maskys.comShurik's Static

Open Loops

10 min read

Ways of seeing and acting

1. How do we build tools for thought to model complex systems?

  • Questions:
    • Which higher-level affordances need to compose, and what should composition mean in practice?
    • How do existing modeling tools move from evidence and assumptions to experiments, decisions, and explanation?
    • What has to remain connected when several models describe the same system?
    • Which serious context would expose the important omissions?
  • Established simulation software already handles parts of modeling, testing, collaboration, and explanation that a new tool should not rediscover.

1.1 How can we see a system across scales and models at once?

  • Yoshiki Schmitz suggested simultanescope for an instrument that could show a system across scales “all at once” (2020 thread).
  • Some New York Times explainers keep one map or model in view while the prose changes what it emphasizes. The reader gets a guided path without losing the whole, and can stop to inspect a detail before continuing. That is the useful precedent here.
  • Tweetscope is an early probe: a way to see a large body of thought as a map rather than one post at a time.
  • Other references include Bret Victor’s Up and Down the Ladder of Abstraction, Chaim Gingold’s Earth: A Primer, and What is the 21st century equivalent to punctuation?.
  • Questions:
    • How can a person inspect a detail without losing its place in the whole?
    • How should a map, timeline, network, simulation, comparison, and evidence trail respond to one another?
    • How can a reader move between a guided explanation and free exploration?
    • How should disagreement and uncertain evidence remain visible?

1.2 What can useful games teach serious tools?

  • Useful games make understanding a system part of play.
  • Cases to examine:
  • Play supports exploration and practice across humans and other animals, but that alone does not show that learning transfers outside a game (Špinka, Newberry, and Bekoff; Burghardt).
  • Questions:
    • What does a player learn: the modeled system, the interface, or the strategy that wins?
    • Which ways of revealing state, constraints, time, and feedback transfer to serious models?
    • How would we test whether anything useful survives outside the game?
    • How can a serious tool preserve evidence, uncertainty, and omitted mechanisms while remaining explorable?

2. How should humans and agents share a thinking environment?

  • Questions:
    • How do we keep useful information from disappearing inside chats?
    • Should conversations, evidence, instructions, and resulting artifacts live in the same medium?
    • What belongs in a transient conversation, and what should become part of the durable artifact?
    • How should an agent propose changes to higher-level objects rather than only raw text or files?
    • How should authorship, sources, uncertainty, disagreement, and voice survive revision and summarization?
    • Which changes can an agent make without prior approval, and which should remain proposals?
    • How should an agent distinguish contextual material from instructions?
  • Recursive Language Models are relevant to keeping artifacts outside temporary model context and operating on them through programmatic handles. They do not answer the longer-term collaboration questions.

3. Which evolutionary mechanisms are useful for design and modeling?

  • Evolution is one of the best designers we know, yet I do not study it nearly enough.
  • I have not found one accessible, maintained resource connecting mechanisms, suitable problems, mature tools, examples, and characteristic failures. That is a search finding, not proof that none exists.
  • Questions:
    1. How should we map mechanisms such as development, plasticity, drift, exaptation, coevolution, niche construction, modularity, redundancy, multilevel selection, and evolvability?
    2. Which tools help us inspect populations, lineages, environments, tradeoffs, robustness, and changing conditions?
    3. Which real problems have evaluators good enough for evolutionary search, and where would the fitness function conceal the unresolved problem?
    4. How should we account for premature convergence, proxy gaming, arms races, path dependence, simulation-to-reality gaps, and destructive real-world experiments?
    5. What would a useful map of evolutionary techniques look like?
  • Possible first probes:
    • Compare biological mechanisms with their computational or design analogues, implementations, applications, and failures.
    • Test a bounded physical problem whose result can be checked against reality.

People and institutions

4. How do institutions shape people, and how do those people reshape institutions?

  • Questions:
    • How do rewards, sanctions, roles, media, and peers change what people notice, value, and do?
    • Which behaviors win money, attention, security, or authority inside a particular institution?
    • Who gains influence as a result?
    • How can those people change the next round of rules and information?
    • Which feedbacks help the institution learn, and which protect the people who already won?
    • Where do perception, persuasion, cognitive security, group norms, biology, and individual agency belong in this account?
  • One lead on the selection and information questions is Daniel Schmachtenberger’s The War on Sensemaking.

5. How do people build durable coordination?

  • Questions:
    • How do the right people find one another?
    • How do they divide labor and make commitments?
    • What shared context do they need?
    • How do they notice failure and recover from conflict?
    • How do they handle defection without turning shared state into surveillance or managerial theater?
    • What remains after a project or event ends?
  • Leads: Vitalik Buterin’s Coordination, Good and Bad and stigmergy.

6. How do we build better systems of governance?

  • One existing thread is my attempt to understand the structure, policies, history, and change of the San Francisco government through CivLab.
  • Questions:
    • Which institutions detect and correct their own failures?
    • How can an institution’s formal powers, informal practices, budgets, dependencies, interfaces, and failure modes be inspected together?
    • What keeps an experiment accountable to the people inside it?
    • How does a better arrangement survive capture, incumbents, scale, and time?
  • Leads: Polis; Vitalik Buterin on balance of power and defensive acceleration.

