Eco gets called "the educational one" so routinely that the phrase has stopped carrying information. What almost nobody who repeats it knows is that the claim was tested — with a control group, by the organisation that paid for the game — and that the result contradicts the obvious reading of it.
Between 2014 and 2017, the U.S. Department of Education put just under $1.2 million into Strange Loop Games across three research awards, commissioning Eco as a classroom instrument years before it was a Steam product. When the funder later reported its own pilot, the students who played finished ahead of the control group on environmental attitudes and on knowledge of systems, and level with them on content knowledge. Playing Eco did not make them better at ecology facts. It made them better at reasoning about how the pieces of a world act on each other — and at arguing, with the world's own data in hand, about what to do next.
That is a narrower claim than "learn ecology through play," and a far more interesting one, because it is precisely what Eco's machinery is built to produce. The game does not lecture about externalities; it gives your pollution a physics, a map layer and a permanent consequence, then hands you a legislative editor and a 24-hour vote. The part a player might reasonably experience as bureaucracy is the part the Department of Education was buying.
Update 14 — Garbage & Recycling, announced for 6 August 2026 — leans into exactly that. It turns industrial waste from an invisible side effect into a physical block with a location, a pollution type, a salvage value, a dedicated profession, and a penalty on the housing standing in it. Here is what the evidence actually says Eco teaches, which mechanics do the teaching, and what it would take to get that out of it yourself.
Eco Started as a Federal Education Contract, Not a Steam Game
The money left a paper trail, and the paper trail is unusually specific. Three award records sit on the Institute of Education Sciences' site, each naming Strange Loop Games as the awardee and John Krajewski as principal investigator, each issued under ED/IES SBIR — the Department of Education's small-business research program, which functions less like a prize and more like seed capital for products the department wants to exist. The organisation page lists exactly three. There is no fourth.
- EDIES14C0044 — $149,833. Phase I development, 1 July 2014 to 31 December 2016. Titled Eco: An Online Virtual World for Secondary School Environmental Literacy and Collaborative Problem Solving.
- EDIES15C0028 — $899,871. Phase II development, 7 May 2015 to 6 May 2017. Same title, with "Middle School" in place of "Secondary School".
- EDIES17C0039 — $149,952. Phase I development, 1 May to 31 October 2017. Titled A Collaborative Interface for Teacher-Student Interaction in Virtual Environments.
Add those up and you get just under $1.2 million — deliberately stated that way, because two U.S. government pages disagree about the third figure, and a false-precision total would be the least interesting kind of wrong.
What each award bought is more revealing than the totals. Phase I bought a prototype: a shared online world with a simulated ecosystem of plants and animals, packaged with teacher resources aligned to classroom learning objectives and guidance on implementing them. Its research questions were the questions a school district asks, not the ones a publisher asks — did the prototype function as intended, could teachers feasibly integrate it into a classroom, and were students engaged. Phase II, the award that paid for most of Eco's actual development, was justified in its own project summary by declining U.S. science performance, citing that over a third of eighth-graders scored below basic level on the 2011 national science assessment. The 2017 award is the one to be careful about: six months and a prototype teacher dashboard that would auto-generate reports on individual student contributions and progress toward curricular learning goals, covering Eco and a second Strange Loop title, Colony, piloted across three middle-school social studies classrooms. That contract closed without publishing findings, and no public record shows the dashboard reaching teachers afterwards. It was funded and it was piloted. That is the whole of what can be said about it.
Strange Loop was not an obvious recipient. Krajewski founded the studio in 2009 and shipped two smaller games praised for their scientific representation and largely ignored commercially; by his own account the application ran to eighty pages. What made it credible was a research partner. From June 2014, the University of Illinois' Impact on Science Education team worked with the studio on the classroom prototype — contributing project direction, putting the ecosystem simulation on a scientific footing, testing Eco in real classrooms, and writing up the results in the report that won the grant. Department of Education funding, in other words, did not land on a finished game and ask it to be educational. It arrived at the beginning, with a university attached.
The brief itself is where the strangeness starts. In the funder's own summary, Eco is a product to "prepare middle and high school students to be environmentally literate and capable citizens," in which students "co-create their civilization by measuring, modeling, and analyzing the underlying ecosystem, and advocate for proposed plans to classmates and make decisions as a group." It was written against Common Core and the Next Generation Science Standards, "focusing especially on 21st century skills of collaboration, self-directed learning, leadership, scientific argumentation, and applied learning," and specified to be played during class or outside it, at home or after school, augmenting instruction rather than replacing it. Read that list of deliverables again and notice what is missing from it. Nobody is buying recall of the nitrogen cycle. The specification names behaviours — measure, model, advocate, decide together — and treats subject matter as the thing those behaviours happen to be pointed at.
That distinction is not an accident of drafting; it is visible in how the game was built. Pre-production ran on surveys of teachers, and those surveys asked two things: which concepts lined up with their learning goals, and what their students commonly get wrong. The misconceptions came back, and the team used them to shape the narrative arc and the kinds of decisions the player is asked to make. Sit with the direction of that: most games begin with a verb someone wants to perform and reverse-engineer a world that rewards it. This one began with a catalogue of wrong beliefs and reverse-engineered the situations that would put a student in front of one. Then the stage closed the way an education project closes rather than the way a build closes — with a pilot of 60 students across five classrooms, checking that the software worked, that teachers could use it, and that students would collaborate. Krajewski's stated premise for the whole exercise was that "school has a history of being terrible at showing students why they should care about a subject," and the method follows from it: you cannot show someone why a concept matters until you know which version of it they are already carrying around.
The commercial game came afterwards, and the sequence matters. In August 2015 — three months into the Phase II award period — Strange Loop went to Kickstarter with a $100,000 goal, came away with more than $200,000, and steered toward the Steam Early Access release that eventually landed in February 2018. The survival-crafting product is downstream of the education contract, not the other way around, which is why so much of what makes Eco awkward as a survival game (the paperwork, the data, the arguing, the votes) is not a designer's indulgence. It is deliverable language.
So the useful move for a reader is to stop assessing "Eco is educational" as a vibe and start assessing it as a specification that somebody signed. Department of Education funding of this kind buys a stated outcome, and the outcome stated here was that students would reason and decide differently, not that they would score higher on vocabulary. That is a testable claim, and both development phases carried research obligations to test it — which is the thing about commissioning a measurement instrument. Sooner or later, somebody measures.
