Before You Commit: The Questions That Make Science Worth Building
A field guide for deciding which scientific possibilities deserve time, capital, craft, and cultural attention—before momentum hardens into inevitability.
Camila ReyesTravel & longformFirst published 9/15/2026 · last revised 9/16/2026 with fresh sources, corrections, and new context. Reader corrections are reviewed and folded into future versions.
Summary
Scientific possibility is expanding faster than our capacity to develop it wisely. Gene editing, generative biology, advanced materials, neurotechnology, climate intervention, and autonomous laboratories can all produce dazzling demonstrations. Yet technical novelty is not the same as durable value. Before committing years, capital, talent, or public trust, builders need a better brief: What human or planetary condition improves? Why is this approach newly possible? Who benefits, who bears the risk, and what evidence would justify scaling? This explainer offers a practical framework for interrogating science before it becomes a product, platform, institution, or cultural fact. The aim is not caution for its own sake. It is higher-quality ambition: choosing work whose usefulness, beauty, governance, and consequences have been considered as seriously as its mechanism.
Key takeaways
- Begin with the condition to improve, not the technology to deploy. A crisp problem statement prevents an impressive capability from searching indefinitely for a market.
- Ask why now. A viable scientific venture usually depends on a specific shift in cost, instrumentation, computation, regulation, supply, or public behavior.
- Separate proof of mechanism, proof of performance, proof of use, and proof of value. A laboratory result is only the first layer.
- Map beneficiaries, risk bearers, missing voices, and potential misuse before scale makes the architecture difficult to change.
- Treat regulation, manufacturing, maintenance, interfaces, and public legitimacy as design materials—not administrative obstacles.
- Define stopping rules and disconfirming evidence in advance. Intellectual honesty becomes harder after identity and capital attach to a project.
- Look for asymmetric value: work where modest technical progress can create large social, ecological, or creative gains.
- The strongest scientific products make complexity legible without falsifying it, pairing technical depth with excellent interaction and institutional design.
Explain like I'm 5
Imagine you discover how to build a very fast boat. Before making a thousand boats, you should ask: Does anyone need to cross this water? Is the water safe? Can people steer the boat? Who repairs it? Could it damage the shore? Is a bridge simpler? Science tells us what may be possible. Building wisely means checking whether the possibility solves a real problem, works outside a perfect test, helps the intended people, and remains acceptable when many others use it. The best question is not merely, ‘Can we make it?’ It is, ‘What becomes better if we do—and how will we know?’
Deep dive
Start with the changed world, not the clever object
A seductive prototype can reverse the natural order of inquiry. Once a team can grow leather without animals, read neural signals, or generate candidate proteins, every conversation begins with the capability. Instead, describe the world after success. Whose day changes? Which cost, delay, hazard, or indignity disappears? What new behavior becomes possible? Use a counterfactual: if the invention vanished tomorrow, who would notice and what would they lose? This distinguishes an essential intervention from a novelty sustained by attention. It also reveals substitutes. A new diagnostic may compete not only with another test but with vaccination, workflow redesign, or better access to an existing clinic. Builders should seek problem–mechanism fit before product–market fit: evidence that the scientific mechanism addresses the actual causal structure of the problem.
Interrogate the ‘why now’
Breakthroughs rarely arrive from discovery alone. They emerge when several curves cross: sequencing gets cheaper, sensors smaller, models more capable, manufacturing more precise, or regulation clearer. CRISPR became transformative after the 2012 demonstration of programmable Cas9 editing, but its translation also required delivery systems, genomic data, clinical infrastructure, and regulatory pathways. Name the enabling stack and test each dependency. Is a critical reagent scarce? Does the model require proprietary data? Can production move from milligrams to tonnes without changing performance? A credible ‘why now’ is specific enough to fail. It identifies the bottleneck that has fallen—and the ones still standing.
