Research Lens Catalog

Humans confuse "how it's always been done" with "how it must be." Separating convention from physics reveals which constraints are voluntary. Most innovations come from questioning what everyone else treats as fixed.

Different altitudes reveal different truths. A CTO sees value propositions; an engineer sees race conditions. Moving between levels catches leaky abstractions and hidden load-bearing decisions that only appear at specific zoom levels.

Systems fail at their dependencies, not at their core. The most dangerous dependencies are the ones you didn't know you had — upstream changes that cascade silently and downstream consumers you didn't design for.

Not all parameters matter equally. Understanding which knobs have outsized effect lets you focus energy where it counts. Cliff edges — where smooth functions suddenly become discontinuities — are where production incidents live.

Technology evolves in recognizable patterns: a new constraint becomes the bottleneck, a new approach removes it, the next constraint surfaces. Seeing the pattern lets you anticipate the next transition before it arrives.

First-order thinking asks "will this work?" Second-order asks "then what?" The most impactful consequences are usually 2–3 steps downstream, invisible to a linear read. Every solution creates new problems worth anticipating.

Without a map, you can't see the empty spaces. Positioning reveals over-crowded quadrants (diminishing returns) and empty ones (unexplored opportunities). The most interesting point on any landscape map is usually the gap.

The brain builds understanding through structural analogy. Mapping research to a domain you understand deeply lets you inherit that domain's intuitions — including solutions they've already discovered for structurally similar problems.

Every design choice is a bet about which tradeoffs matter. Making bets explicit reveals where the researcher's assumptions about priorities might diverge from yours. The tradeoff nobody discusses is often the most important one.

It's easier to explain failure in hindsight than predict it in foresight. Pre-mortems exploit this asymmetry: by imagining the project already failed, you unlock failure narratives that forward-looking risk analysis systematically misses.

Failure analysis is accidental; red-teaming is intentional. A dedicated adversary finds vulnerabilities that passive analysis misses because they're actively trying to break the system, not just cataloging what might go wrong on its own.

The most efficient way to learn is from others' mistakes. Every field develops common mistakes that look like best practices to newcomers. Cataloging anti-patterns saves months of dead ends and prevents reinventing known failures.

Every system operates within constraints it doesn't acknowledge. When an implicit constraint changes (hardware gets cheaper, regulation shifts, user behavior evolves), systems built around the old constraint become liabilities overnight.

Humans anchor on the status quo. Considering the opposite breaks the anchor. Often the inversion isn't viable — but the process of considering it reveals hidden assumptions and sometimes surfaces a genuinely better approach nobody tried.

Much of engineering is optimization within constraints. Relaxing constraints changes the landscape entirely — sometimes a 10x relaxation on one metric buys a 100x improvement on another. The constraint your users care about least is your biggest opportunity.

Individual ideas are well-explored; combinations are not. Most innovations are novel combinations of existing ideas, not genuinely new ones. The combinatorial space grows exponentially — most interesting pairs have never been tried.

The most impactful innovations are often transplants: an idea that's mundane in one field is revolutionary in another. Structural similarity between domains is usually deeper than it appears, and the solution techniques transfer surprisingly well.

"Is this good?" is a useless question. "Under what specific conditions is this the best choice?" is actionable. Decision trees force specificity about context, making recommendations precise instead of vague.

Behavior is not linear. Systems that work beautifully at small scale fail at large scale (and vice versa) because of phase transitions, resource contention, and emergent properties. Knowing where the transitions are is more valuable than steady-state performance.

Nothing reveals gaps like concrete specificity. A vague description of "improved workflow" can hide dozens of unsolved problems. Walking through minute-by-minute exposes them all and forces honesty about the actual experience.

The same system looks completely different to a user, a developer, an ops engineer, and a business owner. Designing for only one perspective creates a product that delights one stakeholder and frustrates the rest.

The gap between "I understand the concept" and "I can use this in production" is where most promising research dies. Mapping the learning journey reveals where adoption gets stuck and where better documentation or tooling is needed most.

Good ideas fail not because they're wrong but because the activation energy for adoption is too high. Understanding what resists change — switching costs, organizational antibodies, "good enough" incumbents — is as important as understanding the idea itself.

What a researcher doesn't say is often more revealing than what they do say. Gaps aren't oversights — they're usually conscious scope decisions. But each gap is both a critical limitation to understand and an opportunity for follow-on work.

Every field, culture, and individual has systematic blind spots. The most valuable insights come from perspectives structurally excluded from the standard discourse. An outsider's "naive" question often points directly at the field's biggest hidden assumption.

Research doesn't just provide answers — it expands the space of meaningful questions. The most valuable output of reading research is often not the answers it gives but the questions it makes possible to ask for the first time. These "horizon questions" are the seeds of the next wave of innovation, and they're invisible until you deliberately look for them.