Sources and research
Everything the paths say about the market and about specific companies, with links, dates and labels. Generated from docs/RESEARCH.md.
Researched 2026-11-19. Refresh with the /refresh-research command (see .claude/commands/refresh-research.md); it rewrites this file in the same shape.
Every claim carries a label:
- [official]: published by the employer itself (careers page, interview guide, a live job post).
- [data]: counted by us from a public job-board API on the date given. Reproducible with the method shown.
- [third-party]: a report, survey, prep vendor or candidate-sourced summary. Useful, not authoritative. Prep vendors sell courses; candidate reports are anecdotes.
The site never presents a guess as a real interview question. Where a lesson says "a loop typically includes", it links back to a row here.
1. Market signals
| # | Claim | Source | Date | Label |
|---|---|---|---|---|
| M1 | AI Engineer is the #1 fastest-growing US role on LinkedIn's Jobs on the Rise list, for the second year running. Forward-Deployed Engineers are named among 1.3M new AI-enabled jobs created globally in two years. | LinkedIn press release | 2026-01-14 | third-party (LinkedIn data) |
| M2 | Most common AI Engineer skills on LinkedIn: LangChain, RAG, PyTorch. MLOps skills (versioning, monitoring, cost, governance) read as table stakes. | Dice summary of LinkedIn list | 2026-01-14 | third-party |
| M3 | Tech postings mentioning AI were ~45% above Feb 2020 while total tech postings were ~34% below. 20%+ of software-development postings mention AI. | Indeed Hiring Lab | 2026-01-22 | third-party |
| M4 | Across 800k+ US tech postings (Jan 2025 to Mar 2026): Python 46%, SQL 45%, Java 21%, JavaScript 19%. SQL leads outside tech-sector employers. | Oxylabs via GlobeNewswire | 2026-06-04 | third-party (vendor research) |
| M5 | US cybersecurity postings Jan to Sep 2026 fell 26.4%, but AI Security Engineer demand rose 268%. ~15% of cyber postings reference AI. | SecurityWeek, CyberSN data | 2026-10-18 | third-party |
| M6 | AI/ML and cybersecurity tie as the hardest IT roles to fill (State of the CIO 2026); software engineering is #6. | CIO.com | 2026-06-15 | third-party (survey) |
| M7 | Snapshot of 224 open FDE roles at 118 companies (Palantir 51, OpenAI 31, Databricks 12). FDE skills: production Python and TypeScript, RAG, evals, agents, observability. | JobsByCulture | 2026-07-14 | third-party (job board; directional) |
| M8 | Claims of "+143% YoY AI Engineer postings" and "56% AI wage premium (PwC)". | HeroHunt.ai | 2026-03-28 | third-party (vendor; secondary citations, treat as weak) |
2. What Anthropic and OpenAI publish
2.1 Anthropic
| # | Claim | Source | Date | Label |
|---|---|---|---|---|
| A1 | Mission line used in every job post: "create reliable, interpretable, and steerable AI systems." | Any post, e.g. Forward Deployed Engineer | read 2026-11-19 | official |
| A2 | Candidate AI policy: draft applications yourself, then refine with Claude; take-homes without Claude unless told otherwise; live interviews are "all you", no AI assistance unless indicated; prep with Claude is encouraged; don't let Claude invent experiences. | Guidance on Candidates' AI Usage | page says last updated 2025-07-10 | official |
| A3 | Careers page lists principles including "Be helpful, honest, and harmless" and "Do the simple thing that works." | anthropic.com/careers | read 2026-11-19 | official |
| A4 | Staff+ Software Engineer, Backend: 8+ years; led multi-month projects as tech lead; large-scale backend systems in production; "influence technical decisions beyond your immediate team"; distributed systems, API design, cloud at scale; navigate ambiguity. Range listed $405k to $485k. | Greenhouse post 5174755008 | updated 2026-10-30 | official |
| A5 | Staff+ Software Engineer, Claude Managed Agents: 8+ years backend/distributed/infra; stateful, long-running systems; durability, consistency, failure modes, cost; "excited by 0 → 1 work"; uses Claude or other AI tools as a core part of building; on-call ownership. Nice to have: agent harnesses, eval infrastructure, durable execution, public APIs/SDKs. | Greenhouse post 5395767008 | updated 2026-08-21 | official |
