Back to the table

52 Intermediate Tk Technology Worldbuilding

Technology

The tools and science that shape the world.

What it is

Invented technology is every ability your story world has that the real world doesn't, together with the rules that come attached to it. Some of it is the showpiece: the drive, the gun, the implant, the network, the cure. A lot of it never gets a second look: a door that opens by itself, a self-regulating cup, a form that comes in two copies with the second one going to the orbital authority. Those quiet items carry much of the load.

In science fiction, technology does the job magic does in fantasy. It sets the range of things your characters can do, and it sets the price of each one. Magic system sits on the opposite side of the table, and the rule the two share is explained there in fantasy terms: any made-up power needs limits and a price, and a power that can do everything stops being interesting. What follows is about the places where technology and magic split apart.

The biggest split is the yardstick. Technology grows out of something real, so the reader arrives already holding a ruler to measure it against. A reader will believe in a ship that crosses between planets in a matter of weeks. The same reader won't believe in a phone that never needs charging.

Here's the surprise. Readers almost never believe in a machine because of its physics. They believe because of the setup around it: who owns it, who has been priced out of it, who repairs it, which agency licenses it, and who sells the stolen copy on the black market. Settle those questions and the plot starts to write itself. You can skip explaining how the thing works, and readers will forgive you. They won't forgive you for not knowing who repairs it, who gets left out, and what it costs the person using it.

Magic gets invented from nothing. The only check a reader can run on it is whether it keeps to the book's own rules. If a spell lets someone speak to the dead on the third night after burial, the reader just watches to see that the book never bends that rule. Technology gets projected forward from something that already exists, so the reader runs a harsher second check: could this believably descend from the real thing?

Readers forgive scale much more easily than they forgive upkeep. A ship that crosses the solar system in six weeks gets a pass. A ship that never needs a spare part, a resupply or a refit doesn't, because every reader has owned a car. Scale is invention. Maintenance is something people live through.

Distance in time buys you freedom. A story set eleven thousand years ahead can claim almost anything. A story set in 2041 gets graded on every claim, since the reader is standing inside the present you're extrapolating from. The closer your setting is to now, the more you have to argue for your technology instead of just announcing it.

Contemporary and historical fiction don't get out of this. Think of a thriller about drone surveillance, a crime novel built on digital forensics, or a historical novel about a man who builds bridges. In each one, research takes the place of invention. Any mistake can be checked, and someone who knows will write to tell you. For the same reason, one of the two books listed below, Dan Koboldt's Putting the Science in Fiction, collects fifty-nine short essays by scientists, engineers, physicians and pilots about what they notice when fiction gets their own field wrong. Nobody could write that book for magic.

Arthur C. Clarke's third law says that technology advanced enough looks, to someone watching it, exactly like sorcery. He first put it in a letter to the journal Science in January 1968. It became the third of his three laws in the 1973 revised edition of his essay collection Profiles of the Future. The law describes an observer who meets technology far beyond anything he knows and can't tell it apart from magic.

Writers often turn that law into a permit. They treat the machine like a spell and skip showing where it came from. The permit has a price. Technology that no longer looks extrapolated gets filed as magic, and then readers hold it to magic's stricter standard of consistency, which is the only standard magic has. A machine can't be excused from physics and from bookkeeping at the same time.

You're allowed to dress magic up in technology's clothes on purpose. Commercial science fiction does it all the time: the psychic, the sword that isn't a sword, the ship that goes faster than light because the plot needs Tuesday's battle to happen on Tuesday. The trouble starts when you make that choice without knowing you made it, and then hear that your world felt thin.

For every invented technology, settle five things. Two of them technology shares with magic, two are where technology sets itself apart, and one is where your plot comes from.

Write these five down for your central technology before you start drafting, and file them in your continuity bible. A good deal of plot comes out with them.

