Another 85+ Ideas for Computing https://samsquire.github.io/ideas2/
This is a list of new ideas since my last 100 ideas. I hope you enjoy them as much as I did thinking about them and writing them down.
Contracts should be digitised so that anyone on any side of any contract can see what they are responsible for and any actions they can take within the contract. There should be a UI for interacting with the contract. If I have a contract with a company and can legally fetch my private data, I should be able to request that data through the UI. Or if I have a phone contract and I am eligible for an upgrade, I should be able to upgrade through the UI.
Existing:
Someone with a phone should be able to just create a business that satisfies all reporting requirements and is a legal entity. I want to see a mobile phone app that implements enterprise resource planning for the gig economy.
Computational problems can be rendered into labelled graphical video streams upon which people vote. People can then solve computational problems with a television screen, monitor or phone screen by looking at numbered graphs and using a keypad that they enter sequences of numbers into. Problems like efficient scheduling of cars for carpooling or deliveries are solvable by human beings quickly with a user interface that allows one to lasso points on the map and add cars. This is for problems that are not strictly numerical. The goal is to give everyone with a screen, monitor or television a source of income.
We need a way to define work. All work can be represented digitally: take A to B, assemble X with Y. This standard is that a piece of work must fit on an A4 piece of paper. A bit like a job description but each page represents a task. So you have a piece of A4 paper describing one piece of work to do. Each piece of work can depend on other pieces of work. Can be integrated into a continuous integration pipeline for human beings. Work appears in your work inbox.
Existing:
People can insert demands for products and services into a demand marketplace within a geographical location. Examples:
In aggregate, many people create demands around geographical areas. This data can be visualised as a smattering of circles around various central points representing each person’s home. The intersecting demands where multiple people want the same thing in the same area it should be obvious what satisfies most people’s wants. This data should be surfaced and downloadable.
A want app exposes the market place of wants and allows people to contact others based on their wants. If people published wants accurately and tracked possessions, they could automatically organise swaps with others based on what they want.
There’s a single app for following any dream or any lifestyle with user created content. You pay money per month for some thing you want to do, maybe a lifestyle of healthy eating, eating a certain way, learning a langauge, niche hobby, sports or gym. Maybe you want to become a musician or you want a lifestyle where you live in the suburbs. The company sends you recommended decisions to make, materials and gear that you need for your level of subscription. Users and companies create and maintain lifestyle paths to accommodate every skill level and combination of lifestyle.
So if you said you want a lifestyle where you can get to work within 10 minutes walk, the app would tell you where to move to and who to apply for jobs with. You could subscribe to a coffee everyday. Or if you want to regularly eat out at an upmarket restaurant, you can because your level of subscription includes a set amount of meals a month at a premium restaurant. Can deploy futures to meet demand, such as futures on restaurant tables. Can browse lifestyles in a digital online shop.
https://lifestyle-subscriptions.com
A filled inbox of questions to answer and questions generated based on my answers to previous questions. Uses random lists of words to ask interesting questions such as “Is thingA better than thingB” or “Is thingA meaningful with thingB?” or “How much money do you have in your bank account?”. The data collected could be used to implement life engine and can be used to study.
Existing:
It should be possible to make decisions in a company and run a company via presentation software if slides accurately represent processes of the company and can control a company. Presentation software can be an interactive system. I imagine mobile devices connecting over Wifi. People can run a company via presentation by creating models (flow diagrams) and frameworks (smartart) that are executable by a workflow engines. How companies say they are going to run, is actually how they run. There could be an app store of backend integrations with presentations for showing information that is up-to-date and live; such as pulling in sales data from internal systems and making changes to backend systems.