What the Government's Own Pilot Measured — and What It Didn't Move
The measurement, when it came, was small and blunt. Eighty-eight students across middle and high schools; forty-four of them played Eco for two to four weeks; the other forty-four carried on with business-as-usual activities; then both groups sat a post-test. Teachers were able to fit the game and the discussions around it inside a unit on environmental science, and every student was able to take part — the feasibility questions the awards had been asking since 2014 came back clean.
Before the result, its provenance, because that governs how much weight it can carry. This is not a journal article. It appears in the Department of Education's own SBIR success story for Eco — a funder-published account of what the money produced, written by the organisation that spent it. It gives no effect sizes and no significance levels. It does not name the instruments the students were tested with, does not report attrition, and does not say how the two groups were formed. Forty-four students per side is a small sample by any standard. Nothing about that makes the finding worthless; it makes it a reported pilot rather than a settled result, and it should be read as one. It is also, to this day, the only comparison of Eco players against a control group that anyone has put in public.
What it reports is oddly specific. The students who played scored higher than the control group on a measure of attitudes towards the environment. They scored higher on a measure of knowledge of systems. And then, in the success story's own words: "There were no differences between the groups on a post-test measure of content knowledge."
That is two different exams, and Eco passed one of them.
The success story itemises what the systems measure was made of — data reference, economy and development, environmental strategy. Translate those into questions a student would have to answer. Can you go and get a number out of the world's own record and use it to support what you are claiming? Can you reason about how production, growth and consequence push on each other rather than treating them as separate topics? Faced with a system you cannot see all of at once, can you pick a course of action and say why? Content knowledge is the other exam entirely: name the trophic level, define a carbon sink, list the stages of the nitrogen cycle, say which gas does what. Four weeks in Eco moved the first set and left the second level with the students who had simply carried on as normal.
The instinct is to read that null as the game falling short of its billing. It is better read as the sharpest description of Eco anyone has published. A month of play installed no facts that ordinary instruction was not installing anyway, and there is no reason it should have — nothing in the world stops to tell you what a trophic level is. What it did was change the kind of thinking students brought to the problem, which is precisely what the brief specified and paid to measure. The systems knowledge result and the content knowledge null are not a win and a loss. They are one finding with a shape.
Keep one thing separate from that finding. The Phase II award's project summary described a planned pilot of 150 students across 10 classrooms, with random assignment to either the Eco intervention or standard instruction and pre- and post-assessments on ecology knowledge and environmental literacy. That design is a plan, and no study matching it appears anywhere as having been carried out. The 88-student pilot is what is reported as run, and the two should never be spoken of as the same thing — the larger, cleaner design is the study Eco was supposed to get, not the study it has.
There is a second strand in the same document, and it is the one that shows systems knowledge before it becomes a score. With 39 middle-school students, researchers went at the problem from the other end: instead of testing anyone, they read the in-game chat logs. The students discussed STEM content and matters relevant to civics and environmental systems. They "conducted investigations in the game, referencing data from graphs as they communicated about the issues at hand". And they "discussed and coordinated with each other to ensure that all of the skills necessary to succeed in the game were present in the community." Sit with that last one. Middle-schoolers, unprompted, auditing their own group for missing capabilities and reallocating themselves to cover the gaps. That is not a lesson about ecology. It is a lesson about a system with people in it, delivered in a chat window, and it carries the same provenance caveats as the pilot — same funder document, no control group, no numbers attached.
Even the marketing quietly agrees. The pull-quote Strange Loop Games chose for its own Steam page is Jeff Grubb's, and what he says he took away is that games are "crucial for understanding our relationship to all kinds of natural and man-made systems" — that he learned a great deal with no instructor connecting the dots. It is a testimonial about systems from a man who names no ecological concept in it.
So the practical form of all this is a decision you make before you spend anything. If the objective is ecology facts — vocabulary, cycles, definitions you could quiz — expect Eco to buy you nothing the lesson you would otherwise have taught does not, because the one time anyone checked, the players and the controls came out level. If the objective is students who can pull evidence out of a system, reason about how its parts act on each other, and decide as a group what to do next, there is a controlled result on your side, with all the caveats attached to it. Pick the learning objective first; the tool is only correct relative to it.
All of which rests on a document the funder wrote about its own product. Which leaves the obvious question: has anyone without a stake in the answer looked at Eco and seen the same shape?
The Independent Research Points the Same Way — and Names the Ceiling
Two independent efforts have looked, and neither went about it in a way the Department of Education would recognise. One sat down with seven Norwegians in 2019 and asked them what they had noticed. The other scraped 1,825 Steam reviews and ran a network analysis over the vocabulary, in a preprint submitted in March 2026. Between them there is no control group, no post-test, no random assignment and no p-value. They share no method, no country, no decade and no subjects. They arrive at the same place, which is the interesting part.
Start with the peer-reviewed one, because it is the only peer-reviewed study of Eco that exists. Kristoffer Fjællingsdal and Christian Klöckner, working at the Norwegian University of Science and Technology, published Gaming Green: The Educational Potential of Eco – A Digital Simulated Ecosystem in Frontiers in Psychology on 19 December 2019. It is qualitative from end to end: semi-structured interviews after people had played, run through thematic analysis. Fifty-nine participants were recruited out of two Norwegian high schools, three university classes and four environment-and-gaming Facebook groups, handed logins from a pack of classroom accounts, and set loose for two to four weeks — mostly in groups of four or five — on what was then an unfinished beta with unintuitive controls and missing textures. Then seven of them agreed to be interviewed. Seven, from fifty-nine, all male; the authors put the shortfall down partly to winter exams and partly to the fact that nothing about the game was tied into anyone's curriculum. There are no numbers in this paper because there was never going to be a number in this paper. Anyone citing it as evidence that Eco produces measured learning gains has not read past the abstract.
What it can do is tell you what the seven said, and what they said lines up uncomfortably well with the null the funder reported. The dominant theme the authors named was contextualising knowledge: players were not acquiring environmental concepts they lacked, they were re-encountering concepts they already had, in a form that finally had weight. "When you get it simulated through Eco then you see it a bit more clearly," one put it. Another, describing what over-extraction looked like from inside a shared world: if you can imagine a hundred people doing the same, there wouldn't be any life left. Participants recognised tragedy-of-the-commons dynamics unprompted, once hoarding started to threaten the group's survival, and several traced the difference to the fact that Eco's world runs out. The comparison they reached for was Minecraft — a game where you can dig forever and the resources are effectively bottomless — and against it, Eco's hard carrying capacity made depletion feel, in their words, more real. That is the same distinction the funder's pilot found on its instruments: not more facts, but a different grip on the ones already there.