Science fiction also gets sorted into hard and soft. The split is real and useful, but people often misread it as a ranking of quality. The term "hard science fiction" first appeared in print in November 1957, in P. Schuyler Miller's review of John W. Campbell's Islands of Space in Astounding Science Fiction. It means fiction that stays inside what's currently thought physically possible and shows its working. The physicist and novelist Gregory Benford describes his own hard SF as playing tennis with the net up: nothing goes into the story that conflicts with known science.

"Soft science fiction" has two meanings. The first is fiction that extrapolates psychology, anthropology, sociology and politics instead of physics and engineering. The second, less helpful meaning is fiction that puts emotional effect ahead of technical plausibility. The critic Peter Nicholls called soft SF "a not very precise item of SF terminology" and thought the hard and soft split was sometimes illogical. Nobody has settled where the lines fall.

For a working novelist, hardness earns you nothing on its own. A rigorous book that nobody finishes has failed. Ursula K. Le Guin's Hainish books extrapolate anthropology, get classed as soft, and rank among the most respected books in the field. Frank Herbert kept the machinery in Dune (1965) blurry on purpose, which left room for the political plot.

Hardness works as a promise you make through your packaging. If the cover, the blurb and the first chapter advertise rigor, readers will check your work, and one glossed-over point hurts you more than it would in a space opera. If the packaging promises adventure, nobody remarks on the gloss. So hard or soft is a reader promise decision. You make it when you position the book, and it binds you after that.

The labels show where a book puts its weight. They don't measure how hard the author worked. Soft doesn't mean lazy. Projecting a society forward is no easier than projecting an engine, and you get fewer reference books to help you do it.

In 2004 Geoff Ryman set out mundane science fiction in a manifesto he produced with a Clarion West class. Its rules: stay on Earth, use only technology that could plausibly grow out of the present, allow no faster-than-light travel, and let no aliens turn up to help. You don't have to adopt it. Write one chapter under those rules as an exercise, though, and you'll see how much of your invented furniture was decoration.

In her 1976 introduction to The Left Hand of Darkness, Le Guin called the novel a thought experiment and said science fiction is "not predictive; it is descriptive". Prediction is the business of prophets and futurologists. A novelist has a different job. That should come as a relief. No judge will ever ask whether your invention showed up in real life. The test is whether your fiction convinces well enough to carry a real observation about the present.

Why it matters

Every ability removes a kind of trouble, and every limit adds one. Instant messaging deletes the ally who can't be reached. Cheap, reliable healing deletes injury as a price. If your middle sags, count what your technology has made free. A world with no difficulty makes a book with none. Technology also hands you a second lever on top of that one: what's possible, and for whom. The sections below on who has it and who runs it come back to that. Worldbuilding makes the full case that limits create plot, so this short version is enough here.

The invention is the idea, and its side effects are the story. Readers will accept only a few new ideas at once. Three, each with its consequences traced, read as a believable future. Eleven feels like a costume shop. So change one thing, then follow it all the way out. The second- and third-order effects are the interesting part, much more than the invention itself.

Look for the traffic jam. Isaac Asimov made this point in his 1953 essay "Social Science Fiction", collected in Modern Science Fiction: Its Meaning and Its Future. Someone in 1880, he said, could easily predict the automobile, and would have a very hard time predicting the traffic problem. Fifteen years later, a Galaxy editorial by Frederik Pohl gave writers the version they know best: the good science fiction story predicts the traffic jam instead of the car. The line is usually credited to Pohl, who denied that it started with him. Who said it first matters less than what it tells you. The car is the idea. The traffic jam is the story. So don't stop at "what does my machine do?" Suppose it exists and it's cheap. Which jobs disappear? What new crime becomes possible? What kind of insurance shows up? What does a bored teenager do with it? What does a parent worry about at two in the morning?

