing where a shortcut is safe, and communication means explaining decisions, trade-offs, and blockers clearly in writing. If you remain uncertain whether the candidate has shipped alone, that uncertainty is the answer.

Compensation that is honest about where equity grants have landed in 2026

A founding engineer offer must reflect both stage and scope. Founders get this wrong in opposite directions, offering too little equity to protect dilution or relying on 2021-era grants that have since compressed. Neither approach starts with the actual job. An engineer taking broad ownership at pre-seed is accepting a different risk from the first engineer outside the founders at Series A. Compensation should make that difference visible.

CRV’s guide gives a reference point: the median equity grant for the first engineer is roughly 1.5% on a four-year vesting schedule. Pre-seed founding engineers may get more equity because they take on more risk and are less sure of the cash, while grants are lower at seed and lower still at Series A. The standard structure is a four-year vest with a one-year cliff and double-trigger acceleration. The available figures do not establish stage-specific numerical ranges, so a precise band for every stage would imply more certainty than the evidence supports.

Cash also changes with stage and location. Seed-stage first engineers in the US command a meaningful base salary, Series A first hires outside the founders command more, and remote roles land lower. If cash is below what the candidate can earn elsewhere, address the gap with more equity and clear stage-appropriate framing. Top candidates will do the equity math themselves. Walking through the cap table, dilution path, and scenario math makes the offer tangible rather than leaving the percentage to do all the selling.

Where the hire actually comes from and why cold job posts rarely close it

A clear role and credible offer still need a channel that reaches the right people. Top founding engineers are often passive candidates, so a cold job post is unlikely to reach them when they are ready to commit. The founder’s prior work network is the strongest starting point, including former colleagues, university classmates, and people who have seen the founder execute. Trust is part of this hiring decision too. Founders should therefore talk to people who can make referrals before posting the job description.

Patreon shows why it helps to know people before the need to hire becomes urgent. An existing relationship can give both sides a basis for trust that a new posting cannot supply. Warm referrals and deliberate outbound extend that search beyond former colleagues. Fit still needs to be evaluated once those conversations begin.

The June 2026 HN “Who is Hiring?” thread includes Hotwash, whose solo founder had 11 paying fire departments and zero churn, seeking a founding engineer to own engineering entirely. Column, started by a co-founder of Plaid, sought infrastructure, product, and backend engineers while keeping a small engineering team reporting to the CEO. Many of its engineers were hired through HN.

When the founder’s network is exhausted, structured external sourcing is the next lever. The role definition should remain anchored to the same bottleneck. Otherwise, changing the channel can become an excuse to broaden the search until almost any senior engineer looks plausible.

Running an interview process that a candidate with real options will actually finish

Strong candidates leave before they get an offer if the interview process is slow or poorly planned. They are considering other companies at the same time. A drawn-out loop therefore risks selecting candidates for their availability. The process should be short and test real work, how candidates make decisions on their own, and how they communicate, without making them repeat themselves in rounds that add nothing.

The founder screen takes about half an hour. Its job is to sell the problem, assess interest, and test whether the conversation is clear and useful. The practical coding session then takes one to one-and-a-half hours, replacing LeetCode with pair-programming on a realistic problem in the candidate’s preferred stack. This gives the technical assessment work that resembles the job. For a non-technical founder, a trusted technical advisor should observe the session.

System design takes about an hour and should begin with an ambiguous problem. The ambiguity is part of the test itself. It shows how the candidate asks questions, makes trade-offs, and explains their choices without a finished spec. For an AI-native startup, an eval harness exercise can test how they would measure whether an AI feature helps the user or just gives output that seems right. That tests product judgment more directly than algorithm recall.

Reference calls are required and should include at least two direct managers. A paid trial project lasting a few hours is useful if the decision remains genuinely uncertain. A candidate who has already supplied the evidence should not need another stage simply because the founder is reluctant to decide.

After the hire, the 60-day test has two parts. By the end of month two, the engineer should be shipping meaningful work with real customer impact and improving codebase quality with each PR. Volume alone is not the standard of success. If neither signal is present, revisit fit then, before month six.