Then the study names its ceiling, and does it more honestly than most funded research ever does. One of the two sub-themes under barriers is titled "Good for Others, Not for Me." The participants who came in with substantial prior environmental education reported getting very little new out of it. One said he could not stop looking under the hood at the mechanics rather than engaging with the environmental content at all — he was reverse-engineering the simulation instead of learning from it. And the authors are explicit about the context they were working in: Norway is a country with unusually strong environmental education already in the curriculum, which they flag as the likely reason their sample learned so little that was new. Seven interviews, all male, on a beta build, in the country least likely to be surprised by any of this.
That ceiling follows directly from what the game does. Eco converts abstraction into experience, so its entire effect sits in the gap between knowing a thing and having lived inside it. A cohort that has been taught this material properly will get a demonstration of what they were told. A cohort that has not will get the thing itself. The value is not in the content; it is in the conversion, and the conversion only pays out once.
The second source is a preprint, and has to be labelled one every time it comes up: arXiv:2604.15317, submitted on 2 March 2026 by Yifan Xu and colleagues, not peer reviewed. Its method is deliberately nothing like an interview study. The authors argue that pre- and post-surveys suffer from response bias, so instead they harvested 1,825 substantial user reviews from Steam and built co-occurrence networks out of the language, comparing Eco against WolfQuest — two environmental games picked for holding opposite philosophies. The preprint concludes that Eco produces what it calls "Socio-Political Cognition": players write about environmental problems as legislative and economic frictions. Not as loss, not as damage to something alive — as rules, costs and things that have to be negotiated with other people. Its subjects are self-selected Steam reviewers rather than learners, no learning outcome was measured anywhere in it, and what it observed is strictly how players talk. Treat it as a description of the discourse Eco generates, which is all its authors claim for it.
The preprint also supplies the sharpest version of the limit, because it puts a name to what Eco is not. Its argument is that environmental education needs a strategic tension between systemic logic and emotional resonance, and that these two games hold one half each: Eco makes you a legislator reasoning about frictions across a whole system, WolfQuest makes you a vulnerable animal whose life can end. Neither one, in the preprint's reading, is sufficient by itself.
So the matching rule falls out of the evidence rather than out of taste. Point Eco at people who have not already been taught this material, and expect what comes back to be arguments — about rules, about who pays, about what a settlement can afford to permit — rather than feeling about nature. If what you want is an emotional connection to the environment, Eco is structurally the wrong half of the pair, and the preprint's own comparison tells you roughly what the right half looks like. If what you want is students who will fight about a tax on a smokestack, you are holding the correct object.
Three sources, no shared method, one shape: Eco teaches reasoning about systems, not the contents of ecology. Each is weak exactly where its own method is weak, and no two are weak in the same place, which is the only reason three thin results are worth more than any one of them alone. Which turns the question mechanical. Something in this game is doing that, deliberately, and it is worth knowing exactly what.
The Simulation Teaches by Making Your Own Damage Measurable
The most useful sentence anyone at Strange Loop Games has said about the design was said in 2016, to a trade publication, about methodology: "When we start a system we think about how to make it connect to as many other systems as possible while being as simple as possible." Both halves are load-bearing. The first is the ambition — coupling everywhere, so that nothing you do is contained. The second is what turns coupling into a lesson instead of a fog. If each individual system is simple, any single link can be held in one person's head, and a chain of simple links is something a fourteen-year-old can walk backwards from an outcome to a cause. A world where everything affects everything and nobody can trace how teaches nothing at all.
The studio's own store copy states the ambition as a promise: "a disruption in one species can have cascading effects across the planet… Chop down all the trees, and habitats for creatures will be destroyed. Pollute the rivers with mining waste, and your farms will become poisoned and die." Marketing for half the survival games on Steam says something like that. What separates Eco is that each of those clauses is implemented as a behaviour with a location, a direction of travel and a number attached, rather than as a scripted event that fires when the designer wants you to feel bad.
Take the one the game is named for. Pollution in Eco is not a single meter; it is two families with genuinely different physics. Ground pollution comes from garbage — items dropped on the ground — from sewage, a waste byproduct of smelting at a blast furnace and of other machinery, from tailings, produced by concentrating dry iron, and from wet tailings, which the wiki calls a more potent byproduct of the ore-concentrating recipes. It includes water, and it spreads using the rules of hydrology, seeping into the blocks next to it. It goes, in other words, wherever water goes: downhill, downstream, and into the low flat ground that is also the best place to farm. Air pollution is smog, produced by burning fuels at crafting tables and by running vehicles, and it behaves in the opposite way — it spreads through the air, affects flora and fauna the same way, and is quick to dissipate if it is given time. Both can be routed through pipes, which is why Eco players build exhaust stacks and sewage runs instead of just walking away from the mess.
Two physics means two mental models, and the game makes you hold both without ever naming them. Air pollution is a flow: stop emitting and it thins. Ground pollution is a stock: it sits in the terrain, it creeps, and it does not care that you have stopped. The wiki's own illustration of the air side is a blast furnace and a cement kiln venting into a single shared exhaust pipe for about forty-five minutes, and the spread it produced is captioned "very aggressive" — one vent, two machines, a compounding output. And because plants and trees will not grow in polluted areas, though some species resist better than others, the consequence lands on the thing you feed yourself with. That is why experienced settlements put industry downwind and downstream of their farms. It is not roleplay. It is reading the topography.
Then there is the consequence that does not come back. Air pollution dissipates, but prolonged air pollution raises sea level, and the wiki is blunt about the character of that: devastating, and permanent. No rate is published anywhere, and it is worth resisting the urge to imagine one — the mechanic's teaching content is its direction, not its magnitude. Almost every resource in a survival game is a meter that refills if you behave. This one is a ratchet. A group that pollutes hard for a fortnight and then reforms does not get its coastline back, and no amount of subsequent virtue converts the loss into a lesson-learned reset.
Reforestation is the counter-rate, and it is published as arithmetic. The wiki lists per-species daily CO₂ absorption, and the interesting content is the ratio rather than the raw figure: an oak absorbs exactly twice what a birch does, and four times what a Joshua tree does. Treat those as relative weights and nothing more — the table carries no version tag and states its units only as "daily rates", so it is not a calibrated real-world value and should not be read as one. Even so, the effect on a player is immediate. Replanting stops being a gesture and becomes a species-selection problem with a rate attached. And since nothing grows in a polluted patch, the pollution you generate degrades the very machinery that removes it. Nobody in the game ever uses the phrase "feedback loop." A player who plants birches downwind of their own smelter discovers one anyway.