| A6 | Applied AI Engineer, Enterprise: 6+ years as SWE, Solutions Architect, FDE or founder; owning the technical win (evals, POCs, security and architecture reviews); production LLM apps (prompting, context engineering, agent architectures, evaluation); C-level communication; enterprise security, compliance, data residency. | Greenhouse post 5390795008 | updated 2026-09-14 | official |
| A7 | Applied AI Architect, Enterprise Tech: 5+ years in customer-facing technical roles (Solutions Architect, Sales Engineer, TAM); scalable cloud architectures integrated with enterprise systems; "comfortable with python"; help customers build eval frameworks. Range listed $240k to $380k. | Greenhouse post 5383335008 | updated 2026-10-16 | official |
| A8 | Forward Deployed Engineer: works inside customer systems to ship production apps on Claude; delivers MCP servers, sub-agents and agent skills; autonomy and ambiguity. | Greenhouse post 5391021008 | updated 2026-08-28 | official |
| A9 | Third-party loop reports: recruiter screen, a 60 to 90 min technical screen (often a timed online assessment, Python strongly preferred), then roughly 4 to 5 rounds of ~55 min: coding, system design, hiring manager / project, and a values/culture round that guides call the hardest to pass. Reports differ on whether loops are split across two days. | Aced/Exponent, IGotAnOffer | undated pages, read 2026-11-19 | third-party (candidate-sourced) |
2.2 OpenAI
| # | Claim | Source | Date | Label |
|---|---|---|---|---|
| O1 | One interview guide for all roles. Stages: résumé review (~1 week), intro calls, skills-based assessment (pair coding, take-home or technical tests, varies by team), final interviews, decision within ~1 week. | openai.com/interview-guide | no date on page; read 2026-11-19 | official |
| O2 | Finals: "4–6 hours of final interviews with 4–6 people over 1–2 days", virtual by default. | same | same | official |
| O3 | Engineering bar: "well-designed solutions to the challenge, high-quality code, optimal performance, and good test coverage", plus communication and collaboration. | same | same | official |
| O4 | "We are not credential-driven"; looks for people who ramp quickly in a new domain and produce results. Values: collaboration, communication, openness to feedback, mission alignment. | same | same | official |
| O5 | The guide states no rule about AI use during interviews. Ask the recruiter. | same | same | official (absence noted) |
| O6 | Forward Deployed Engineer (Healthcare): 6+ years SWE / ML / solutions / consulting; own discovery, architecture, implementation, evaluation, productionization and handoff; success = production adoption and eval-backed launch criteria; up to 50% travel. | Ashby post | published 2026-08-06 | official |
| O7 | Applied AI Engineer: success measured by "production systems, sustained adoption, and meaningful customer impact—not simply activity or successful demonstrations"; decisions across models, agents, retrieval, tools, latency, cost, safety, governance. | Ashby post | published 2026-03-25 | official |
| O8 | Software Engineer, Codex Enterprise: deep backend and distributed systems; owns new products from ambiguity to launch; builds plugins, skills, hooks; works with FDEs; founder experience valued. | Ashby post | published 2026-01-09 | official |
| O9 | Third-party loop description: recruiter, hiring manager, coding, system design, a technical project deep-dive, behavioral. Coding topics reported: arrays/strings, hash maps, trees, graphs, DP, heaps. | Educative | 2026-10-26 | third-party (course vendor) |
2.3 Live job-board counts (our data)
Pulled 2026-11-19 from the public board APIs: boards-api.greenhouse.io/v1/boards/anthropic/jobs?content=true and api.ashbyhq.com/posting-api/job-board/openai. "Engineering" means titles with engineer, architect or developer, excluding data-center, IT, AV and sales titles. Keyword hits are regex matches on the post body, so they are approximate.