Human suffering could be eliminated if everyone had a virtual placement assigned. Everyone can have what they need and want if every other thing is at the right place at the right time. Place is virtual. There are virtual places such as receive X amount of Y food at Z, shop at X, sleep at X, be at X, receive X, go to X, take A to B, Complete X amount of Y work. Do Y at X time. Everyone has a set of places at any given time. Not everyone needs a virtual place at all times. If virtual places are scheduled, traffic, commuting, waiting can be eliminated. Placement is representation of a position to take in the world - it could be a place in a queue. Everyone can buy and sell placements and schedules. Basically every thing is a market and it can be financialised.
Depending on a library could also involve pinging the libraries’ software infrastructure so that whenever the library is changed, the dependents are also rebuilt. This would catch code that causes downstream changes to APIs. API changes are expensive.
Existing:
GUIs can be declarative. GUIs can be represented as a commutative set of statements. In the form of widgetA leftOf widgetB. Represent the commands that generate GUIs as a monotonically increasing set of statements about the GUI and declaratively render a GUI based on propositions of widgets in relationship to one another. Has the property of being able to extend existing GUIs by simply making more statements. See additive guis Imagine if a set of tweets could render a GUI.
Wouldn’t it be nice if you could recommend a song to play with a web page as background music. Perhaps a meta tag with a URL to a YouTube video.
The entire GUI could be generatable as a composable query.
A GUI is a two dimensional representation of a multidimensional data structure so a query is a plane slicing a high dimensional data structure.
As the user mouses around the GUI, a URL should be generated which is a representation of what the user is doing at any point in time including all context from the user’s perspective. This is like a very deep link into an application, as the user is a resource. See user urls on github
Replace job interviews at individual companies with interviews with an Accreditation app company to vet people for a certain kind of role. People can look at shop exterior and see a logo of an Accreditation company and work there if they have been vetted by the app. Transport can be branded by Accreditation company and accreditation app companies can specialise in the quality of the people they hire.
Desktop computers could be accelerated by doing things remotely. If a user’s filesystem is mirrored elsewhere, it can be searched remotely and results served locally without incurring any performance penalty locally. See my cloud desktop repo on github
An app that downloads data from various sources online to fill 1 gigabyte of data (reddit, blog posts, articles, free books), implements a viewer and always gives you something to read while you’re offline.
A CMS for webapps: create serverside and clientside routes, register handlers for the server and client side routes, create facts about the GUI.
Would be best integrated with 11. Additive GUIs
The OSI model makes a whole deal of sense, one layer adds promises on-top of the layers below to create an emergent complexity. The network stack manages to offer some very complicated behaviour with the illusion of simplicity at layers above. Each layer makes promises to other layers. What if normal software could take advantage of this property of network stacks by breaking up each responsibility into layers and use wrapped message passing to solve problems.
What if we had a visualization of a multi layered spreadsheet whereby each layer is a layer of the architecture of the software. Instead of formulas calculating an answer, we have message passing between cells that represent promise keepers. So if you wanted a CRUD system, you would send a message to a CRUD service and configure the CRUD service with your messages. You can explore different layers by going up and down the stack whereby the communication between services is at a different layer of messages.
Credit cards can be used to access buildings.
Churn is painful in software development. Things change without warning and things no longer compile or work together. I would like to see a more mature strategy to breaking changes and how they ripple through a software ecosystem. Software change is not cheap. Python 2 to Python 3 is some thing I’d like to avoid going forward. I want to depend on some thing that shouldn’t break, that means depending on an abstraction or an interface but is depending on some thing much larger than a single interface in code. I call this depending on a promise. The interface I depend on should not break between major releases, it’s stable and it’s locked into the software. Javascript in the browser keeps working but my NodeJS, Ruby and Java projects break all the time due to refactoring being done on underlying dependencies. Developers need accurate statistics of code coverage of who is using what. They also need to know if they’ve broken some thing downstream with their changes.
I propose code coverage tools are used to create a trace of API usages ands software projects can depend on these traces to work by sharing the trace file with upstream. Tooling would ensure that an API would never break because we’re depending on a promise of an API.