The one published figure for containing ground pollution is a case study in why you should check dates. The wiki states that pollution reduction for ground pollution is based on how many solid blocks sit above it — every five solid blocks halves it, so ten blocks gives a 75% reduction and fifteen gives 87.5% — and tags that rule explicitly to Version 0.9.2, many releases behind the build people are playing now. Cite it as a documented rule from an old version, not as current behaviour. What survives the version question is the shape: burying waste attenuates it, and attenuation is not deletion. The stuff is still there, under your floor, doing a fraction of what it did.
All of which converges on the single loop that is Eco's whole argument in miniature. Every world runs under a meteor on a countdown, and the technology that stops it is built out of metal. Krajewski put the mechanism at its bluntest in 2019: "In Eco, players generate CO2 by making iron, an important resource." Be careful with that as a recipe line — it is a causal shape described at a conference in 2019, and the current wiki attributes smog to burning fuels at crafting tables generally and tailings specifically to the concentrating of ore, so the emission attaches to the industrial process around iron rather than to one labelled step. The shape is what matters, and the shape is vicious. The settlement that smelts hardest is the settlement most likely to beat its own deadline, and the settlement most likely to have ruined the planet it was racing to save. When a group responds by capping how much iron gets produced, or by taxing it, they are not choosing between industry and nature as a slogan. They are trading days off their own survival clock against the state of their own atmosphere, and both sides of that trade are quantities they can look up.
That is the part people skip, and it is the part the Department of Education was paying for. The simulation's state is queryable. In-game graphs and heat maps cover plants, animals, climate and player activity, and the World Layers overlay opens a menu of exactly what the world is willing to tell you about itself: Property, Districts, and Environment Data broken into Animal, Biome, Moisture, Nutrient, Plant Group and Pollution — with Pollution itself decomposing into Air Pollution, Debris, Ground Pollution, Player Activity, Player Trampled, Trampled Spread and Trampled Sum. The game keeps a layer for where feet have been. Tick "Show Overlay in Main View" and the selected layer stops being a map and becomes a projection painted onto the terrain you are standing on, so the bloom of ground pollution creeping out from your own operation renders across your own valley while you walk through it.
This is the teaching surface. Not the description text, not the pollution's colour palette, not the premise about a meteor — the readout. The transferable skill Eco drills is reading a system's state before acting on it, and it drills that skill by making the state cheap to read and the consequences expensive to ignore. You do not come away having learned that industry produces externalities. You come away having watched yours spread, in a layer you switched on yourself, over ground you chose.
Which has a practical edge for anyone deciding whether this game is worth the trouble: if you play Eco, or run it for a group, and nobody ever opens World Layers, you are using the least educational half of it. The survival-crafting half works fine without the data. The part that was commissioned does not.
And once the damage is visible, and attributable to somebody, the next thing that happens on every server is that someone proposes doing something about it. Eco's answer to "do something" is not a button.
The Bureaucracy Is the Curriculum: From Heat Map to Passed Law
It is a hotkey. Press L, and Eco opens the server's own web page in your browser, and what is waiting there is a rules engine. Laws in Eco are proposed at a Court, and the law editor's format is fixed: if X happens, then do Y. You assemble the rule out of colour-coded blocks — Triggers, then Else If, then Then — filling dropdowns as you go, and above them the editor renders your finished rule back into plain English so you can read what you have actually written rather than what you meant to write. The header carries two buttons: Law Guide, and Add New Law to an Election. That is the whole shape of the thing. A survival game about chopping trees ships an authoring environment with documentation, a plain-language decompiler, and a ballot attached to the save button.
Here is one, read off an official screenshot of somebody's server. It is titled the Clean Air Pollution Act, and it opens the way a real bill opens — Section 1, then a preamble the player wrote in their own words, describing "a forward-thinking piece of legislation aimed at curbing air pollution" whose revenue is reinvested into clean-air efforts. That part is rhetoric, and it does nothing. Underneath it is the part that runs:
On event Pollute Air / if Citizen is a citizen of Lighthouse Point / then Tax Citizen 15 TonyPepproni Credit into TonyTown Treasury.
Expand the blocks and you can see every decision the author made. The Trigger is Pollute Air, with Pollution Source: Any — no carve-outs for anyone's favourite industry. The Else If condition is Is Citizen Of Settlement → Lighthouse Point. The Then is Tax, targeting the citizen, paying into the government account TonyTown Treasury, in the currency TonyPepproni Credit, amount 15. Lighthouse Point, TonyTown and the TonyPepproni Credit are one server's names for its own settlement, its own treasury and its own money, not defaults you will find in your world; what generalises is the grammar, not the nouns.
Two of the smaller fields are where the actual thinking lives. Prevent If Unable To Pay: No means the law does not stop you from polluting when your bank balance is empty. That single dropdown is the difference between a prohibition and a price, and the author chose price. Infobox On Success: Yes means the game tells you, each time, that you have just been charged for the smoke you made. Put the whole thing together and it is a Pigouvian tax: a levy set at the point of emission, sized to make the polluter carry a cost they were previously handing to everyone downwind, with the proceeds going to the public purse. Nobody involved has to have heard that word, and almost certainly nobody had. They had a smokestack problem and a law editor, and the editor's grammar pushed them into economics.
Then it goes to a vote, and the voting machinery is as specified as the drafting machinery. A world's Constitution is authored at a Capitol, and the predefined one is a basic democracy: every change to Laws, Demographics, District Maps, Elected Titles, Executive Actions or the Election Process needs a majority of all players. Once something is proposed, everyone has 24 hours to vote on it. Elections start at a Government Office or a Court; contests for Elected Titles run from a Ballot Box, where any qualified player can stand, and a title can be defined to carry real powers — access to the treasury, or a veto over proposed laws. Districts are drawn inside a law proposal itself: you open the law menu, choose Redefine districts, and colour areas directly onto the map, after which rules can be written that apply to those areas and not others. A student who wants a rule to bite on the industrial quarter but not the farms has to go and draw the industrial quarter.
