Interview tomorrow · field guide · researched 21 July 2026

PM’ing the agent runtime.

A product-first primer on OpenAI’s Agents SDK, the contrasting thesis behind Vercel Eve, and the judgment required to build an agents platform that developers can trust in production.

8:18 VIDEO PRIMERPRODUCT STRATEGYARCHITECTURE20 INTERVIEW DRILLS
THE MODEL
DECIDES
THE RUNTIME
DELIVERS
The interview thesis: the agent-runtime PM owns the boundary between probabilistic intelligence and dependable product execution.
01 · Watch first

The eight-minute briefing.

The video is deliberately 40% architecture and 60% product judgment. Watch once end-to-end, then use the rest of this document to rehearse answers.

Agent SDK PM Interview Primer · 1280×720 delivery · 8 minutes 18 seconds · open the player directly
What to sound like in the interviewTechnically fluent enough to choose a boundary; product-minded enough to explain which user pain that boundary removes; disciplined enough to name the tradeoff and the metric.
02 · Mental model

The model call is the small part.

An agent is not a chatbot with a tool. It is an application that plans, acts through tools, maintains enough state to finish multi-step work, and sometimes transfers control or pauses for a human.

INPUT + AGENTinstructions · toolsCALL MODELResponses by defaultINSPECT ITEMSfinal · tool · handoffEXECUTE / PAUSEvalidate · approveFINISHor repeatTRACING · USAGE · STATE · GUARDRAILS ATTACH TO THE LIFECYCLE
01

Decision

The model proposes output items, including tool calls or a transfer.

02

Execution

The runtime validates, invokes, records, and feeds results back into the loop.

03

Operation

The product exposes state, cost, safety, traces, interruptions, and recovery.

PM reframingThe product is not “make the model more intelligent.” It is “make intelligent action faster to build, safer to run, and easier to improve.”
03 · OpenAI product

Responses when you own the loop. Agents SDK when the SDK should run it.

The official guide draws a clean boundary. The Responses API is the model-response primitive. The Agents SDK provides an agent-run abstraction and manages recurring orchestration such as repeated tool calls, branching, handoffs, sessions, guardrails, tracing, and resumable approvals. The application’s server still owns deployment, tool implementations, storage choices, and approval policy.

DimensionResponses APIAgents SDKPM implication
Core abstractionA model responseAn agent runThe SDK’s product surface is lifecycle semantics, not only request syntax.
LoopApplication implements itRunner performs itValue equals repeated orchestration removed without blocking escape hatches.
Multi-agentBuild routing/delegationAgents-as-tools and handoffsMake ownership explicit; do not reward topology complexity.
StateHistory, response chaining, ConversationsSame options plus SDK sessions and resumable stateState ownership must be understandable, portable, and debuggable.
SafetyTool approvals; broader controls are customInput/output/tool guardrails and approval flowsSafety is a workflow UX, not a boolean filter.
ObservabilityResponse objects and API logsTraces across calls, tools, agents, guardrails, handoffsTrace-to-fix is a first-class developer productivity outcome.

Product surfaces you should be able to explain

Agent + Runner

Agent is the reusable configuration: model, instructions, tools, output contract, policies, and specialists. Runner owns the loop and stop conditions through async, sync, or streamed entry points.

Tools + orchestration

Hosted tools, function tools, MCP, and agents-as-tools give the model action. Manager-style composition retains central ownership; handoffs transfer active ownership.

State + humans

Sessions manage conversation memory. Run state can serialize interruptions, collect approval/rejection, and resume. Long-lived work can use durable integrations.

