AI Product Design is the hottest new interview topic in 2025-2026. Core areas: when to use AI (not every problem needs it), human-in-the-loop design patterns (when to let humans intervene), trust building (how to make users believe AI output), AI product challenges (hallucination, latency, cost), and AI product evaluation metrics.
Product Builder behavioral interviews differ from SWE — they don't just test teamwork, they specifically test how you drive things without formal authority. Core skills: influence narratives (how to convince engineers to build your feature), conflict resolution (disagreements with designers/engineers/stakeholders), vision expression (how to make someone understand your product direction in 30 seconds), and failure stories (learning from failure without deflecting blame).
Execution interviews test whether you can turn ideas into deliverables. Core skills: roadmap planning (how to prioritize with limited resources), priority defense (why A before B), cross-team collaboration (how to drive engineering and design), stakeholder management (how to handle conflicts), and the ability to track progress with data.
Growth interviews don't test whether you can growth hack — they test whether you have systematic growth thinking. Core skills: growth loop design (the acquisition → activation → retention → referral flywheel), experiment design (the full hypothesis → metric → experiment → analysis process), retention strategy (finding the aha moment, designing habit loops), and using data to decide what's worth continued investment.
Metrics interviews test whether you can make decisions with numbers, not how much statistics you know. Core skills: north star metric selection logic (why this one and not that one), metric tree decomposition (finding actionable levers), funnel analysis (which step's drop-off is most worth fixing), A/B testing design and pitfalls, and judgment when facing counterintuitive data.
A Product Builder isn't a traditional PM — you need to build from 0 to 1, not just write PRDs. Interviews test the intersection of product intuition, metrics thinking, technical understanding, and execution ability. Prep strategy: first figure out whether your target company wants a PM or a Builder, then allocate time across nine dimensions.
Product Design interviews don't test whether you can draw wireframes — they test how you go from problem to solution. Core skills: MVP scope judgment (what to build and what not to), trade-off analysis (speed vs completeness, generic vs custom), communication of design decisions (why A instead of B), and iterative thinking.
Product Sense interviews don't test how many features you can think of — they test whether you can find the problem truly worth solving within a vague requirement. Core skills: user segmentation thinking, problem reframing (turning 'add a feature' into 'what problem are we solving'), structured reasoning for feature prioritization, and the ability to hold or revise your judgment under follow-up questions.
Strategy interviews don't test whether you can recite frameworks — they test whether you can make judgments with incomplete information. Core skills: market sizing (the practical use of TAM/SAM/SOM, not rote numbers), competitive moat analysis (network effects, switching costs, brand), go/no-go decisions for new markets, and using elimination rather than addition for strategic trade-offs.
Technical PM interviews don't require you to write production code, but you need to be able to read trade-offs. Core skills: API design fundamentals (RESTful, versioning, error handling), high-level system architecture understanding (microservices, database selection, caching), collaboration patterns with engineers (RFC process, technical spec review), and making product decisions under technical constraints.
Agricultural spraying is the one drone application whose unit economics are fully transparent: the billing unit is the fen (about 970 m²), rates run NT$150–300, spraying one fen takes about two minutes, equipment costs NT$300–500k, and even the substitute's price is public (manual spraying, roughly NT$200 per fen). And precisely because anyone can run the arithmetic, everyone did — operators who entered in 2018 sprayed over a thousand hectares a year and cleared over a million; later entrants broke even after two or three years and quit. Licensed operators charge NT$300 per fen; unlicensed ones undercut to NT$150. This is the industry cycle in miniature, compressed into about six years.
I assumed inspection was blocked by beyond-visual-line-of-sight rules the way logistics is. It isn't. Bridge, transmission tower, and high-speed rail viaduct inspection are all running with hard numbers: one bridge went from 8 inspectors, 4 vehicles and 2 days to 5 people, 1 vehicle and half a day with no traffic control at all, at 60% of conventional cost; high-speed rail crews covered at most 700 metres a day on foot and now save 3–5x; a private power plant cut headcount by three quarters and cost by half with no outage required. The reason is that all of these are segmented, fixed-point tasks completable within visual line of sight. What BVLOS actually blocks is continuous long-range routes, not 'inspection' as a category.
I assumed Taiwan had no real drone logistics because it has no BVLOS framework. Wrong: the CAA has approved 10 cases across 24 flight corridors, and the Institute of Transportation has run a six-year PoC → PoS → PoB path since 2020, entering commercial validation in 2025. The way it routes around regulation is the mirror image of inspection — inspection cuts work down to within visual line of sight, logistics gets corridors approved one at a time. And its value isn't being cheaper than a boat: during the typhoon sailing suspensions at Liuqiu, a drone made the crossing in a bit over ten minutes, which is what you have when the boats don't run.
Agricultural spraying runs NT$150–300 per fen, computable to the decimal. Search and rescue has no such number, because a life recovered has no price. That difference determines two things: first, the specification is written by terrain rather than performance — the defining feature of Taiwan's fire agency drones is that they do NOT depend on GPS, because mountain signal drops and they must thread through trees; second, when the legislature moved to cut the budget, the ministry could only point to one man pulled from a flooded river in June. Defending a budget with an anecdote is fragile, and this application has no better weapon.
Hardware runs 35–55% gross margin under permanent DJI price pressure; autonomy software and DaaS subscriptions run 60–80% and recur. Skydio's software subscriptions were already ~30% of revenue in 2023 at a 38% blended margin; India's Garuda had DaaS at 62% of FY24 revenue with a 351-day cash conversion cycle against a defense-heavy peer's 597. Taiwan is almost entirely concentrated in the lowest-margin, most substitutable cell.
The unit of validation is an assumption, not an idea. Kohavi's data shows the industry median experiment success rate is ~10%, which means roughly 22% of 'winning' experiments at p<0.05 are false positives. Sean Ellis's 40% threshold has no publicly available dataset. AI product retention should be baselined at M3 rather than M0, and GRR splits from 23% below $50/mo to 70% above $250/mo.
A Product Builder runs the full loop from problem discovery to design to build, alone. The core difference from a PM: PMs influence execution through authority, Product Builders influence it by directly shipping working products. LinkedIn replaced its APM program with an Associate Product Builder track, and PayFit defined the role back in 2019.
Using my own 30+ RAG/Agent posts to audit the blog itself, I identified a prioritized improvement list spanning content quality, site tech, RAG design fixes, harness infrastructure, and AI agent applications — no phases, just priorities.
A breakdown of the three-layer digital ecosystem structure: LINE's super-app, Shopify App Store flywheel, and Taiwan MarTech integration strategies. The core mechanism is using APIs and data flows to create mutual dependency among participants, collectively reinforcing the moat.
MOOC completion rates hover at just 5–15%, and the problem isn't course quality — it's the execution gap. DaoDao positions itself as a 'Learning OS,' using public commitments, community interaction, and AI recommendations to make learning visible and sustainable.
DaoDao is not a content platform -- it's a learning connector. Using anti-perfectionism design, community co-learning, and zero-decision recommendations, it helps learners bridge the execution gap -- from vague ideas to actionable plans.
NobodyClimb uses RAG to tackle scattered climbing route information, ties quota limits to community engagement, and leverages Cloudflare Workers AI to bring inference costs close to zero.
The climbing community doesn't lack the will to share — it lacks a place to connect and preserve its culture.
To consolidate scattered notes and showcase diverse interests, I built a personal blog using Astro + Cloudflare Workers D1, paired with a Claude post skill for zero-friction writing.