Bodies and environments

7. How do coral reefs survive this century?

  • Coral reefs are the closest I have come to experiencing an alien world. They cover less than 1% of the ocean floor but host roughly a quarter of marine species, protect coastlines, and support food and livelihoods for hundreds of millions of people.
  • From January 2023 through September 2025, bleaching-level heat stress affected about 84.4% of the world’s reef area. NOAA describes the fourth global bleaching event as the largest recorded so far (NOAA Coral Reef Watch).
  • Pressures include:
    • Ocean warming and acidification.
    • Agricultural and wastewater runoff.
    • Chemical and material pollution.
    • Loss of herbivores and predators.
    • Destructive fishing, dredging, anchor damage, coastal development, and poorly managed tourism.
  • Work to follow:
    • Citizen science and monitoring (Reef Check, Allen Coral Atlas).
    • Coral gardening and artificial structures (Coral Vita, Mossy Earth).
    • Assisted evolution.
    • Watersheds, wastewater, local protection, reef insurance, and incentives tied to measured reef health.
  • Questions:
    • Why are some coral genotypes and reefs more heat-tolerant than others?
    • How much belongs to coral genetics, symbionts, or the wider microbiome?
    • Where are the thermal refugia, and are they protected?
    • What are the tipping-point dynamics, and can we detect useful early-warning signals?
    • When does local restoration remain useful under global warming and acidification?

8. How do we understand and reduce microplastics?

  • Microplastics and nanoplastics have been reported in human tissues. Detection, exposure, and causal disease burden are separate questions; recent reviews do not establish causal relationships with specific clinical diseases (review).
  • Source estimates depend on what is being counted. The OECD identifies road transport as the largest assessed source of global microplastic leakage in 2019; an IUCN study of primary microplastics entering oceans found synthetic textiles and tire abrasion dominant within that narrower category (OECD, IUCN).
  • Possible intervention points:
    • Road transport: lower-shedding materials, road-runoff treatment, and collecting tire dust as it is produced (The Tyre Collective).
    • Synthetic textiles: different fibers and washing-machine filtration (PlanetCare).
    • Water and food: filtration and source control. Could okra-fenugreek polysaccharides, or similarly cheap methods, work at municipal scale?
    • Existing pollution: interception and cleanup (The Ocean Cleanup).
    • Measurement: cheaper comparable testing and public interpretation (Ocean Diagnostics).
  • Questions:
    • Which sources dominate human exposure?
    • Which particle sizes, polymers, shapes, additives, pollutants, doses, and routes cause which effects?
    • Which interventions remove enough material to matter without creating a worse substitute?

9. How do we better screen the chemicals around us?

  • Chemical examples include:
    • Endocrine disruptors and plasticizers.
    • PFAS.
    • 6PPD and other tire-related compounds.
    • Heavy metals.
  • Related material and biological exposures include microplastics, mold, and other indoor-air contaminants.
  • Questions:
    • What is present in a product, building, food, water supply, workplace, neighborhood, or body?
    • What might be hazardous?
    • What are people exposed to, by which routes and at which doses?
    • Which signals deserve confirmatory testing or further study?
    • When should something be substituted, restricted, remediated, or avoided?
  • Leads: PlasticList, Human Chemical Co., Insilica, and Remedy Scientific.

10. How do we understand what enters our bodies and what it does?

  • A Research Breakdown for anything:
    • Could we build something like Examine.com’s Research Breakdown for any substance, product, supplement, drug, food additive, or environmental exposure?
    • How do we identify the substance and dose accurately?
    • How should product testing, mechanisms, epidemiology, trials, and expert review appear together without flattening their evidential quality?
    • Leads: Elicit for research synthesis and PlasticList for product testing.
  • Human lab rats:
    • How can we learn more from people already experimenting on themselves without treating anecdotes as controlled evidence?
    • What intervention, dose, context, baseline, outcomes, and adverse events need to be recorded?
    • How should the system represent different risk tolerances?

11. How do we scale healthy—or at least non-unhealthy—food to everyone?

Changing physical constraints

12. How will cheaper solar change the economics of fuels and chemicals?

  • Solar is taking off. Solar PV capacity additions passed 600 GW in 2025, and solar generation grew by about 600 TWh—both records. Solar supplied more than three-quarters of new renewable capacity (IEA).
  • Electricity still has hard constraints. Solar output varies by time and weather; storage, transmission, grid connections, flexible demand, and other sources of supply have to bridge the mismatch. Grid investment is lagging generation, with more than 2,500 GW of generation, storage, and large-load projects stalled in connection queues worldwide (IEA, DOE).
  • Solar can also become fuel. Renewable electricity can be used with water, CO₂, heat, and chemical processes to make hydrogen, hydrocarbons, ammonia, and other fuels or feedstocks compatible with systems already built around molecules (DOE).
  • Efforts to follow: Terraform Industries, Aircela, and Synhelion.
  • Questions:
    • Which fuels and chemicals become economical first?
    • Can production use cheap or otherwise curtailed solar without making poor use of expensive equipment?
    • When are storage, transport, energy density, and compatibility with existing infrastructure worth the conversion losses?
    • What is the actual cost stack, including capacity factor, CO₂ source, water, catalysts, compression, distribution, regulation, and finance?
    • What does lifecycle accounting show once the sources of energy and carbon are included?

13. Can robots learn many kinds of physical work?

  • These efforts share a bet on learned systems that can acquire many physical skills and use them across new tasks, objects, and environments.
  • Efforts to follow: Physical Intelligence, Figure, 1X Technologies, Sunday Robotics, and Unitree Robotics.
  • Questions:
    • How well do skills transfer to unfamiliar tasks, objects, layouts, and robot bodies?
    • Which combinations of human demonstrations, teleoperation, simulation, and experience on real robots produce useful learning?
    • How much apparent autonomy depends on teleoperation or supervision?
    • What do reliability, safety, maintenance, and cost look like outside demonstrations?

Thoughts, suggestions, counterarguments, recommendations? Working on one of these? Email me at contact at maskys dot com.

Backlinks

Graph View