Cursor’s $60B valuation dwarfs Sourcegraph’s $50M revenue

Seven Roles, Two Weeks, One Signal

Sourcegraph currently lists seven open roles across its career pages — a cluster that reads like a coordinated sprint, not routine backfill. Zero G Talent's first-party data shows one role added in the past seven days, with the latest postings including Senior Product Marketing Manager [IC4], Software Engineer - Platform [IC3], Agent Engineer [IC4], Software Engineer/Tech Lead - Code Plane [IC5], Security Engineer [IC3], and Regional Sales Director [M4], all tagged Remote. The IC3-through-IC5 and M4 designations reveal intent: Sourcegraph is hiring across the full seniority ladder, not padding junior headcount.

The timing matches the company's stated pivot toward agent-driven workflows. Sourcegraph's strategy page frames the problem as a "tidal wave of code" that agents cannot fully contextualize; each agent sees only fragments, rebuilding context per task. The Agent Engineer [IC4] sits at the center of that thesis. So does the Code Plane tech lead [IC5], a role that implies ownership of the infrastructure layer where SCIP-powered context gets assembled for MCP servers. Stripe's public adoption of the Sourcegraph MCP server for its internal agent fleet gives that layer a production proof point most startups lack.

All seven listings carry "United States" and "Remote" tags, consistent with Sourcegraph's all-remote model. The company's twice-yearly Merge onsites — recent venues include Amsterdam, Montreal, Cancun, and Rome — serve as the only mandatory colocation. Benefits include 100% company-paid medical, dental, and vision across the US, UK, and Canada; unlimited PTO with a 30-day minimum; a $25,000 family-planning benefit; and dedicated travel budgets for Merge gatherings.

The Greenhouse board also shows seven open roles, though the composition has shifted toward platform and agent engineering (roles that did not exist in the same form a year ago). Both listings suggest simultaneous hardening of the enterprise trust layer and expansion of the go-to-market motion. Sourcegraph already counts 200-plus enterprise engineering teams as customers, with SOC2 Type II and ISO27001 compliance, zero data retention, and dedicated account management as baseline expectations.

Six technical roles plus the evergreen Talent Community entry mean the screening funnel is about to widen. The next question is whether the bar moves with it.

Inside the Screen: What Sourcegraph Actually Looks For

Sourcegraph publishes its entire engineering interview process in a public handbook: a 30-minute recruiter screen, a 30-to-60-minute hiring-manager screen, a 60-minute resume deep dive, two technical interviews of 45 to 60 minutes each, a 60-minute cross-functional session with a product manager and a designer, a 30-minute values interview, a 30-minute leadership interview, reference checks, then an offer. The structure is deliberate. Co-founder and CTO Beyang Liu designed the loop so candidates focus on an area they know well, simulating a technical discussion between future peers and using that dialogue as a proxy for programming aptitude, all while keeping the loop time-efficient for both sides.

The technical bar varies by discipline. General engineering tracks run architecture reviews, code walkthroughs, pairing exercises, and live-coding sessions. Security candidates face an application-security walkthrough and a security-operations case study. AI and ML roles add a hiring-manager screen on ML breadth, a technical background conversation with Liu, a technical deep dive, a pairing exercise, and an AI coding exercise that tests problem selection, system design, metric choice, large-dataset handling, and scalable implementation. Leadership candidates interview with Head of Engineering Steve Yegge, then with Liu, followed by a peer interview, a technical interview, and cross-functional collaboration. The handbook lists each format by name so applicants can prepare for the exact session they will face.

Across every track, Sourcegraph flags four cultural signals it treats as non-negotiable. Engineers must be team-oriented. They must communicate clearly in an asynchronous, fully remote environment where much of the day involves reading code they did not write and explaining domain expertise to colleagues they rarely meet live. They must tailor technical depth to the audience: a product manager needs a different explanation than a senior platform engineer. And they must teach effectively; the handbook explicitly calls out "can effectively teach and share their knowledge" as a hiring criterion. Security roles add two more: confidence identifying weaknesses from varied sources, and the pragmatism to advocate for better practices without grinding product velocity to a halt.

A 2022 Blind thread asking whether Go knowledge is mandatory for backend roles suggests the company still weighs language-specific fluency, though the handbook frames technical interviews around systems thinking rather than syntax quizzes.

Glassdoor reviews consistently praise the public documentation — sample questions, expectations, even the interview rubric are posted — but several reviewers note a drop-off in follow-up communication after the final round. That friction point matters: a candidate who clears every technical and cultural gate can still sit in limbo while references are checked or headcount is reconfirmed. The handbook's stated goal of a "time-efficient process for yourself and Sourcegraph" does not always survive the last mile.

For the seven roles currently open, the screen will weight different slices of this rubric. The Agent Engineer and Code Plane lead will face the AI/ML track plus the standard architecture and pairing rounds. The Security Engineer walks through the application-security walkthrough and the operations case study. Every candidate, regardless of level, hits the values interview and the leadership interview. The company's bet is that the same cultural signals — communication, teaching, async fluency — predict success across product lines, while the technical formats flex to the domain.