Modern systems such as Elasticsearch, Consul require clusters of computers to install against. What we need is installers that can be configured to install across multiple machines, not just independent machines.
Existing:
Have you ever been looking at the output of a command and wanted to change the output of the command directly? Perhaps it’s Kubernetes output or the output of ip a.
You click on a piece of output text, it pops up in an editor and you change it. You then are told what you need to run to make the changed output a reality.
Could a programming language be designed for refactoring simplicity? Java does really well as being a language that supports automated refactorings. I want software breakage caused by refactoring to end. Perhaps the author of a refactoring of a broken signature change could provide an equivalent piece of code that acts as a bridge between the old way of doing a thing and the new way of doing the thing. This way people who upgrade can acquire the code that does the bridge and there is no software breakage.
Deeply nested YAML is hard to edit because you do not know what context you’re in. If you’re modifying Kuberentes YAML, it’s hard to know what is valid where. If you indent a block to the left you make the document invalid. Nested YAML editors stop you from indenting to the left by accident. Validating means that the editor tries to deserialize the YAML into the data structure used by the program and shows you that program’s interpretation of the input YAML to tell you it’s valid or not.
Every thing typed into a web browser should be saved in a new document that is indexed by website. This way we can have an archive of our contributions to any website. With browser synchronization, no data should slip through the cracks. We don’t have to worry if a site doesn’t permit a backup of our data.
Existing:
# 27. Call site refactorer
In a system with multiple microservices and use of shared libraries, it is difficult to carry out refactorings across codebases. The call site refactorer is a system that allows refactoring across code bases. It accomplishes this by identifying all uses of a particular signature and then showing every call site in a list. For this to work, automated refactorings must occur against the call site signature. Automated refactorings such as extract variable should work on every call site simultaneously, a bit like multiple cursor support in text editors.
All libraries and systems should be documented by examples. Examples that include import statements, filenames and any other basic precursor knowledge that should be known to run the example to be included so that anybody can run with it. I would argue that Flask and other simple libraries are so successful because they have a simple easy-to-run minimalist example on their home pages.
Have a lot of data to transfer to a mobile client to create a useful interface? Just send an sqlite database over the wire to seed the client. Then stream updates over HTTP.
I want an editor that looks like the Github issue editor but renders to plain text - in the format of this idea document and accepts this document as input. So I can click New Idea and have it render to this plain text markdown format.
I’d like my desktop environment to be social. A desktop shouldn’t feel so alone. I want every screen provided by the desktop to have a chat room associated. People can discuss the desktop and vote on things regarding the GUI.
Ever thought a piece of software did something right? There’s no central collection of patterns that are of software done right. I would like to see a website which collects each piece of software’s installation process and says what is right about it. Installation process is just an example. There’s other things that we can discuss who does something right or wrong.
Like a chat bot that messages you instead of you messaging it. As part of the Human Query Engine.
Run a key logger that logs data entered into every program on a computer to keep a record of personal data.
Can software development be done as a conversation? Spin up Docker containers from a chat room. Write SQL. Write JSX React components in chat. Write Redux actions in the chat room. Everyone runs the same code and anyone can make a change anywhere. Without touching a text editor.
A wiki that is integrated with a kubernetes or docker swarm environment with pages for various problems. I can simply begin using different kinds of software that I can find via the wiki without an active deployment step. I can surf into some software. The distributed computing page provides implementations and various interfaces for distributed computing that I can use. I can write code that depends on a problem page. Each page is like an API. I should be able to import a page in a programming language.
import problem.wiki/a*-algorithm
import problem.wiki/map
map = initialize_map("mapfile")
find_shortest_path(map, "start-position", "goal-position")
Store your unavoidable file based secrets in /home/${USER}/secrets and have a shell script called secrets-logout that uses tar and GPG to encrypt your file based secrets when you go home. In the morning, log in with secrets-login to reverse the process.