Near-future technology goes stale. Magic never does. A spell written in 1968 reads exactly the way it did the day it was published. Near-future technology picks up a visible expiry date, and the closer the setting is to the present, the sooner that date comes. Philip K. Dick's Do Androids Dream of Electric Sheep? was published in 1968 and set in 1992. Editions issued after the 1982 film moved the setting to 2021. The year never mattered to the story, and changing it didn't keep the book from being outrun. The parts of a future that expire are usually the parts carrying the least weight.

The part nobody can predict keeps growing. Charles Stross gave a keynote at the Chaos Communication Congress in Leipzig in December 2017 and posted it to his blog the following January. He said he used to assume that ninety percent of the next ten years was already here. As of 2017 he put it nearer eighty percent already here, fifteen percent foreseeable but not yet arrived, and five percent "utterly unpredictable deep craziness". That unforeseeable share has grown, and it's the share that ages a book. For a literary novelist who writes one book a decade, that's an interesting puzzle. For a self-publisher whose backlist should keep selling for fifteen years, it's a business problem, since a book that feels like its period earns for a shorter time. The aging is lopsided, too. A 1968 novel is forgiven for being 1968. A 2019 novel isn't forgiven for being 2019.

Better forecasting won't save you. The fix has nothing to do with learning to "predict better". Three habits work instead, and the next three paragraphs take them one at a time.

Choose a specific invented detail over a forecast. A striking made-up detail lasts longer than a careful real-world one. "She paid the arrears on the implant before the license lapsed" should still work in twenty years. A named phone model won't.

Decide what stays unexplained. Be vague about how the thing works and exact about what it does. Most instinct runs the other way. Vagueness only helps you when you spend it on purpose.

Strip out the surface that dates. Product names, edition numbers, screen sizes, transfer speeds, calendar years and any statistic a reader can check go stale first. You rarely need them, because what the thing does to the person is the point.

Technology spreads unevenly. In fantasy, the circle of people who can use magic tends to stay fixed, by birth, by training or by keeping others out. The circle of people who hold a technology moves over decades, and it can move inside a single book. Here's a path worth inventing for yours. The thing starts as a lab prototype. Scarce units go to the state, the military or the rich. It gets cheaper, and people find uses nobody planned. Regulation arrives late and does a poor job. Black-market versions appear without the safety features, the license fee or the logging. Finally it sinks into infrastructure, and nobody notices it until it breaks. All along, the previous generation keeps working wherever the new one costs too much.

The future is already here, unevenly spread. That idea belongs to William Gibson, who says the future has already arrived but isn't evenly distributed. Its history is shakier than people tend to assume. The earliest printed version is in a 1992 San Francisco Examiner article by the journalist Scott Rosenberg, quoting Gibson. Gibson repeated the line in a 1999 interview on NPR, and he has said he had the thought long before he first said it in public. So the line is his, though it started out as speech rather than writing.

Put your characters at different points on the curve. Place different characters in one book at different stages of the technology's spread, and conflict shows up with no villain needed. One character has the new model. One has the old model. One carries a stolen unit that reports to someone else. One used to do the job the machine now does. Organized groups often form around control of access, so this is a good place to look for your factions.

Readers believe the people and paperwork around a machine. Science fiction convinces through what surrounds its technology, meaning the people, the companies and the rules, and almost never through the quality of its physics. Real technology comes with a maker who wants paying and a standards body that chose the connector. It comes with a regulator years behind, an insurer that won't cover modified goods, and a bored technician who has seen this fault before. Somewhere there's a supply chain with one point of failure in a country nobody in the book has ever visited. There's also a training course, a trade union, an instruction manual with a page missing, and a resale market. All of that is free. None of it needs another invented idea.

Institutions make friction, and friction makes scenes. Writers drop the institutions as paperwork and end up with a flat future. With no institutions there's no friction. Resistance is what makes an object feel real, and the story lives in that resistance. A gun anyone can pick up is scenery. A gun that needs a license, which needs a clean record the hero doesn't have, gives you three scenes. A quick rule: give every invented device one company, one regulation and one person who gets paid to fix it. Little of that will reach the page, but it will show anyway.