Climb the ladder of evidence
Evidence changes meaning as science leaves the laboratory. Proof of mechanism shows that an effect can occur. Proof of performance shows repeatability under defined conditions. Proof of use asks whether people can integrate it into real routines. Proof of value compares outcomes, costs, and harms against current practice. Finally, proof of system fit tests infrastructure, incentives, regulation, and maintenance. Teams often leap from mechanism to market narrative. Resist this compression. Predefine metrics, baselines, sample sizes, and failure thresholds. Replicate decisive findings independently where possible. For AI-enabled science, separate benchmark performance from prospective performance on unfamiliar data. For biomaterials, test aging, repair, toxicity, and end-of-life—not merely the pristine sample photographed at launch.
Design the whole consequence chain
Every scientific product distributes benefits and burdens. List direct users, non-users affected by deployment, workers handling production, communities near extraction or disposal, future maintainers, and plausible malicious actors. Then trace first-, second-, and third-order effects. A cheap sensor may improve diagnosis, increase surveillance, and generate an attractive insurance dataset. Cultivated ingredients may reduce land pressure while concentrating intellectual property and energy demand. This is not a reason to freeze. It is a design prompt. Privacy-preserving architecture, repairable hardware, tiered access, benefit-sharing, reversible pilots, and secure defaults can be built early. Values become expensive retrofits once scale, contracts, and habits settle.
Treat adoption as a cultural and aesthetic problem
People encounter science through forms: a pill, dashboard, clinic, package, story, consent screen, or public landscape. These forms communicate authority and shape behavior. Good design makes uncertainty visible without making a product unusable. It tells a clinician when an algorithm is outside its validated domain; it gives a consumer meaningful provenance rather than decorative green claims; it makes a laboratory instrument calm, maintainable, and hard to misuse. Cultural context matters equally. Communities may reject a technically sound intervention because institutions have broken trust, benefits feel remote, or the language of progress erases local knowledge. Legitimacy is not marketing. It is earned through participation, transparency, recourse, and fair distribution.
Create a commitment gate
Before a major investment, write a one-page commitment case. State the condition to improve, the mechanism, enabling shift, strongest evidence, unresolved uncertainty, affected groups, pathway to scale, governance model, and measurable definition of success. Add three kill criteria: findings that would make the team stop, narrow, or redirect the project. Then stage commitment through bounded experiments. Fund the next uncertainty, not the grandest story. A useful portfolio balances near-term validation with long-horizon options and includes budget for safety, social research, design, and replication. The result is not timid science. It is disciplined imagination—the ability to pursue radical possibilities without surrendering judgment to novelty, sunk cost, or inevitability.
- 1945Vannevar Bush publishes ‘Science, the Endless Frontier,’ shaping the postwar model of publicly funded basic research while linking scientific capacity to national welfare.
- 1975The Asilomar Conference establishes an influential model of scientist-led deliberation on recombinant DNA risks, containment, and research norms.
- 1986The U.S. Office of Technology Assessment publishes work on technology assessment methods; the agency serves Congress until 1995, illustrating the value—and fragility—of institutional foresight.
- 2003The Human Genome Project announces completion after 13 years, with roughly $3 billion in U.S. public investment and a commitment to rapid data release.
- 2012Jennifer Doudna, Emmanuelle Charpentier, and colleagues demonstrate RNA-guided Cas9 genome editing, turning a microbial mechanism into a programmable tool.
- 2018He Jiankui announces the birth of CRISPR-edited babies, provoking international condemnation and exposing gaps between technical capability, governance, and consent.
- 2021WHO issues its global framework for governance of human genome editing, emphasizing registries, oversight, international cooperation, and public engagement.
- 2022AlphaFold Protein Structure Database expands to more than 200 million predicted structures, showing how open scientific infrastructure can reshape discovery workflows.
- 2023The U.S. FDA approves Casgevy, the first CRISPR-based therapy approved in the United States, demonstrating both the clinical power and delivery complexity of gene editing.
Glossary
- Problem–mechanism fit
- Evidence that a scientific mechanism addresses the causal structure of a meaningful problem, before a specific product or market is assumed.
- Translational research
- Work that moves findings from basic discovery toward practical applications, often across preclinical, clinical, engineering, and implementation stages.
- Technology readiness level (TRL)
- A nine-level scale, developed by NASA, used to describe maturity from basic principles at TRL 1 to proven operational systems at TRL 9.
- Externality
- A cost or benefit imposed on people or environments outside the immediate transaction, such as pollution, surveillance, or shared knowledge.