Do not use your name or your personal email address for logging into websites. Your username is 50% of what’s needed to get into your account. Instead, generate a new email alias and use that to login, such as name+1tvsdradv4@example.com.
Promises in a digital advert can be directly purchased and appear in a special account with one click. If I see a deal for a product at a certain price, I can click purchase on it and instantly purchase the service or product. You could be driving down a motorway and see an advert for a product on a banner and instantly purchase it with a button on your car wheel. What you’re purchasing is the promise specified by the advert. It could be a product at a certain price or more likely, a capability. Each purchase is itself a marketplace as it could be that you’re buying from a reseller, you’re buying a promise, which means any product that satisfies the promise can ultimately be provided to you.
Instantly purchase an advert. Software advertisements that contain code to install themselves and a puzzle fitter that works out how to integrate the software with your infrastructure. One click installation and integration into continuous integration.
A ticket is for saying where you’re trying to go, not that you’ve paid. Ticketless travel is possible through credit cards loaded with digital tickets and NFC.
Imagine if a web request response had to be constructed by a human being. The incoming request sits on a queue. The UI has conveyor belts that take relevant information to the human operator. The human operator interacts with these request objects and constructs a response object
A programming language whereby you import language features to depend upon. Python has a PEP process and Java has JSRs. You directly depend on named groups of functionality and this determines your language experience. Perhaps the first line of the file after the shebang says what language features to enable, such as “feature1,feature2,feature3”
Territories can be occupied by people who believe in the same things. If enough people used an app and voted for what they wanted on the app and agreed to move towards where other people believe the same things, people could begin to get what they want. This idea is called world splitting because you divide the world in half when you agree with a policy.
Society learns that borders delineate acceptability. It’s what a neighbourhood is. What is acceptable or what people like within a set of borders, block level, strip mall level, street level, floor level. People bet on what they want, not what makes the most money and it becomes a form of voting which provides funding for said ideas for both sides. Betting forms the basis of rules that you have chosen to live by. Parallel societies - societies can fragment by what is wanted by each neighbourhood. You can create a new neighbourhood by living in a place with a set of beliefs, a placement company would try place you near to people who believe the same things as you. Everyone gets what they want.
A general purpose CRUD editor with the flexibility and power of a file explorer.
Can the state of an organisation be ran as a giant JSON structure?
Headless CMSes such as Strapi require expensive migrations to keep up-to-date. I’d like a CMS that promises to be compatible from early versions with minimal impact.
Take a single file with special comments to delineate directories and render in an editor as a folder with multiple files.
Titlebar of running applications reflects state of an upcoming build of the same application that is being compiled on CI.
When I open a number of programs, I would like this list of programs to be persisted so that I can shut down the computer and return to them again. I’d like the same for the terminal too.
Existing:
The ability to schedule disk space to various purposes and to schedule files to different regions of empty space.
I have OneDrive, Google Drive, Box and six hard drives. I want my data to be synced across all of them.
My virtual disks in my VMs should also resize (to grow) automatically.
When an application wants to show a tip, it should be crowdsourced.
Open source community could come up with an organisation that is digital and runs every thing via code.
Integrate with chatops and case management ops to offer jobs to people on mobile phones to do customer support via chat.
An IDE that guarantees if you can open the project, you can build it reliably the way others build it. Dependencies are fetched.
Like CURL but for querying APIs and handling pagination for you.
Every form of information collected by a corporation to go to a GUI server whereby data forms can be monitored and centrally tested.
If I was an desktop environment or office software vendor I would have a server for defining every information collecting screen and drive it from this place. Can generate documentation and do data collection centrally.
Another program in an Office suite: An editor for streams, such as Bash pipelines or Kafka streams. Would allow you to replay data between points in the pipeline and test each stage independently.
What can your computer do for you? The operating system and IDE advertises features and things it can do based on what is installed.
My computer and software I interact with raises events all the time. I’d like to be able to hook into events that are raised and add behaviour to them.