And the vote is where the pollution readout stops being a curiosity. The same web page that holds the ballot holds Scientific Data: a category tree covering Farming, Finance, Harvesting, Pollution, Drop Garbage, Settlements, Specialties, Work Party, Plants, Trees, Animals, Environment, Citizens, Civics and Economy, plotted into a graph panel with a timeline scrubber, next to a ground-pollution heat map of the planet. Alongside it sits Elected Officials, with running elections and a scroll of previous ones. So the full loop is one sitting at one desk: switch on the pollution layer; open the web interface and pull the matching time series out of the Environment category; then attach it to the election, because the comment function exists so players can discuss a proposal and "substantiate the argument with statistical evidence (in the form of graphs)." The store page is not being coy about this. It calls the ability to debate with scientific evidence "your greatest weapon against the destruction of your world." That is measurement, argument and enactment inside a single interface, and the interface was built in that order on purpose.
Taxes get the same treatment. There are four kinds, and all of them flow into the Treasury: Sales Tax, Crafting Fee Tax on a station owner charging others for use of their machine, Contract Tax paid by whoever receives money when a contract completes, and Direct Transfer Tax on account-to-account payments. The official wiki then does something no survival game's documentation has any business doing, which is warn you about behavioural responses. Tax direct transfers too hard, it says, and players simply route around you with multiple bank accounts — so keep that one low or zero. And a government-minted currency that gets hoarded "slows down and finally stops the economy." Those are two lessons — tax avoidance responds to the rate, and money has to circulate to be money — sitting in a wiki page beside instructions for placing a ballot box.
What makes the law editor the centre of gravity rather than a side system is that Eco leaves you nothing else. There is no PvP focus and no player death; conflict here is economic, legal and ideological, which means you cannot resolve a dispute by removing the person causing it. And by Krajewski's own account of how every server begins, "people, by default, will not consider the group" — early play is individual tools, individual houses, no thought spared for the commons, right up until the aggregate becomes visible. At that point the only instrument in the box is a proposal. You cannot burn down the smelter. You have to convince a majority of strangers, inside 24 hours, with a graph.
They do it, at volume. The status bar in one official screenshot reads: Online Players 7/765 · Day 30, 02:39 · Meteor Impact 21 hrs 20 mins · Plants 497,533 · Laws 60 · World Size 2.56 km². Sixty laws, written by one community, with twenty-one hours left on the clock.
Gamers.Wiki's own review of Eco, A Bold Civilization Sim Buried in Bureaucracy, reads exactly this machinery as the thing that smothers the game. Nothing here disputes that as a description of the play experience — the argument is narrower and more awkward: the paperwork is not a layer on top of the lesson, it is the lesson, and the Department of Education's brief bought a product where students "advocate for proposed plans to classmates and make decisions as a group." A system can be a drag on an evening's entertainment and the entire pedagogical payload at the same time. Those are not competing verdicts. They are the same mechanism, assessed against two different jobs.
The limit deserves stating as carefully as the mechanism, and the official wiki states it better than most funded research does. Government in Eco resembles the real thing, but: players have real lives outside the world; players can always decide to leave the server and play somewhere else; and Eco societies are small, so social effects will differ — bigger impact, more dynamic. Keep those three in mind, it says, before importing real-world political or economic theory. That is the honest boundary on this whole section. A citizen who can exit a polity by closing a window has not learned what it is to be stuck inside one, and a settlement of thirty is not a country. What survives the transfer is not the politics. It is the procedure.
Which is the thing worth taking away, and it is a sequence rather than a subject: measure, argue with the measurement, write a rule that enforces itself, then live under the rule you wrote. Each step is ordinary on its own — students graph things, students debate, students learn what a tax is. Doing all four in order, on the same data, with the outcome binding on you afterwards, is the part almost nothing else does; the fourth step in particular has no classroom equivalent, because a class debate ends when the bell goes and a passed law in Eco charges you fifteen credits every time you light the furnace for the rest of the world's life. If you are running this with a group and the sequence never completes — if the graphs get pulled and no proposal ever follows, or laws get written that nobody bothered to evidence — you have the parts and not the machine.
None of which explains why a settlement bothers. A government is expensive in attention and argument, and a player who could simply do everything themselves would never convene one.
Eco Doesn't Preach Interdependence — Update 13 Prices It
They can't, and the reason is arithmetic rather than atmosphere. On 8 April 2026, Update 13 rewrote Eco's progression from the ground up, pushing stars toward being a currency rather than a milestone — players earn significantly more of them and are made to spend more, with the stated aim of keeping good food and housing worth chasing. Buried in that rework is the rule that does the actual work of pressing a group of strangers into a society, and it is short enough to quote in full.
Specialties no longer cost the same. Logging costs 1 star, Farming 2, Baking 3, Oil Drilling 4, and Composites 5 — a ladder where the further a trade sits from a stone axe, the more of your progression it eats. Then the part that matters: by default, for every specialty you have already learned, your next specialty costs an additional star on top of its base cost. Note that this is a 2026 change and not some ancient design principle. Before Update 13, every specialty cost one star, flat, and a player who wanted six trades simply bought six trades. The current build charges the sixth far more than the first, and charges you for the sin of having been curious.
Work it with real numbers, because the numbers are the whole argument. Baking costs 3 stars to a player who holds nothing else. To a player who already holds two specialties, the same skill — same recipes, same bench, same bread — costs 3 + 2 = 5 stars. Two-thirds again the price, and nothing about the skill changed; the only variable is who is buying it. Follow the rule further and it steepens: a fourth specialty pays base + 3, a fifth base + 4. Meanwhile the same update made talents cost a star each, with several retakeable at further star cost, so depth inside the trade you already have is bidding against breadth for the same pool. The player trying to be a one-person civilisation is not merely slower than the settlement. They are paying a surcharge that grows with every step of the thing they are trying to do, and paying it in the currency they would otherwise spend on being genuinely good at anything.
That is the entire mechanism behind a claim Eco's marketing has made for a decade. A settlement does not need a cook, a logger and a driller because cooperation is nice. It needs them because the alternative — thirty people who can each do a bit of everything — is the single most expensive configuration the game allows. Interdependence here is priced, not preached, and a price does something a moral never does: it changes behaviour in people who were not listening.
You can watch it work in the wild. A player with 475 hours describes, in a Steam review, what amounts to the designed behaviour arriving unprompted: they joined a world, walked around, talked to the inhabitants, pitched a tent in the middle of everyone, built a simple house — and then surveyed the neighbours, noticed the settlement had no good cook, and became one. Nobody assigned that. No quest asked for a cook. They read a community's gaps and priced their own labour against them, which is the behaviour the Department of Education's brief described as making decisions as a group, showing up in a store review written by an adult who was just trying to have a good time.