| Anthropic | OpenAI | |
|---|---|---|
| Open posts | 636 | 849 |
| Engineering/architect posts | 255 | 412 |
| Mentions Python | 53% | 45% |
| Mentions TypeScript | 12% | 11% |
| Mentions Rust | 19% | 12% |
| Mentions C# or .NET | 0% | 0% |
| Mentions Kubernetes | 18% | 9% |
| Mentions Azure | 9% | 5% |
| Mentions evals | 49% | 39% |
| Mentions agents | 42% | 40% |
| Mentions customers | 46% | 51% |
| Customer-facing AI titles | Applied AI Architect 37, Applied AI Engineer 12, FDE 8 (incl. managers) | "Applied AI" department 128, "Forward Deployed Engineering" department 40 |
| Titles naming Codex / ChatGPT / API / Infrastructure / Safety Systems | n/a | 22 / 15 / 14 / 51 / 2 |
[data] label. Counts move daily.
3. What this means for David's paths (re-check)
| Path (brief order) | Verdict | Change made |
|---|---|---|
| 1. AI Engineer | Holds. #1 on LinkedIn (M1); evals and agents appear in ~40 to 50% of lab engineering posts. | None. Track 3 (LLM & agents) leads. |
| 2. Forward Deployed / Applied AI Engineer | Holds and widens. Both labs hire heavily under Applied AI Architect as well (A7, §2.3). That title asks for 5+ yrs Solutions Architect experience and Python comfort, which is David's closest match. | Path renamed Forward Deployed / Applied AI Engineer & Architect; Applied AI Architect called out as the best-fit entry. |
| 3. Anthropic | Holds. Posts match the brief (A4 to A8). Live interviews forbid AI help (A2), so mock rounds are practised unaided. | Path covers Staff+ SWE and Applied AI (Engineer + Architect). |
| 4. OpenAI | Holds. Guide matches the brief (O1 to O4). Teams confirmed in titles: Codex, ChatGPT, API, infrastructure; safety systems is small (2 titles). | Practice set weights "test coverage" and project deep-dive. |
| 5. Staff Software Engineer | Holds, but general SWE demand is flat (M3, M6). | None. |
| 6. Principal Engineer / Enterprise Architect | Holds; strongest fit with David's background, smallest posting volume. | None. |
| 7. AI Platform / ML Infrastructure | Holds; infra is the largest engineering group at OpenAI by title (§2.3). | None. |
| 8. Security Engineer (AI and cloud) | Rising. AI Security Engineer +268% while cyber overall fell 26% (M5); tied hardest to fill (M6). | Kept at #8 for David's fit, but the path page notes it is the fastest-rising niche. |
| 9. Cloud Solutions Architect | Holds; overlaps with Applied AI Architect. | Path points to path 2 as the AI-era version. |
| 10. Engineering Manager / Head of Engineering | Holds; both labs list many EM roles (Anthropic ~110 manager titles, many non-engineering). | None. |
Language finding. C# and .NET appear in zero Anthropic or OpenAI engineering posts today (§2.3), and Python leads every source (M4). The coding track therefore shows Python first and C# second on every pattern. C# stays because enterprise loops (paths 5, 6, 9, 10) often let you choose, and it is David's strongest language.
4. Not verified / open questions
- Anthropic's careers page sections "What we are looking for" and "The interview process" did not render in a text fetch. Loop details here come from third-party reports (A9) and should be confirmed with the recruiter.
- OpenAI's interview guide carries no publication date.
- One source (FourWeekMBA, "developer jobs up 15%") failed to load and is not used.
- No source was found for Staff-level interview rubrics at either lab. Rubrics on this site are our own and are labelled as such.