An interface to create asynchronous tasks definitions. I’m thinking of complicated workflows in the backends of controllers whereby multiple services need to be contacted and queried and where there are complicated error handling.
A office suite application for allocating resources to problems.
People should be able to do work via a mobile phone and be paid for it. We need to convert problems to be runnable by mobile phone.
People with the same title do different things. We could decentralize by voting what tasks constitutes a newly named role and eliminate burnout by creating accurate titles. We could identify who does most of the work in society.
Imagine a database of personal information, facts about yourself and various collected resources. I want to crowd source queries to this database to create interesting GUIs and for the data in the database not to require manual migrations. Development should always be carried out with an example dataset, so developers do not get used to started from empty all the time.
We can automate job descriptions by building GUIs and workflow rule engines.
Ordering food could be a commodity like web server software. We need an open source food ordering platform.
Monoliths can be stem cell servers - they can startup and choose what kind of servers to be. When software needs to communicate with a service boundary, it should go through a library. The library is configured to transform requests into either method calls, REST calls or MQ messages based on configuration to talk to other deployed instances of the service.
Communication libraries can be configured with retry logic and circuit breakers.
Use a web interface to render desktop environment screens.
Active network design software. Build a cloud network with a network designer GUI and render to Terraform or cloud API calls.
I want to allocate computing resources to different purposes. I’d like a GUI like spreadsheet that lets me allocate work to different CPUs and to different machines perhaps running containers.
My computer has lots of spare resources. My computer is powerful enough to run demanding applications because it is multicore. What my computer needs is models that take resources to run. I’d like a community of people generating workloads for computers to run.
There should be a menu in a desktop that is like a search for a task. So if you want to add numbers, you might be guided towards a spreadsheet. A shop of things that people can do with their computers. Not necessarily apps but the interoperation of multiple services installed to create some behaviour. For example, you could be guided through the process to create a website and then the following things happen:
"Now you have a website, the following tasks are open to you"
* Password Protect my photo album
"Now you have a database, the following tasks are open to you:"
* Setup [Accounts]
"Now you have a virtual network you can set up the following useful programs"
* File [Synchronization]
I can search for ways of doing things, such a search for “track spending” and then I am introduced to someone’s technical solution to this problem which is to show forms that collect spend amounts. If I don’t like this approach, I can try another one. The line between software installation and use of someone’s software is blurred.
My phone should be contactable via my desktop and by my laptops.
Existing:
A reddit or forum like website for managing large fleets of servers in a peer to peer driven way. Each post represents an administration task or some thing that can be executed on a server such as an Ansible task. For example, in the backup part of the website, there’s a post for beginning a general system backups and a nested tree of actual backups. There’s a post for each system resource such as listening ports and active connections.
Create machines in the cloud as easily as creating a post.
A community of thoughts around “wouldn’t it be nice if”.
See my repository https://github.com/samsquire/society for an optimistic view of the future.
A tool that either generates configuration files or command line options to try out different combinations of configuration options together. Use a hill climbing algorithm to see how each option affects run time performance or some other property of the system such as file size or safety. I had such a script that experimented with compression and encryption options of GPG but I have since lost it.
Model dataflow between infrastructure provisioning tools such as Terraform, Ansible, Packer, Kubernetes and execute the resulting graph in parallel.
See devops-pipeline https://github.com/samsquire/devops-pipeline
CSV of VMs/images to spin up in the cloud.
Embed a sub-spreadsheet in a row.
Log rich data structures and assert across the data structures that were logged by each machine. Could be used to test distributed systems. Use Complex event processing (CEP) to do unit tests across logs.
Write an end-to-end flow example log for your software and assign what machines carries out which log line so that communication between machines can be inferred by log entries by jumping between machines. For example, the following log definition CSV handles registration whereby the server contacts the email service to send an email.
Registering user :username,server
Sending verify email,emailservice
Handling order,orderservice
Generating invoice,invoiceservice
The log lines act as an interface to implement against and automatically pass context information automatically to the next service.