The reader needs only enough understanding to weigh the cost. Call this the competence question: how much does the reader need to understand for the climax to land? Explain too little and the finale looks like a trick, science fiction's version of the last-minute rescue. Explain too much and you burn the reader's attention on mechanism nobody needed. The reader starts skimming and skips the one paragraph that mattered. The usual test, "does the reader follow it", is the weaker one. The better test asks whether, when the climax comes, the reader can weigh what the decision will cost. Nobody needs to understand the drive. The reader does need to grasp that a burn this hard will probably kill the passenger stored in the hold, and the protagonist knows it too. Aim for clear stakes instead of full understanding. Stakes and cause and effect both lean on this, and each one fails when the reader can't judge the ending.

Too much explanation has had a name from the 1970s: the info-dump. In dialogue it looks like two characters telling each other things they both already know. You have four ways to deliver what the reader does need, and you can combine them. Exposition covers the general mechanics. The next four paragraphs show each one at work on technology.

Show it used under pressure. Let a character use the thing with skill, under pressure, for a reason, before you describe it at all. Chapter four of Orson Scott Card's How to Write Science Fiction & Fantasy, the other book listed below, rests on the classic case. In Robert A. Heinlein's Beyond This Horizon, serialized in Astounding Science Fiction in April and May 1942, a character walks toward a door and "the door dilated". Those three words explain nothing. They claim a whole future, and the story keeps moving. They do more work than three pages of description would.

Give the knowledge to a character who already has it. Put the expertise inside a viewpoint character who thinks about the technology in shorthand, contempt or daily irritation. Experts complain. They don't explain. A technician muttering about a coupling from a production run recalled for the third time tells you about the part, its maker, the agency that regulates it and the speaker's own skill, all in one remark.

Show the consequence before the mechanism. Show the harm or the benefit to a person first. Explain afterward, or never. Then the explanation reads as a reward instead of homework.

Lecture on purpose when the mechanism is the draw. Sometimes the mechanism is what readers came for: procedural science fiction, novels built around an engineering problem, readers who buy the appendix on purpose. Andy Weir's The Martian is the obvious case. The book works because every calculation is the main character trying to stay alive. Explanation that isn't also someone trying to get something isn't a scene.

Every invented machine argues something. An invented technology nearly always makes an argument, and readers feel it whether you meant one or not. The argument is what separates science fiction proper from a richly decorated adventure. Memory you can edit takes a stand on guilt. Free labor takes a stand on what people are for. A huge new ability that shifts no power, no money and no class takes a stand too, and a lot of readers won't buy that world, though they can't say why.

Chiang and Stross both read the fear of machines as a portrait of their makers. Ted Chiang's essay "Silicon Valley Is Turning Into Its Own Worst Fear" (BuzzFeed News, December 18, 2017) reads the tech industry's dread of a runaway optimizing machine as a self-portrait it never meant to paint. Silicon Valley, he argues, has "unconsciously created a devil in their own image". Stross makes a parallel case in his keynote: corporations are the oldest AIs around, and they've been optimizing badly for centuries. You don't need to preach. You do need to know what your machine stands for. If you don't, the book won't stay neutral. It will make an argument by accident. Theme is how you make that argument on purpose.

What a technology can't do is where the story is. The interesting part of any technology is what it fails to do, does badly or does wrong. Limits come in three kinds, and each is more useful than the one before. Hard limits are walls. The plot routes around them, and then they just sit there. Costs force a choice. Failure modes are the best of the three, because a failure is an event, and it comes with a countdown. Some samples: a healing device calibrated for adults gets used on a nine-year-old patient; a drive's coolant circuit was certified as serviced on paper with no work done; a network is a network, so someone reads it.

Most real failures have a person behind them. Each of those samples traces back to a human or organizational fault: a calibration decision, a tired technician, someone holding access. Real technology mostly fails through a person, a process or an organization, and rarely through physics. Failures rooted in institutions ring truer and play better on the page, because the blame can land on someone.