Operate + improve

Tracing records lifecycle spans; usage aggregates requests and tokens; hooks expose events; evals turn observed failures into repeatable tests.

from agents import Agent, Runner, function_tool

@function_tool
def lookup_order(order_id: str) -> dict:
    return {"order_id": order_id, "status": "shipped"}

agent = Agent(
    name="Support specialist",
    instructions="Resolve order questions; ask approval before refunds.",
    tools=[lookup_order],
)

result = await Runner.run(agent, "Where is order 1042?")
print(result.final_output)

The visible code is small because the runtime owns schema handling, repeated model turns, result injection, stop conditions, tracing, and usage. That is the activation wedge—not the whole long-term value.

04 · Eve contrast

Eve turns project structure into product opinion.

Vercel Eve calls itself a filesystem-first framework for durable agents. Conventional locations describe capabilities before the system runs.

OpenAI Agents SDK

Code-first runtime primitives

  • Agent and Runner organize orchestration.
  • Typed application code in Python or TypeScript.
  • Integrates into storage, deployment, tools, and policy that the team owns.
  • Durability is available through explicit integrations and resumable state.
Vercel Eve · beta

Filesystem-first application harness

agent/
├── instructions.md
├── tools/
├── skills/
├── channels/
├── schedules/
├── sandbox/
└── subagents/
  • Discovery and compilation derive a runtime surface from files.
  • Workflow-backed sessions checkpoint and resume by default.
  • Channels, schedules, sandbox, skills, and evals share one project model.
QuestionAgents SDK thesisEve thesisInterview insight
Where is authoring?Objects and typed codeConventional files and foldersDiscoverability can be a product feature, not documentation work.
Where is durability?Resumable state + integrationsWorkflow checkpoints as a defaultDefaults shift activation, operations, and lock-in simultaneously.
How broad is scope?Runtime/orchestration SDKAgent application frameworkCategory strategy is a choice of abstraction boundary.
What is the risk?Users rebuild surrounding app scaffoldingOpinionated conventions constrain unusual systemsThe roadmap should make the common path paved and the boundary permeable.
Do not answer “which is better?”Answer: “For which segment, job, and operating model is each boundary better—and what migration path prevents today’s convenience from becoming tomorrow’s lock-in?”
05 · Users and jobs

Segment by maturity and responsibility.

Explorer / prototype builder

Job: prove an agent can complete a useful task today.

Needs: quickstart, sample tools, sensible defaults, immediate traces.

Failure: installation works, but the first real tool loop does not.

Application engineer — primary wedge

Job: ship a reliable workflow inside an existing product.

Needs: typed tools, stable state, streaming, approval UX, testability.

Failure: the demo cannot survive production data and failure modes.

AI platform team

Job: provide governed primitives to many internal teams.

Needs: policy, provider/model governance, reusable components, fleet metrics.

Failure: every app invents incompatible state, traces, and safety semantics.

Security + operations

Job: control risk and diagnose incidents without blocking innovation.

Needs: approvals, audit trails, redaction, budgets, retention controls.

Failure: observability itself leaks sensitive model or tool data.

A coherent progression is: fast first run → guided productionization → explicit control at scale. The platform should reveal complexity as the job requires it, not at installation time.

06 · Product principles

The boundary is the strategy.

01 · Paved road

Strong defaults with real escape hatches

Make the safe common path short. Let sophisticated teams replace models, storage, tracing processors, tool behavior, and infrastructure. An escape hatch that bypasses observability or changes semantics silently is not a good escape hatch.

02 · Semantics

Observable and consistent lifecycle contracts

A handoff, guardrail, approval, retry, and interruption should mean the same thing in code, traces, results, and evals. This reduces both cognitive load and support burden.

03 · Complexity

Multi-agent complexity must be earned

More agents increase calls, context, latency, cost, and evaluation surface. Encourage specialists when ownership or policy genuinely differs—not because a diagram looks sophisticated.

04 · Recovery

Make failure resumable, not merely visible

Observability answers “what happened?” A production runtime also helps answer “what can safely continue?” State snapshots, idempotency guidance, approval resumption, and durable execution belong to one recovery story.

05 · Learning

Turn traces into evals

The highest-leverage flywheel is production failure → understandable trace → reproducible eval → verified fix → safer rollout. The product should shorten every edge in that loop.