See https://github.com/samsquire/structured-logging-2-system
The ability to create a new environment by changing a single string. See devops-pipeline https://github.com/samsquire/devops-pipeline
Use things that are on the screen by right clicking on them and selecting ‘Use’.
Slice a block of contiguous code up and specify where each slice should run using comments. Introduces implicit communication calls to convey context data through the code. Write one program and split it up between servers via communication slices. You specify where each side should run, on the client, on a server or in a microservice.
# server /login
login_request = Request()
password = login_request.get("password")
username = login_request.get("username")
hashed_password = bcrypt.hash(password)
users = db.users.find(username=username, hashed_password=hashed_password)
if users:
response = make_response(redirect('/'))
response.set_cookie('session_id', s.session_id)
# server /
feed_items = db.feed.find(username=username)
response = make_response(render_template('feed.html', items=feed_items))
# clientside /feed
recommendation_service.fetch_recommendations()
recent_posts_service.fetch_recent_posts()
# recommendation_service /recommendations
recommended_items = db.recommendations.find(username=username)
response = make_response(render_template('feed.html', items=recommended_items))
# recent_posts_service /recent_posts
recent_items = db.recents.find(username=username)
response = make_response(render_template('feed.html', items=recent_items))
A spreadsheet of objects, rather than numerical cells. The code for a web application can live in a cell and communication to database and other services can be threaded through different cells. Would have a GUI where each cell can act like a REPL to your database, ORM system and CLI.
A reader that you can keep open on a monitor while eating without any mouse or keyboard input and it shows content and scrolls through content.
With data from the Human Query Engine and user generated opinions, we can produce highly personalised advice for people in the form of notifications in an advice feed. The advice should be highly personalised based on previous answers.
Take surfed pages and turn them into a spaced repetition feed.
When you need to make a decision, you open a program in your office suite called Decision. You fill out what needs to be decided and file it in a system. The people who need to approve the decision can do so as they get notified. Automated decisions based on rules can be defined.
My CPU has 64 threads in order to do things in parallel. I am looking for things I can use my CPU for.
Ideas that are not large enough for their own section.
Operations can be visually queued by interacting with objects that represent a pending operation as if it had been completed. See https://github.com/samsquire/gui-thunks
Spreadsheets that run across multiple machines for performance and parallelism. Cells can be configured to come from other computers.
A development server that can be the destination of any web form. Using the primitives map, filter, reduce, a spreadsheet that lets you see incoming data and transform it into a web page.
Spreadsheets are helpful views of data. Can open a database with a spreadsheet viewer and convert spreadsheet formulas into SQL.
A language where you can request properties of your data structure to select an implementation. This could enhance cost aware computing.
Imagine needing to import a sort algorithm, I can pass properties to the import command such as needing to import a sort algorithm that doesn’t care about stable sorting and can deal with partially presorted data:
import "sort" stable=true partiall_sorted=true n=50000
Every UI container should be tracking how many items are inside it. From this data we can decide what data structures and widgets to use for every UI widget on the screen. From this we can collect:
A marketplace of business space, business services, employment, products and services that can be provisioned by data and code through the commodisation of all business plans so that real life businesses can be created by software. Schedule people, buildings, vehicles from one central place.
A coffee shop can be represented as code and data structures and the systems needed to provision a coffee shop should be provisionable by code.
I am still working on this idea.
Imagine you have a very long lever, you can cause action at a distance. Deep Lever means to use a data structure to drive out behaviour far from where the data structure is defined to exert control over it. Examples:
https://github.com/samsquire/ideas#53-web-state-machines-for-testing
Have a table of data? Use Pivottables to change the shape of the data until it is convenient to process.
A GUI that is a live system that takes declarative rules of what should be where.
Every thing on the screen originates from streams that have been mapped and filtered.
There’s more ideas in ideas3.