A limit can drive a whole run of stories. Isaac Asimov's Three Laws of Robotics took their familiar form in "Runaround", in the March 1942 issue of Astounding Science Fiction, with the wording credited to his conversations with the editor John W. Campbell. Read as engineering, the three laws are safety rules. Read as craft, they're a plot engine. The stories set the laws against each other, so a robot behaves in baffling ways until someone works out which law is winning. The limit is the story.

Real-world examples

Neuromancer (1984): the arcade, the payphones and the job market that lasted

William Gibson's Neuromancer is the richest single case for this element. It shows extrapolation that worked and the way a book ages, both in the same novel.

Gibson told The Paris Review in 2011 that cyberspace didn't come from computers. It came from walking around Vancouver and passing "a video arcade, which was a new sort of business at that time". He had never used a personal computer. What caught him was the kids playing console-style games built out of plywood. He extrapolated an attitude instead of a machine. The players acted as though a place lay behind the screen, and the computer-savvy people around him seemed to believe in that space literally.

Then he needed a name. On a yellow pad, with a Sharpie, he wrote "infospace, data space" first. He got to "cyberspace" on the third try, and he liked it because "it sounded like it meant something while still being essentially hollow". A professional picked a word on purpose because he knew he would never have to explain how it worked. The book never does explain how the matrix works. Instead, the novel is exact about how jacking in feels, and about what the hacker Case loses when he's locked out of it.

Some of the details have dated. The artificial intelligence Wintermute tries to reach Case by ringing a line of public payphones, one after another, as he walks past them. In 1984 that was an eerie picture of an intelligence reaching out through public infrastructure. Since then it has turned into a period marker. Nobody in the book carries a personal cellphone.

Writing in Ploughshares in April 2017, Frances Katz set those overtaken gadgets beside the part that held up: "the story of hackers for hire stealing corporate secrets and doing the bidding of faceless corporate entities is almost shockingly familiar". The hardware dated. The institutional setup, deniable contractors doing dirty work for bodiless entities, didn't. Gibson projected a job market along with a gadget, and the job market is what lasted.

The Expanse (2011 to 2021): one engine change and everything it set off

James S.A. Corey is the shared byline of Daniel Abraham and Ty Franck. The Expanse runs nine novels, from Leviathan Wakes (2011) to Leviathan Falls (2021). The series is the cleanest case of the change-one-thing habit you'll find, and it shows how much plot a single limit can produce.

The one change is the Epstein drive, a very efficient fusion rocket that makes trips between planets take weeks instead of years. On the Planetary Society's Planetary Radio in December 2019, Franck described it as basically a fusion reactor with a little afterburner. Abraham said plainly that they weren't reaching for a warp drive. The books have no faster-than-light travel and no other repeal of physics.

Watch what follows from that one change, in order. The engine is a plain rocket with no loophole, so distance still costs time, and time becomes politics. Trips between Earth, Mars and the outer settlements take days or weeks. Sending a fleet is a diplomatic statement, and nobody can get to a crisis fast.

Next, thrust becomes the scarce resource, so the limit moves to the human body. Crews ride out hard burns in crash couches with a cocktail of drugs in their veins, and a chase turns into a medical judgment about how much injury the passengers can survive.

Then there are the Belters. People who spend their whole lives in low gravity grow up tall and fragile, and they're uncomfortable on Earth. A Belter's body is a political identity before anyone says a word.

Last, the asteroid belt is within reach but harsh, so water and air become trade goods. That gives someone leverage, then a grievance, then a faction.

Very little of that is invention. Everything after the engine comes from demography, medicine, labor and resentment.

The series also shows restraint. It has a second speculative element, an alien artifact whose workings are never explained. The characters can't understand it either, so the vagueness creates dread instead of looking like the authors dodged a question. In the same interview, Franck said robots were left out because neither author finds them interesting to write. Choices like that come from what writers can build stories with, not from forecasts.