07 · Metrics

Count outcomes, then guard the outcome.

Candidate north starWeekly production tasks completed successfully through SDK-managed runs, with the intended product-quality and safety gates satisfied.
Activation
Median time to first successful tool-using run; quickstart completion; percent reaching a second capability such as sessions, guardrails, or tracing; first-week retained projects.
Adoption depth
Production projects; active agents/tools; workflows using state, approvals, or guardrails; share with trace review and eval coverage.
Reliability
Task completion; tool failure; max-turn exits; retry rate; resume success; stale-state conflicts; approval abandonment; handoff failure.
Efficiency
End-to-end latency; model calls, input/output tokens, cache efficiency, and cost per successful task—not cost per request.
Developer productivity
Trace-to-root-cause time; trace-to-verified-fix time; upgrade success; support tickets per production project; time to implement a new tool.
Safety + trust
Risky actions routed to approval; guardrail precision/recall; false-positive burden; unauthorized tool attempts prevented; percent of traces compliant with retention policy.

Anti-metrics: raw run count, raw tool-call count, and number of agents per workflow. Each can rise while user value, reliability, or margins fall.

08 · Roadmap judgment

Prioritize the graduation bottleneck.

For most platform products, the existential gap is not “can a developer make one demo?” It is “can that team graduate to production and continue improving?” Diagnose the leaky stage before proposing the feature.

A

Activation gap

Improve quickstarts, errors, templates, and the first successful tool loop.

B

Production gap

Improve state, approvals, traces, evals, privacy, and cost controls.

C

Scale gap

Improve governance, reuse, fleet operations, migration, and compatibility.

A practical scoring frame

USER
FREQUENCY
PAIN
SEVERITY
STRATEGIC
FIT
PLATFORM
LEVERAGE
DELIVERY
RISK
UNIQUE
RIGHT TO WIN
Default bet: a production-readiness cockpit that converts trace patterns into actionable failure classes and reusable eval cases—because it strengthens the learning loop without hiding execution.

Validation plan: cohort the funnel from first run to production; interview recently activated and stalled teams; conduct concierge trace reviews; prototype failure clustering; measure reduction in time-to-root-cause and time-to-verified-fix; watch for privacy and false-confidence risks.

09 · Safety and trust

Approval is a product flow, not a modal.

Agent safety lives across the lifecycle: what instructions and tools are available, what input reaches which specialist, what tool arguments are allowed, which actions require approval, how a paused run survives, what the approver sees, what resumes after rejection, and what gets logged.

LayerQuestionFailure modeProduct response
InputShould this request enter the workflow?Irrelevant or prohibited work consumes resourcesInput guardrail with clear scope and user recovery
ToolIs this action and argument safe now?Model proposes a risky or malformed callSchema validation, tool guardrail, approval, idempotency
OutputMay this result be returned?Final content violates policy or contractOutput validation and explicit failure behavior
HumanCan an accountable person make the decision?Opaque approval, abandonment, or stale contextSummarize intent, impact, evidence, reversibility, expiration
ObservabilityCan we debug without leaking data?Trace retention becomes a privacy incidentRedaction, sensitive-data controls, retention policy, export governance
Sharp interview pointGuardrail false positives are not merely model-quality errors. They are product friction that pushes developers to disable safety. Track them alongside unsafe-action prevention.
10 · Product cases

Use this answer operating system.

  1. Name the user and job. Avoid “developers” as the only segment.
  2. Identify the repeated failure or orchestration tax. Tie it to lifecycle evidence.
  3. State the principle and chosen abstraction boundary. Explain why the platform should own it.
  4. Propose the smallest leverage point and accepted tradeoff. Do not jump to a suite.
  5. Define success, guardrails, and learning plan. Include counter-metrics.

Case: “Should OpenAI build more multi-agent features?”