Candidate Reaction: How Applicants Are Preparing

The split between Sourcegraph and Amp has created a preparation trap. Blind threads from mid-2024 show candidates prepping for the wrong company because nearly every write-up published before the split treats them as a single entity. Candidates studying Amp's agentic-coding workflows arrive at Sourcegraph screens expecting prompts about autonomous agents; the actual evaluation targets code-intelligence infrastructure, distributed systems, and Go fluency. The mismatch wastes weeks.

LeetCode has become the universal tax. Blind's "sourcegraph interview" query returns 106 results as of June 1, 2026, and forum threads show candidates grinding 150–200 problems while simultaneously building study groups. One poster reported posting for a study buddy and fielding dozens of replies. Another described a regimen that started in October 2024: "consistent with the planning, efforts, applying and studying." A frontend engineer with 12 years' experience wrote in May 2024: "I've been a frontend engineer for 12 years and never needed to do LeetCode. But in this job market, I had to bite the bullet."

Referral hunting runs parallel. A November 2022 post read: "Seeking Sourcegraph referral... Can anyone from Sourcegraph provide referral to this poor soul?" By mid-2024 the language had hardened: "Grinding LC, Started Study Group, Need Referrals." Candidates treat internal connections as a required checkpoint, not a nice-to-have.

Format preference splits the crowd. An August 2024 poll asked: "If you had an onsite interview and you had the option to do a 4 hour take home assessment or do two interviews of system design plus leetcode, which would you rather do?" The code-pairing round, specifically called out in an August 2024 thread ("Anyone know what to expect in the code pairing interview for sourcegraph?"), has become a focal study target because it tests real-time collaboration on unfamiliar codebases, not algorithm recall.

Public documentation earns rare praise. Glassdoor reviewers noted "detailed and publicly available documentation about the interview process, which included sample questions and expectations"; a transparency that candidates convert into mock-interview scripts. Yet communication gaps undercut the goodwill. Glassdoor summaries confirm "a positive atmosphere during interviews but noted a lack of follow-up communication." A January 2023 applicant for a Product Designer role waited three weeks without a reply despite emailing recruiting. A Reddit poster described being ghosted after multiple rounds: "The last interview I had with them was pretty conversational."

International candidates face an extra layer. A Sri Lanka-based engineer in July 2021 asked: "I have never faced an interview of an international company. So, what should I expect in my first interview?" By 2022, a Bangalore-based backend engineer with nine years' experience was still asking: "Is knowledge of Go mandatory?" The answer, reinforced by current role specs listing Go for Platform and Code Plane tracks, is effectively yes.

The net effect: a preparation ecosystem that looks more like a certification pipeline than a job search. Candidates chain LeetCode, system-design templates, Go refresher courses, and referral outreach into a months-long sprint, all for seven open roles.

From Hires to Product: Linking Talent to the AI Code‑Intelligence Roadmap

The six roles Sourcegraph added recently read like a product roadmap rendered in headcount. Each maps to a specific surface the company has signaled it will expand in 2026.

The Agent Engineer role sits at the center of the most visible push. Sourcegraph's February 2026 blog announcing version 7.0 framed the release as "the beginning of a new chapter," positioning the platform as "the shared intelligence layer for both developers and AI agents." The same post described Agentic Batch Changes as "AI agent for large-scale code changes. Migrate, modernize, and remediate across every repository with full control." That feature, and the MCP Server that feeds agents "complete, SCIP-powered context to produce reliable results with fewer retries and lower inference spend," need engineers who understand agent orchestration, context retrieval, and the failure modes of LLMs operating on million-file codebases. The IC4 Agent Engineer hire is a direct bet on that surface.

That IC5 lead targets the layer beneath. Sourcegraph's engineering surface spans code intelligence (search, navigation, symbol resolution), platform infrastructure, Cody AI, enterprise product, and developer-experience tooling. The Code Plane is the substrate that makes symbol resolution deterministic across millions of repositories, the "deep code-understanding across millions of repositories" that the company argues is its durable differentiation. A tech lead at IC5 signals an intention to harden that substrate for higher query throughput, lower latency, and broader language coverage as enterprise indexes grow.

Software Engineer - Platform [IC3] and Security Engineer [IC3] reinforce the enterprise trust chain. Sourcegraph lists SOC2 Type II, ISO27001, zero data retention, and dedicated support as baseline for its customer base: Reddit, MathWorks, Leidos, Indeed, Canva, MongoDB, Dropbox, HubSpot, Stripe, Blackstone, Midjourney, Uber, Dropbox, Yelp, large banks, Palo Alto Networks. Stripe uses the MCP server to connect internal agents ("Minions") to internal docs, ticket details, build statuses, and code intelligence via Sourcegraph search. Rakuten reported 80-plus hours saved onboarding new developers in their first month. Sourcegraph's website reports CERN indexes 15 million-plus lines of code. Each new enterprise deployment expands the attack surface and the compliance burden. The security hire is not incidental; it is a prerequisite for the next tier of regulated customers.