The Murderbot Diaries (from 2017): a machine that comes with a contract

Martha Wells's The Murderbot Diaries shows the institutional layer doing the heavy lifting. The series is the strongest recent evidence that what surrounds a machine holds more interest than the machine itself. The first book, All Systems Red, came out from Tor.com in May 2017. It won the Nebula Award for Best Novella in the 2017 cycle and the Hugo in the 2018 cycle.

The narrator is a SecUnit, a security construct built from machinery and cloned organic tissue. It was made, not born. It calls itself Murderbot, and it would rather be watching serials. Murderbot is no free agent. It counts as inventory, and the whole book hangs on that word. Murderbot comes supplied under contract along with a survey package. It carries a governor module meant to punish any action nobody approved. Murderbot has secretly disabled its own governor and works hard at looking compliant.

The clients it guards are academics from Preservation, a self-governing polity outside the Corporation Rim. They have to operate inside law and commerce built by companies for companies, and they took the survey contract because it was cheap.

The plot turns on paperwork. There's no rogue machine and no physics failure. Hazard data has been deleted from the survey packet. Sections of the map are missing. A document has been tampered with. The threat is a rival company, and the way out runs through a signal beacon, a compromised system and a contract. The paperwork is scarier than any breakdown would be, because every reader has been handed an incomplete record by an institution that didn't care.

Two moves are worth borrowing. First, ownership gives you politics at no cost. A leased device behaves differently in a story from one owned outright. Second, Murderbot describes its own abilities with the weary precision of a worker describing a job: what its scanners pick up and what they miss, what its contract makes it record, which of its own systems it doesn't trust. Murderbot is the knowing character kept up across a whole book. The detail never stops the story, because it's also building character.

Gattaca (1997) and Minority Report (2002): hide every surface or design every one

Gattaca was written and directed by Andrew Niccol and released by Columbia Pictures in October 1997. The film answers the aging problem. It comes from 1997 and doesn't feel dated. Its premise is that parents screen their embryos, so a person's genome is known from birth, and because it predicts so well, it gets treated as destiny.

The effects are institutional. A blood sample at the entrance decides who gets hired. Insurers sort people the same way. The screenplay coined a word for this kind of discrimination, "genoism", and the word later turned up in discussions of real cases. The protagonist invents nothing. His goal is to pass as someone else, handing over another person's samples well enough to get through the daily screening at work.

The production design holds almost nothing you could date. The electric cars are shaped like 1960s models, the Rover P6 and the Citroën DS. The headquarters is the Marin County Civic Center, designed by Frank Lloyd Wright, a 1960s building filmed as if it were the future. There are no invented-interface screens, no holograms and no visible year. The deliberate retrofuturism takes the film out of the timeline altogether. The film was nominated for an Academy Award for art direction, and the design was doing craft work as well as decoration. By refusing every surface that could date, it protected the one idea worth protecting.

Minority Report (2002) took the opposite route. In 1999, Steven Spielberg brought together fifteen experts, among them architects, computer scientists, urban planners and biomedical researchers, for a three-day meeting in Santa Monica. The production designer, Alex McDowell, turned what they worked out into a reference document of roughly eighty pages on the world of 2054, nicknamed the 2054 bible. The result is among the most influential designs of future technology in film. Even so, its gesture controls were being shown at trade shows soon after the film's 2002 release, which is a strange fate for a world set in 2054.

Both strategies work, and each one ages in its own way. Detailed surfaces get outrun. Withheld surfaces leave the idea intact.

An everyday scenario: the license renewal

Here's a made-up case. Picture a writer working on a near-future thriller. The protagonist, Ziva Ostrowski, has a retinal implant that lets her see in the dark. The writer has worked out in depth how it functions and explains it across two pages in chapter three. A beta reader says the world feels flat and admits to skipping the technical passage.