I would not start with topology. I would segment workflows where ownership, policy, or context genuinely differs, then measure whether current handoffs and agents-as-tools fail those jobs. If the pain is debugging and evaluation rather than expression, I would improve trace semantics and eval support before adding another orchestration primitive.

Case: “Should the SDK become more like Eve?”

Adopt the insight, not necessarily the scope. Eve shows that discoverable conventions and durable defaults reduce scaffolding. I would test optional project conventions, diagnostics, and production templates while preserving the SDK’s strength as a composable runtime inside existing systems.

Case: “How would you improve onboarding?”

Optimize for the first successful tool loop, not package installation. Instrument the funnel through API key, first agent, first tool schema, first tool execution, and first final output. Pair every step with an inspectable trace and an error that explains the next action. Then guide the second capability based on intent: state, approval, or evaluation.
11 · Interview drills

Questions and model answers.

1. Why should the Agents SDK exist if the Responses API already exists?
Answer shape · boundary → tax → value → guardrail

The Responses API is the right primitive when an application wants to own model interactions and orchestration. The SDK exists because tool loops, specialist routing, state, guardrails, approvals, tracing, and result handling repeat across applications. Standardizing that lifecycle reduces orchestration tax and creates consistent observability and safety. The guardrail is that the SDK must preserve application ownership and escape hatches; otherwise convenience becomes lock-in.

2. Who is the primary user?

I would choose the application engineer graduating an agent from prototype to production. Explorers create adoption volume, but the application engineer experiences the full differentiated pain: tool semantics, state, approvals, latency, cost, debugging, and testing. Platform and security teams are critical secondary users whose requirements shape enterprise readiness.

3. What is the hardest product problem in an agents SDK?

Choosing stable lifecycle semantics while models, tools, and workflows evolve quickly. A runtime has to make common patterns easy without hiding enough detail that failures become mysterious. The PM challenge is to define contracts—run, state, handoff, approval, trace—that remain understandable across local development, production, and evaluation.

4. How do you decide between manager-as-tools and handoffs?

Start with ownership. Use a manager when one agent should synthesize results and enforce central policy. Use a handoff when a specialist should own the continuing conversation. Then compare context needs, latency, policy boundaries, and evaluation complexity. Multi-agent is justified by real specialization, not by the number of tasks.

5. What does Eve teach OpenAI?

Eve demonstrates a coherent demand for filesystem discoverability, stronger conventions, integrated channels and schedules, sandboxing, and durability by default. The lesson is that framework-level opinion can reduce production scaffolding. The strategic question is which of those insights should become optional SDK conventions versus a broader application platform, because scope expansion can weaken composability.

6. What would you prioritize next?

Absent data, I would prioritize the demo-to-production gap: trace clarity, eval integration, resumable state, approval ergonomics, and cost controls. My concrete bet is a production-readiness cockpit that turns observed trace failures into reusable eval cases. I would validate the assumption with lifecycle cohorts and trace interviews before committing.

7. What is the north-star metric?

Weekly successful production tasks completed through SDK-managed runs, qualified by product-specific correctness and safety gates. I would pair it with activation, reliability, efficiency, developer-productivity, and trust metrics. Raw runs are an input, not value.

8. How would you measure developer experience?

Use behavior, not only surveys: time to first successful tool run; time to add a second capability; trace-to-root-cause; trace-to-verified-fix; upgrade success; support incidents per production project; and retention of activated projects. Qualitative studies explain where cognitive load appears.

9. How should pricing influence product design?

The runtime should optimize cost per successful task, not minimize calls in isolation. Product surfaces should expose call count, token use, cache behavior, latency, and retries at the run and step level. Budgets and stop conditions should be composable. Pricing must not create incentives to hide observability or discourage safety checks.

10. How do you handle provider breadth versus consistent semantics?

Define a small portable core, publish capability differences clearly, and make provider-specific features explicit rather than silently degrading. Measure portability on real workflows. Consistency is valuable only when it does not erase differentiated model or platform capabilities.