None of this is fixed in stone, and Update 13 says so out loud: modders can adjust, remove or add talents at any level, including talents that grant stars in exchange for negative effects. The price list is editable. Which matters, because there is a switch in the base game that edits it wholesale.
The escalating cost applies by default — except on the No Collaboration and Low Collaboration difficulty settings. Turn the difficulty down and the surcharge for holding multiple specialties simply stops existing. And the only public how-to for running Eco with a class, the official Education page on the developer's own wiki, tells teachers to do exactly that: depending on group size and the experience of the players, choose low collaboration or no collaboration, a setting the page describes as affecting gained experience points, collected resources and "other game parameters."
Before reading intent into that, read the dates. That page was last edited on 9 July 2024 — roughly twenty-one months before Update 13 attached progression economics to the collaboration setting. When the advice was written, "other game parameters" meant XP and resource yields; it could not have meant an escalating specialty cost, because there wasn't one. This is not a studio quietly undermining its own lesson. It is guidance that has been overtaken by a patch, and nobody has gone back to the page since. The tension is real regardless of how it happened, and it is worth saying plainly, because the page is still the first thing a teacher finds.
Here is what it produces. A class of twelve, configured on the advice of the developer's own education page to Low or No Collaboration, plays a version of Eco in which a third specialty at Baking's tier costs 3 stars rather than 5, a fourth costs its base rather than base + 3, and one diligent student can quietly become the cook, the logger and the driller by themselves. The headline mechanic — the one that makes Eco different from every other survival game with a crafting tree — is switched off in the configuration most likely to be used in the room it was built for. The same page still warns, correctly, that "even with no collaboration settings, it is hard to succeed in the game with no teamwork." That warning is doing a lot of work: teamwork stays useful, but it goes back to being a good idea rather than a bill. And a good idea is precisely the thing a class was already capable of ignoring before anyone bought a licence.
So treat the difficulty slider as a pedagogical decision rather than a comfort setting, because that is what it now is. Lowering it makes Eco kinder — fewer dead ends, less waiting on somebody who logged off, a gentler run for players who have never touched the game — and it makes the specialty star costs, and therefore the lesson those costs teach, weaker in exactly the same motion. Both of those are real, and you do not get to have both. If you turn the collaboration setting down, know that you have taken the mechanic that enforces interdependence out of the world and made yourself responsible for supplying it by hand: assign the trades, refuse to let one student cover three, and expect to be the reason the group specialises rather than letting the price list do it. If you leave it up, the game will do that work for you, and the arguing about who becomes the cook will start on its own.
Which is the pattern worth holding onto as the game keeps moving. Eco teaches by putting a number on things a group shares — the cost of a skill, the cost of a smokestack, the cost of doing it all yourself. So the obvious question to ask of the update Strange Loop Games has announced for 6 August 2026 is whether it adds another shared cost.
Update 14, Announced for 6 August 2026, Turns Waste Into a Shared Bill
It does, and the thing it puts a price on is rubbish. Update 14 — Garbage & Recycling — was announced for 6 August 2026 in a Steam post dated 31 July 2026, and the interesting part is not that it exists. It is that it is the third release running to tighten the same coupling, and the first two arrived quietly enough that almost nobody noticed an arc was under way.
Start in February. Maintenance update 12.0.7, released on 11 February 2026, is the kind of patch nobody reads: election bugs, settlement permissions, a fix for a law trigger named Drop Or Pickup Garbage that had been firing when a player merely opened a garbage bag — so the legal apparatus for regulating rubbish predates the update that makes rubbish matter. Buried under Gameplay is one line: "Added: A mechanic that reduces the housing score of rooms affected by ground pollution." Then Update 13, on 8 April 2026, promoted it in a balance note that is easy to skim past: "Pollution at the location of a room now negatively affects its housing score by default." Read the last word. In February the coupling was a mechanic somebody had added. In April it was how the game behaves unless a server operator decides otherwise.
Then there is the image attached to the 31 July 2026 launch announcement, which is the part worth sitting with. It is not concept art or a hero render. It is a map of cumulative ground pollution across the official Update 14 playtest server — the studio running the same readout over its own test world that a player runs over their own valley — captioned to say that residences inside those areas had "significantly decreased housing score due to the pollution effects." Strange Loop Games advertised its update with a picture of its own contamination and a note about what that did to the rent. Significantly is their word, and it is the only word on offer: nothing published puts a number on the penalty, and there is no honest way to estimate one.
What Update 14 adds, as described in the 24 July 2026 preview, is the supply side of that same problem. Crafting gains byproducts: a range of new blocks that, in the studio's words, provide "differentiated and significant impact on the environment through different pollution types" — so waste stops being one undifferentiated bad thing and becomes several with distinct behaviours — while still containing "valuable resources" you can extract with the new Recycling profession, moved around by reworked modular vehicles built for hauling it. Pollution also stops being solid-only: the preview says operations will need their input fluids filtered, and their outputs too if you want to keep your surroundings clean, with a desalinator required for any machine not sitting near fresh water.
Put the three together and waste occupies three roles at once. It is an input, because the byproduct block on your floor holds resources somebody can recover. It is a job, because Recycling arrives as a new profession, and somebody in the settlement has to be the one who takes it up — so the group is choosing between a recycler and whatever else that person could have become. And it is a bill on your neighbours, because the ground pollution it produces lands on the housing score of whoever lives in it, which need not be you if you put the works uphill and out of town. Stock, labour and rent, from the same pile of slag.
That is what converts a preference into a policy problem. A settlement has always been able to decide that pollution is bad; holding the opinion costs nothing. From 6 August 2026 the same settlement faces a question with quantities on both sides — whether to zone the industry, whether it can spare somebody to become the recycler, whether to tax the byproduct where it is made or price the housing that ends up standing in it. Those are precisely the coupled, collective decisions the Department of Education's pilot measured students getting better at, and they are arguable with the same graphs, in the same web interface, on the same 24-hour clock.