The Senior Product Marketing Manager [IC4] and Regional Sales Director [M4] complete the loop. Sourcegraph 7.0's messaging — "If you're evaluating how to enable effective agent workflows at enterprise scale, Sourcegraph 7.0 provides the authoritative intelligence layer that both your developers and your AI agents can depend on" — requires a go-to-market motion that translates technical depth into buyer language. The marketing role will shape how the "intelligence layer" narrative lands with CTOs and VPs of Engineering. The sales director will carry it into territories where the company's current footprint is thin.

Sourcegraph has said it ships on a weekly cadence for Sourcegraph Cloud "so improvements reach you faster." The structure "evolves based on product priorities more than traditional reorgs." That cadence only holds if the hiring pipeline feeds the right specialties at the right time. The current cohort — heavy on agent systems, code-plane infrastructure, and enterprise hardening — suggests the next two quarters will deepen the intelligence layer rather than broaden the feature set. Cody, already called "the fastest-growing part of the company," will absorb the agent work. The Code Plane will absorb the scale. The security and compliance work will unblock the next wave of regulated deals. The marketing and sales hires will measure whether the market buys the "shared intelligence layer" thesis at the price Sourcegraph needs to command.

Market Context: How Sourcegraph's Talent War Stacks Up Against Peers

The AI coding assistant market has consolidated around three poles. GitHub Copilot sits at the center with 20 million users, deployment in 90 percent of the Fortune 100, and $2 billion-plus in annual recurring revenue. Cursor, through its parent Anysphere, just commanded a $60 billion valuation in an all-stock acquisition by SpaceX announced June 16, 2026 (four days after SpaceX's own IPO) on $2 billion in revenue. Sourcegraph occupies the third pole: $50 million in revenue as of March 2025, 800,000 developers served, 54 billion lines of code indexed, and 200-plus enterprise engineering teams on contract.

The competitive dynamics explain why Sourcegraph's seven open roles read the way they do. Copilot's moat is distribution: it ships inside Visual Studio Code, the editor most developers already use. Cursor's moat is the editor itself; Anysphere built a fork of VS Code that bakes AI into every interaction. Sourcegraph's moat is the code graph. The company spent a decade indexing multi-repository, multi-host codebases at enterprise scale before layering Cody on top. That history shapes its hiring: the Code Plane and Agent Engineer titles signal investment in the retrieval and agent infrastructure that differentiates Cody from single-model assistants.

Copilot relies on OpenAI's GPT family. Cody defaults to Anthropic's Claude 3 but supports Mistral, GPT, Gemini, and Llama, a model-agnostic architecture that enterprise buyers treat as insurance against vendor lock-in. That flexibility demands engineers who understand LLM routing, context assembly, and evaluation across model families. The Agent Engineer role maps directly to that requirement. The Code Plane tech lead maps to it as well: the code graph is the retrieval layer that makes multi-model Cody work at enterprise scale.

Security and compliance appear in the role list for the same reason. Sourcegraph advertises SOC 2 Type II, ISO 27001, zero data retention, and SSO/SCIM/RBAC authentication. Copilot has faced scrutiny over training-data provenance and IP risk; Cody's codebase-grounded approach reduces hallucination and copyright exposure. The Security Engineer [IC3] hire reinforces that enterprise trust motion.

The sales role reveals the go-to-market gap. Copilot sells bottom-up: individual developers swipe a credit card, then IT ratifies. Sourcegraph sells top-down: dedicated account managers, support engineers, and multi-year contracts. A Regional Sales Director [M4] suggests the company is still building the field motion to convert its technical lead into revenue at Copilot's velocity.

Cursor's $60 billion exit changes the calculus for every engineer in this space. Anysphere proved a code editor can become a generational company in three years. Sourcegraph's split into Sourcegraph and Amp, noted for 2026, hints at a similar structural bet: separate the platform from the application, let each raise and hire on its own logic. The current hiring wave may be the last before that separation formalizes.

Talent follows the technical problem. Engineers who want to scale a single-model autocomplete to millions join GitHub. Engineers who want to build the editor of the future joined Anysphere. Engineers who want to solve retrieval-augmented generation across 54 billion lines of fragmented enterprise code (with model freedom, audit trails, and CI integration still missing from every vendor) are the ones Sourcegraph is screening for now. The bar is high because the problem is narrower, deeper, and less forgiving of shortcuts. The seven roles posted in two weeks are the clearest signal yet: Sourcegraph is not chasing the autocomplete market. It is staffing the infrastructure that makes agents trustworthy at enterprise scale.

Working in frontier tech? Zero G Talent tracks the openings: see every open Sourcegraph role, browse frontier tech jobs, the companies hiring, and the people building the field.