11. What is a bad roadmap signal?

Requests for more orchestration primitives without evidence that current users cannot express high-value workflows. The underlying problem may be debugging, evaluation, state, latency, documentation, or organizational ownership. Diagnose the lifecycle bottleneck before adding topology.

12. How would you build trust in human approvals?

Give the approver enough context to judge: intended action, arguments, expected impact, evidence, reversibility, and expiration. Preserve the run durably while waiting. Make reject/edit paths clear. Measure abandonment, decision latency, reversals, and unsafe bypasses.

13. What is the moat?

Not the loop by itself—it is reproducible. The moat is an integrated learning system: best-in-class model/tool capabilities, stable lifecycle semantics, production traces, eval feedback, safety, ecosystem adoption, and a migration path that compounds developer trust.

14. How do you prevent abstraction leakage?

You do not eliminate it; you make it legible. Expose typed run items, raw responses where needed, trace spans, explicit state choices, provider capability differences, and well-documented escape hatches. A good abstraction compresses the common case and reveals the exceptional case.

15. How would you evaluate an agent runtime release?

Run contract tests for lifecycle semantics, representative workflow evals, reliability and latency benchmarks, upgrade/migration tests, safety red-team cases, and trace usability studies. Roll out by project cohort with rollback and compare task success, error classes, cost, and trace-to-fix time.

16. When should the product be opinionated?

Be opinionated where repeated failure has a clear safe default and where inconsistency creates ecosystem cost—tracing conventions, error taxonomy, approval state, schema validation. Be flexible where user systems genuinely differ—storage, deployment, domain policy, provider choice, and business logic.

17. How do you think about open source?

Open source can accelerate trust, debugging, contributions, and ecosystem reach for a developer runtime. The product strategy should clearly separate portable client/runtime value from hosted advantages such as trace operations, managed tools, evaluation, and model integration—without making the open layer artificially weak.

18. What could make the SDK fail as a product?

Unstable semantics, confusing state ownership, poor failure diagnosis, provider behavior that silently diverges, safety controls developers disable, or a gap between quickstart success and production readiness. A broad API with low trust is less valuable than a smaller one with predictable behavior.

19. What would you do in your first 90 days?

First 30: map lifecycle cohorts, top failure classes, support data, trace journeys, and stakeholder constraints. Next 30: select one graduation bottleneck and prototype with design partners. Final 30: ship an instrumented beta, define the metric tree, document the contract, and build the trace-to-eval feedback loop.

20. Give the 30-second product pitch.
The Agents SDK helps teams turn model decisions into reliable multi-step execution. It runs the tool and handoff loop, manages state and approvals, and makes the full lifecycle traceable—so developers reach production faster without surrendering control of their application.
12 · Questions to ask the interviewer

Ask about boundaries and failure.

  1. Where do activated users most often fail between first successful run and production?
  2. Which lifecycle guarantees—state, handoff, approval, trace—are considered sacred contracts?
  3. How does the team decide what belongs in Responses, the Agents SDK, hosted platform surfaces, or ecosystem integrations?
  4. Which segment is currently the primary design center: application engineers, AI platform teams, or a broader developer audience?
  5. What do production traces reveal that support tickets and evals do not?
  6. Where does provider portability create value, and where does it constrain OpenAI-specific innovation?
  7. How does the team measure safety friction, especially guardrail false positives and approval abandonment?
  8. What would have to be true for the SDK to become more opinionated about durability, project structure, or deployment?
Best closing question“Which repeated failure mode do you most want the runtime to make impossible—or at least immediately obvious?”
13 · Source trail

Primary sources.

Research note: Eve is labeled beta in its repository. Both products evolve quickly; verify renamed surfaces before quoting exact APIs in a live interview.

Final memory hook

Reduce orchestration tax. Preserve control. Make failure obvious. Turn traces into learning.