Two practical notes if you have a world already running. The preview's migration warnings are stated for the playtest branch and the shipped build may differ, but as written: older worlds load and are unsupported; pipes stop counting as a building material, so any room whose structure includes them fails until they are replaced; machines require fresh water, so anything drawing from the ocean stops until that water is desalinated; and modules are stripped from every table, converted to new equivalents and dropped into void storage. And the launch post promises release notes containing a breakdown of the new pollution mechanics with tips on minimising their effects — written, the studio says, so that arrivals start better prepared than the playtest participants were. Read that before you build anything downwind or downstream.
None of it is a new lesson. It is the 2014 brief — measure the ecosystem, model it, argue with classmates about what to do — with one more quantity bolted to it, and it only works because eight years of machinery already exist to make a bag of rubbish legislable. Which raises a question no patch note answers. The lesson keeps getting sharper. Is anyone still in the room to receive it?
The Machinery Still Exists; the Classroom Evidence Is Old
Nobody can say, and the shape of the not-knowing is the honest answer. Everything Eco was built to do in a classroom is still in the software and still documented. Almost nothing dated after 2019 says anyone is doing it. Both halves have to be held at once, because it is easy to collapse them into either "Eco is a school product" or "the school thing is over," and the evidence supports neither.
Start at the studio's front door. Strange Loop Games' homepage in August 2026 opens with "We Create Worlds" and an invitation to "build a connected network of gaming worlds that are owned and governed by their citizens." The pitch runs in three parts — Players as Citizens, where "citizens aren't merely the customers that are sold value, they are the value"; Worlds of Consequence; and User-Generated Societies, described as "connecting people and worlds together in a bottom-up metaverse." A second world is teased as coming soon, with no name attached. There is no mention of education, schools, classrooms, teachers or curriculum anywhere on it. The vocabulary survived intact — citizens, governance, consequence — and the buyer changed. The same civics machinery the Department of Education commissioned is now being sold to players as a metaverse proposition. The news feed underneath it has not moved in years: the latest entries are "Road Map to Update 10," "Update 9.7 released!" and a developer blog about Eco Infinite, several major versions behind a game that shipped Update 13 in April 2026 and has announced Update 14 for 6 August 2026.
A quiet marketing site is not proof that a programme ended, and it should not be read as one. The Education page on the developer's own wiki is still up, the citizen-account model it describes still exists, and none of that requires a paragraph on a homepage to keep functioning. What the silence does mean is narrower and still worth knowing: a teacher who arrives at the studio in 2026 will find nothing addressed to them.
Look for the classrooms themselves and the trail stops in the same decade. Every named school-level deployment predates 2020 — the pilots reported in the funder's own account, a class at Hathaway Brown School that produced a video about its experience playing Eco in January 2019, and the Norwegian recruitment through two high schools and three university classes that became the paper published that December. Nothing since. That is not evidence of absence; a private world with a dozen student logins on it generates no press, and a teacher who runs a good term of Eco has no reason to publish anything about it. But it does mean that nobody writing about Eco as a classroom tool in 2026 — this article included — can point at a room and say: there, that one, last spring.
Which is why the best story anyone can tell about Eco in a school is still the one John Krajewski told from a stage at GamesBeat Summit in April 2019. One class played through a world and used every available tree to build their houses. The next class, joining the same persistent server, arrived to a treeless world — and knocked down the first class's houses for the materials. Nobody set that as an exercise. No teacher stood up and said resources are finite. The first cohort's consumption was simply still standing there as the second cohort's constraint, because the server does not reset between periods, and the second cohort's answer — take the timber from wherever the timber now is — was precisely the behaviour the first had modelled for them. The lesson was delivered by the state of the world, on a delay, to people who had not been in the room when it was caused. It is Eco's entire pedagogical claim compressed into an anecdote, and it happened seven years ago, in front of an audience, to students who have since left school.
Set the recognition record beside it and that stops in the same year. The Department of Education's own success story lists, in its own compilation: the Grand Prize at the Climate Game Challenge at the Games for Change Festival in June 2016; People's Choice at Earth Games in May 2017; VentureBeat's Game of the Year for 2018, in December of that year; and in September 2019, an invitation to the UN Climate Summit as an intervention on the climate crisis, with coverage in a UN press release, USA Today and the New York Post. That is the funder's listing rather than four independently checked citations, and it is a good record — but it is also four entries across four years, ending exactly where the classroom evidence ends. Eco's peak institutional visibility and its last documented school both sit in 2019.
The product, meanwhile, has never declared itself finished. Eco is still in Early Access on 1 August 2026, more than eight years after the 6 February 2018 launch. It has no DLC, but it carries Steam's In-App Purchases label and sells cosmetics for Eco Credits alongside the base purchase. The announced 1.0 destination, Eco Infinite, is reported to have been revealed in 2021 — on the studio's own site it survives only as a stale developer-blog title — and it still has no release date. A long Early Access is not a scandal by itself; plenty of good games have had one. It does mean that anyone adopting Eco for a term is adopting a build nobody has called done, aimed at an endpoint named five years ago and never scheduled.
None of which describes an abandoned game, and the numbers say so. Steam's all-language standing on 1 August 2026 is Very Positive: 13,984 reviews, 11,290 positive against 2,694 negative, or 80.7%. Filtered to English it reads Mostly Positive, 78% of 6,608. The last thirty days sit at Mixed, 58% — on a recent sample of 39 reviews, which is far too small to describe as players turning on the game and too small to read a direction into at all. Updates ship. A playtest branch is live, the wiki's maintainers have begun updating it for 14.x, and a major release is dated for 6 August 2026. What has gone quiet is not the game. It is the education programme around it.
So the frame that survives the evidence is this: in 2026, Eco is a documented mechanism you would be operating yourself, not a programme you would be joining. There is no institutional scaffolding behind it, no current case study to copy, no dashboard waiting for you, and nobody at the studio whose job is your class. What does still exist — citizen accounts, a private world you control, statistics you can pull off your own server — exists as software and a wiki page, not as a service with a person behind it. That is not disqualifying. It simply moves the entire weight of the thing onto whoever decides to try it, which makes the remaining question a strictly practical one about what that person would be taking on.
What It Would Actually Take to Run Eco With a Class
The instructions exist, and they are on the developer's own wiki, under Education, in seven numbered steps. That page was last edited on 9 July 2024 and nothing on it has been revised since, so everything below is what the studio documented then rather than what it would quote you today — but nothing has replaced it either, and it remains the only public how-to there is.