- Dual use
- Research or technology that can support beneficial purposes but could also be misapplied to cause harm.
- Reversibility
- The degree to which a deployment can be stopped, withdrawn, or restored without persistent damage.
- Precautionary principle
- An approach supporting protective action when plausible serious harm exists even if scientific certainty is incomplete.
- Responsible Research and Innovation (RRI)
- A framework emphasizing anticipation, inclusion, reflection, and responsiveness throughout research and innovation.
- Kill criterion
- A predefined result or threshold that triggers termination, narrowing, or redesign of a project.
FAQs
What is the first question to ask before funding scientific work?+
Ask which measurable human, ecological, or productive condition should improve. Then verify that the proposed mechanism addresses a cause of that condition rather than an attractive symptom.
How is this different from product–market fit?+
Product–market fit concerns demand for a product. Earlier-stage science needs problem–mechanism fit: confidence that the underlying effect is real, relevant, and capable of changing the target outcome.
How much evidence is enough to commit?+
Match evidence to the irreversibility and potential harm of the decision. A reversible software pilot may justify modest evidence; a germline intervention or environmental release requires far stronger validation, oversight, and consent.
Should founders avoid fields with unclear regulation?+
Not automatically. Regulatory ambiguity can signal opportunity, but the venture should have a credible pathway to safety, classification, trials, and compliance. A business model that works only without oversight is structurally weak.
How can a non-scientist evaluate a technical claim?+
Ask for the baseline, sample size, uncertainty, replication status, operating conditions, known failure modes, and comparison with current practice. Consult independent domain experts who have no financial stake.
When should ethics enter the process?+
At problem selection and system design, not after launch. Data rights, access, misuse, labor conditions, and environmental effects often determine architecture and cannot be repaired cheaply later.
What makes a scientific project culturally legitimate?+
Legitimacy grows from transparent evidence, meaningful participation, fair distribution of benefits and burdens, accountable institutions, and accessible routes for challenge or remedy.
How should teams handle uncertainty publicly?+
State what is known, unknown, assumed, and being tested. Use ranges and scenarios where appropriate. Precision theater may win attention briefly, but calibrated candor builds durable trust.
Predictions
- By 2030, scientific due diligence will increasingly combine technical review with security, lifecycle, regulatory, and social-impact audits before major funding rounds.
- Autonomous laboratories will make experiments cheaper and faster, shifting competitive advantage from raw execution toward question selection, proprietary measurement, and experimental taste.
- Regulators will demand more prospective validation of AI systems used in discovery and clinical settings, especially when models encounter changing populations or instruments.
- Product provenance will become an interface layer: materials, biological inputs, energy use, model lineage, and chain of custody will be exposed through machine-readable records.
- Reversible deployment will become a premium design feature in climate technology, synthetic biology, and urban sensing, allowing pilots to earn legitimacy before expansion.
- Public-benefit infrastructure—shared datasets, reference materials, standards, and biobanks—will become one of the highest-leverage categories for mission-driven capital.
Risks
- Capability pull: teams build around what a laboratory or model can do rather than what society meaningfully needs.
- Evidence inflation: preliminary, benchmark, or animal results are narrated as if clinical, industrial, or ecological performance were established.
- Scale discontinuity: a process that works in microlitres or controlled facilities fails economically, safely, or consistently at commercial volume.
- Governance debt: privacy, consent, security, access, and accountability are postponed until contracts and architecture make change costly.
- Dual-use exposure: tools for synthesis, automation, sensing, or manipulation lower barriers for malicious or negligent actors.
- Distributional harm: benefits accrue to wealthy customers while environmental, labor, or surveillance burdens fall on less powerful communities.
- Narrative lock-in: publicity, sunk capital, and founder identity suppress disconfirming evidence and prevent an intelligent pivot.
Opportunities
- Build decision-support platforms that translate scientific evidence into explicit assumptions, uncertainty ranges, readiness levels, and stage-gate recommendations.
- Create independent replication and validation services for biotech, materials, climate, robotics, and AI-for-science ventures.
- Design scientific instruments and interfaces around maintenance, graceful failure, accessibility, and uncertainty—not just peak performance.