The writer runs the five decisions. Two of the five are answered: what the implant does, and roughly what it costs to install. Holders, governor and breakdown are still open, so the writer fills them in. Few people have the implant, because it was built for a mining consortium and the civilian version is expensive. Ziva's is secondhand, bought from someone leaving the industry and set up for another person's skull. The governor is Corbel Dynamics, the maker, which charges a monthly firmware license. An unlicensed unit keeps working for a while, then starts losing range at the low end. A man two districts over reflashes units for cash, but a reflashed unit can never be updated again. The implant fails in the cold and when the license lapses, and both problems are worse in a resold unit that was never fitted to her.

Now the two pages of explanation go. The information comes back as pressure on Ziva instead. Money is tight in chapter three. By chapter nine she's behind on her license payments. The reader learns what the implant does by watching her use it. The reader learns its price by watching her weigh the renewal against something else she needs. The reader learns who controls it by watching her think about the man two districts over, and about losing updates for good. A climax can now turn on the bottom of her range giving out in the cold, and the reader has been set up for it without noticing. Nobody is to blame. There's only poverty, a resold unit and a lapsed license.

The manuscript never needed more technology. It needed three institutional facts, delivered as subtext and inconvenience instead of explanation. The repair takes an afternoon.

Try it yourself

Start by picking the one technology that carries the most weight in your book. If you can't tell which one that is, you've already learned something. The steps below apply a method Suyi Davies Okungbowa lays out in a short SFWA piece on worldbuilding through extrapolation: change one thing, then trace the consequences outward into the economy, the law and everyday custom.

  1. Write the big idea in one sentence. Leave out the word "and". A model: "Embryos can be screened, so a person's future is knowable at birth." If you need "and", you have two ideas, and one of them is probably decoration. Mark the one the story couldn't survive losing.
  2. Fill in the five decisions on one page. Write the ability (narrowly, with something it can't do), the price, the holders, the governor and the breakdown. The thing it can't do matters more than three paragraphs about what it can: "it heals, but does nothing for burns".
  3. Do the traffic-jam pass. List ten second-order consequences. Write down what changed around the technology, not what the technology does: a job that disappeared, a new crime, a new way to cheat at school, an insurance product, a law that came too late, a subculture, a health problem, a worry for parents, a smell, a slang word. Keep three of them, the ones your protagonist would notice.
  4. Draw the diffusion curve. Make a timeline covering the last thirty years of your invented world's past, or all of it if the world is younger than that. Mark where the technology was a prototype, expensive, cheap, misused and regulated, where the black market appeared, and where the previous generation is still in service. Then put every main character somewhere on the line. Any two characters at different points make a scene you haven't written yet.
  5. Do the institution pass. For your central technology, write down five entries: a maker or licensor, a rule or license that governs it, a person paid to repair it, a weak point in its supply chain, and a route to the illegal version. Only some of these will be proper names, and most of them will stay off the page.
  6. Clear out the surface that dates. Search the manuscript for brand names, model numbers, screen sizes, storage capacities, transfer speeds, specific years and any number a reader could check. For each one, ask whether the scene needs the figure or only its effect. Replace them with their effects unless a scene truly can't work without the figure.
  7. Test the climax for competence. Write one sentence that finishes this thought: at the moment of the climax, the reader needs to understand exactly this much. Then check that the reader can weigh what the decision will cost. Nobody needs to understand the mechanism. If the ending depends on something the machine does that was never shown in a scene, you have a cavalry problem, and more explanation won't fix it.
  8. Convert your longest explanation. Find the biggest block of technical exposition in the book. Try three replacements and keep the shortest: skilled use under pressure, the knowledge held by a viewpoint character who grumbles in shorthand, and the outcome shown first with the mechanism left out. Then reread the chapter without the original block. About nine times in ten, the chapter loses nothing.
  9. Find the institutional failure. Write three ways the technology fails. Rewrite each one so a person, a process or an organization is at fault instead of physics: a calibration set for the wrong patient, a form signed with no work done, an operator who read the logs. Pick the failure that fits one person best and give it to a named character.
  10. Name the argument. Finish this sentence: "This story is saying that ____ about the present." The sentence doesn't need to appear in the book, and it shouldn't go into dialogue. You do need to know it, or the argument arrives by accident.
  11. Keep the ledger. Log every rule, limit, price and failure you've settled in your continuity bible, along with the chapter where each one first appears on the page. The ledger stops you from softening a limit in book three because the ending wants it. Paste this into your notes:

Technology: ____ One-sentence idea: ____ Can do: ____ Cannot do: ____ Costs: ____ Who has it: ____ Who does not: ____ Who controls it: ____ Fails when: ____ Whose fault: ____ At the climax the reader must understand: ____ This is saying ____ about the present.

Common mistakes and misunderstandings

Giving the hero a new ability near the end. An ability that shows up or grows in the last fifty pages feels like a cheat, even when nobody can point to the clause that was broken. Here's the test: could you write the finale using only the limits set up by chapter four? If not, set up the ability well before the end, with a price attached, or find a different ending.

Piling on specifications with no consequences. Three paragraphs on how the drive works and none on what it costs to run, who services it or who can't afford a ticket. Detailed specs don't add texture. A reader told the ticket price believes in the ship more than a reader told the thrust.

Never maintaining anything. No refits, no firmware updates, a colony water plant that has run for ninety years without a single swearing technician. Maintenance is realism that costs you almost nothing, and it drives plot too, because the person who fixes the thing knows what nobody else does.

Leaving the technology without an owner. A device nobody owns just exists, like weather. Real abilities have makers who want paying, governors who lag behind and sellers who cut corners. A device with no owner can't create politics, so the only conflict it can create comes from breaking.

Having everyone adopt it at once. Everyone owns it, everyone uses it the same way, and nobody keeps the old one. Real worlds have a hospital still running the previous generation, a religious community that refuses the technology, an area where no license is on sale, and a professional who prefers the manual method and does it better. Each of those is a character.

Reading Clarke's law as a license. "Sufficiently advanced technology is indistinguishable from magic" describes a bewildered observer. The law doesn't excuse you from doing the work. A machine that isn't extrapolated from anything gets held to magic's stricter standard of consistency instead.

Thinking "Hard" means "good". Rigor is a promise to a particular set of readers. It doesn't measure quality. The cover and the first chapter make the promise. In a space opera, a dozen impossibilities cost you less than one broken promise in a hard SF novel. Pick your book and keep the bargain.

Making the reader understand more than the ending needs. Authors who did deep research want the research to show. They lay out the whole mechanism and use up the reader's attention before the part that counts. Aim for a price the reader can see clearly at the moment of choice.

Inventing a character to receive a lecture. The newcomer, the reporter or the trainee who asks whatever the reader is assumed not to know is an info-dump in disguise. Give that character a goal of their own in the scene, so the lecture turns into bargaining. Otherwise cut the scene.

Forgetting that old technology sticks around. New abilities get layered on top of old ones. Paper receipts outlasted the systems that were supposed to end them. Cheaper, worse earlier versions serve wherever your protagonist can afford to be. A future that's all one vintage looks like a showroom.

Confusing a rule with a limit. "Faster-than-light travel requires a license from the Navigators' Guild" is a rule. A rule can be bribed, broken or ignored, and sometimes a rule is exactly what you want. "Faster-than-light travel is impossible" is a limit, and it forces the plot to find another path. The two do different jobs. Writers who need a limit tend to write a rule, then wonder where the pressure went.

Being vague by accident. Failing to decide isn't the same as choosing to be vague. Deliberate vagueness, exact about consequence and silent about mechanism, reads as confidence. Accidental vagueness reads as thin, and it gets summed up as "the world didn't feel real". That note is hard to act on, because no single thing is wrong.

Related elements