The flow it describes runs like this. You buy a standard licence for yourself, because you need one to create a world. Your students get citizen accounts instead: cheaper accounts, locked to your world and no other, issued for a duration you choose, defaulting to 30 days with monthly extensions available. You get server space — rentable directly from Strange Loop Games by the month, or from another host, or self-hosted on a PC that has to run constantly, and the page recommends renting for one-off educational use rather than volunteering your own machine. You download the server package and run ecoserver.exe. Then you email Strange Loop Games to order the citizen accounts, supplying your Strange Loop Games account ID, how many accounts you want and an address to send them to, and you pay by PayPal only. When they arrive you assign a name and password to each one under hosted worlds/citizens, pasting in the server ID from your network.eco file so each account knows which world it belongs to. Students log in at accounts.play.eco/login and download the client. You join by typing in the server's IP or hostname, usually on port 3000.
Read that back and notice what kind of transaction it is. There is no education portal, no purchase-order flow, no seat-management console — there is an email address, a human on the other end of it, and PayPal. That is the shape of a small, manual arrangement, and it is worth knowing before you plan around it, because everything about it assumes low volume and a person willing to wait for a reply.
Now the bill, using the page's own July-2024 figures. A citizen account was priced at ten dollars, one third of what the standard licence cost. Server space from the studio was fifteen dollars a month. So a class of thirty comes out at three hundred dollars for the accounts, one standard licence on top so you can make the world in the first place, and fifteen for the first month of hosting: roughly $345, ordered by email and settled by PayPal, before a single student has learned anything. As education spending goes that is not much — it is a rounding error against a set of textbooks. Two things about it are still worth flagging. The accounts run 30 days by default and extensions are sold monthly, but no extension price is published, so a term longer than a month is a cost the page does not let you calculate in advance. And those numbers are two years old and unconfirmed; treat the $345 as the right order of magnitude and the right structure — per-student, plus a licence, plus rent — rather than as a quote.
The rent is the part people underestimate, because it buys something unusual: continuity. Eco's clock runs in real time, and the world keeps existing whether anyone is logged in or not. Trees grow, crops mature, sewage seeps and the countdown counts while your students are in double maths. That is exactly why the thing teaches — the state of the world on Thursday is the consequence of what happened on Tuesday, and no one gets to reset it — and it is also why "a PC that must run constantly" is not a throwaway line in the setup notes. If the box goes to sleep at four o'clock every afternoon, you have bought a save file, not a simulation.
Then group size, which is the decision that most determines whether any of this works, and the one the evidence is thinnest on. A server host that runs Eco worlds professionally puts the floor bluntly: it can work with groups as small as five if you tweak the settings enough — with the tweaking doing real work in that sentence — while many servers carry thirty-plus regulars and a few run past a hundred. That is a direct observation rather than a measurement, and it is the most useful number anyone has published, because it brackets the range: five is survivable with the game bent around you, thirty is where a settlement behaves like a settlement.
Note also that you do not need everyone online at once. MaxConnections, set in Network.eco, defaults to roughly a hundred — that figure is documented by a third-party host rather than by the studio, and it is an editable default, not a ceiling — and what it caps is simultaneous connections, nothing else. A world keeps property, skills and economic history for every citizen who has ever joined it, so two class periods can share one planet without ever meeting, which is precisely how the most famous thing that ever happened in an Eco classroom happened at all.
A healthy all-time review score is a verdict on the product, not a headcount for any given world, and what the public game's population does not tell you is whether your world will be busy: 739 people were playing worldwide at one instantaneous sample on 1 August 2026, and recent reviews from players with anywhere between 82 and 1,178 hours describe public servers as thin. One reading and a handful of complaints are not a trend, and they say nothing about a private world with your own students in it — but they do mean that dropping a class onto somebody else's server instead of running your own is a gamble on strangers.
The calendar is the next thing to settle, and it is a single line in a config file. The disasters section ships with generate meteor = true and meteor impact in days = 30, and after thirty real days the meteor lands and the world ends. The Education page says, in as many words, that for some educational purposes it might be better to change those settings. Take that seriously, because a term is not a countdown. Thirty days from a Monday in September expires in the middle of October, taking the world with it regardless of where your unit has got to; and a class that meets twice a week gets roughly eight or nine sessions inside a deadline built for people who log in nightly. Stretch the timer to fit the term, or switch the meteor off and let the settlement be the point, and make that call before you generate the world rather than after your students have built something they care about.
And then the wall, which is where classroom Eco actually fails, and it has nothing to do with ecology. Strange Loop Games named three design pillars for this game, and the third was eliminating grinding. It is the one current players say it misses hardest. A reviewer with five hours on record gave up inside the non-skippable tutorial — glacially slow, they said, with the construction UI refusing to cooperate. Another spent thirty-five of their forty minutes staring at load screens. A third, thirty hours in, advises simply keeping the wiki open while you play, and reports two-and-a-half-hour chains of prerequisite tasks standing between you and the thing you sat down to do. Those are adults who chose this game and paid for it. A fourteen-year-old with one period a week and no particular investment will hit the same wall harder and quit sooner, and they will quit before reaching anything the article above has described. The Education page's own advice is the right response and is easy to skim past: play the tutorial yourself, at least once, before your students touch it, and run the first session together in the same room, because a joined start minimises beginners' errors and gets people teaching each other instead of flailing alone. Budget the first session for the interface. You are not losing teaching time; you are protecting all the teaching time that comes after it.
What you get for all of that is narrower than the marketing and better than it sounds. You get a world whose statistics you can save off your own server — the page names population growth, sea level and air pollution — and which it describes, correctly and unglamorously, as convenient material for discussions and exercises. That is the assessment surface: not a grade the software computes, but a record of what a group of students did to a planet, in numbers they generated and can argue about. It is also, exactly, the thing the Department of Education's pilot found moved — data reference, economy and development, environmental strategy — and after 6 August 2026 the most legible argument in the file will be cumulative ground pollution set against where people can still afford to live.
So make three decisions before you spend anything. Group size, because interdependence is the lesson and a thin group either seizes up or ends up playing with the mechanic that enforces it switched off — decide how many bodies you can put in the world before you decide anything else. The calendar, because the meteor timer is a setting and your term is not thirty consecutive days. The first session, because the tutorial is what loses people, so spend that hour on the interface with everyone in one room and accept that no ecology gets taught on day one. Then buy the licence, order the accounts, and start the world — and judge the whole experiment by one thing: whether, by week three, your students are arguing about a rule with the server's own graphs open in front of them. That is the systems reasoning the one controlled test of this game found it moved, and it is the thing worth the $345 and the wall.