- Develop provenance systems for biomaterials, engineered organisms, laboratory data, and AI-generated scientific outputs.
- Offer governance-as-a-product: consent infrastructure, secure access controls, incident reporting, model monitoring, and benefit-sharing workflows.
- Finance enabling commons such as curated datasets, standards, reference samples, open protocols, and interoperable laboratory software.
- Pair scientists with artists, anthropologists, and service designers to prototype how emerging technologies will feel, signify, and function in public life.
- Create venture studios focused on neglected but measurable needs where modest scientific advances can unlock outsized value, including diagnostics, water quality, cooling, and repairable assistive technology.
| Pressure | Opening | |
|---|---|---|
| #1 | Capability pull: teams build around what a laboratory or model can do rather than what society meaningfully needs. | Build decision-support platforms that translate scientific evidence into explicit assumptions, uncertainty ranges, readiness levels, and stage-gate recommendations. |
| #2 | Evidence inflation: preliminary, benchmark, or animal results are narrated as if clinical, industrial, or ecological performance were established. | Create independent replication and validation services for biotech, materials, climate, robotics, and AI-for-science ventures. |
| #3 | Scale discontinuity: a process that works in microlitres or controlled facilities fails economically, safely, or consistently at commercial volume. | Design scientific instruments and interfaces around maintenance, graceful failure, accessibility, and uncertainty—not just peak performance. |
| #4 | Governance debt: privacy, consent, security, access, and accountability are postponed until contracts and architecture make change costly. | Develop provenance systems for biomaterials, engineered organisms, laboratory data, and AI-generated scientific outputs. |
| #5 | Dual-use exposure: tools for synthesis, automation, sensing, or manipulation lower barriers for malicious or negligent actors. | Offer governance-as-a-product: consent infrastructure, secure access controls, incident reporting, model monitoring, and benefit-sharing workflows. |
For professionals
For founders, investors, design leaders, and R&D teams, use a 90-minute pre-commitment review before approving a major build. Score ten dimensions from 1 to 5: problem severity, mechanism relevance, evidence quality, why-now strength, manufacturability, unit economics, adoption fit, governance readiness, reversibility, and distributional fairness. No average score should conceal a critical weakness: any score of 1 requires a mitigation plan before expansion. Assign an owner and dated test to the three largest uncertainties. Define one success metric, one safety metric, one equity metric, and three kill criteria. Invite at least one independent technical skeptic and one affected stakeholder who is not a customer or funder. Repeat the review at each transition—prototype, field pilot, regulated trial, first production, and geographic scale-up. The output should be a living commitment case, not a ceremonial memo. Its purpose is to preserve the team’s freedom to learn before capital, reputation, and infrastructure make learning politically difficult.
Sources & references
- Science, the Endless Frontier — National Science Foundation
- WHO Global Guidance Framework for Human Genome Editing
- Framework for Responsible Innovation — UK Research and Innovation
- Technology Readiness Level Definitions — NASA
- Reproducibility and Replicability in Science — National Academies Press
- Recommendation on the Ethics of Artificial Intelligence — UNESCO
- Casgevy Approval Information — U.S. Food and Drug Administration
- AlphaFold Protein Structure Database — EMBL-EBI
A field guide to the studios turning cells, fungi, microbes, and living systems into cultural experiments—and potentially the next generation of materials, products, and ventures.
A design-led field guide to turning waste streams, living systems, and circular chemistry into objects people desire—and businesses built for a resource-constrained century.
A field guide for deciding when evidence deserves belief, when uncertainty deserves patience, and when a promising idea is ready to shape products, culture, or capital.
Iceland turns geology, climate, darkness, distance, and small scale into a living laboratory—revealing how scientific constraints can become products, culture, and strategic advantage.
From geothermal wells to glacier instruments and cultured proteins, Iceland reveals why scientific progress is governed as much by seasons, infrastructure and trust as by ideas.
A field guide to the institutions, infrastructures and communities turning Iceland’s volcanoes, genomes, fisheries and creative culture into consequential knowledge.
From our own rounds
Measured on The Curator, from real sessions people played on this site — not a third-party dataset.
- Rounds played here
- 167
- Questions per round
- 1.7