This set of guiding development and business principles is in no way set, and may be revised over time as experience reveals new truths.
Cooperate and conquer:
- Humans are best at creativity, and machines are best at redundant or mundane tasks. Both tend to specialize individually, but humans tend to have greater flexibility and range in more than one specialty. Perform or delegate all tasks accordingly.
- If any task is done the same way more than twice (within predictable variations) , it should be automated with machines. Start creating machine automation tasks for it the second time you do it, and increase those tasks’ priorities each subsequent time it is needed.
- Automating the same task for others more than twice is itself a task that requires automation. It is the communication between the task’s author and the machine’s performance that requires some form of automated improvements. If anyone asks you to automate it for them again, that means your interface to the machine isn’t useful enough yet. Computer programming language designers don’t realize they have all failed at this to date. The fact that human-machine translation is considered a stand-alone specialization called “computer programmer” is proof alone of their failure. Interface designer is a more fitting notion, but automating the complete human-machine task translation is the real goal.
- The act of newly automating tasks requires the most flexibility and creativity, so that should always be each humans’ first priority.
- If any two automated tools in the world are authored with similar components or results, these tools should be broken down into reusable components and shared. Hiding your proprietary tools from anyone with common needs is a foolish waste of your effort more than their’s.
- Any tools, processes, or documents created in the progress of business that are not your core market distinction should be openly shared with the world. You are really just tricking the world into helping you update and maintain these valuable items, while you lose nothing. All markets are implicitly transactional, thus they are all social. Attribution and public confirmation are of superior value to notions of proprietary ownership in all social contexts.
- Anything that is an obfuscated market differentiator for another company — especially one you are dependent on — should be recreated openly to share with the world. Openness provides better competition and freedom from dependency than you can ever create alone. If others are similarly dependent on your market differentiator, then open it and find a different way to be unique. Beat them to changing the market while maintaining a defensibly unique position.
- All obfuscation of useful information is temporary at best. Data must be revealed to be useful, so it is unprotected to every user at their first time of use. Data can be copied infinitely at each time of use.
- This is why rude DRM is impossible. The only secure way for any license purchaser to DRM-encrypt anything is by politely asking each user to delete their key and any decrypted copies after their license expires. The only secure way to enforce such deletion is by legal contract, never via hardware encumbrance.
- Obfuscation of data that isn’t proven outwardly useful is similar to deletion — like a lock with a key that no one ever bothers to use. The key might as well be lost, rendering the lock useless. If you intend it to be seen, why even lock it in the first place?
- The best protection of privacy you can ever hope for is strong locks with careful key management among trusted friends, and frequent lock changes. Unless your are both willing and able to manage all keys of the available lock, you can not protect the data with any lock. You can never be sure that anyone else is both able and willing to manage these for you.
- Only deterministic machines can be trusted to do what they are told, and only then when you completely understand their process and limitations. Making this process understood is a task to be automated (see 1.b.).
- All keys are just as permanent as their locks, and can be copied infinitely. There is no such thing as an “original” key that is different from its copies. No key can be destroyed, except in its use by destroying its lock. Thus keys can only be given, never lent.
- Even in abstract systems like encryption, the same lock will always open to the same keys, and keys are infinitely copyable by their users, so to eliminate future access you must change the locks. This makes the balance between security and access very difficult. To ease this burden, keep the current key to each lock for each valuable item inside shared-key lock boxes, and give each user a unique key to the box. As the valid user groups change, you only have to change the box rather than every key. As closed items or locks change, just change the key inside the box rather than the whole box.
- The only way to catch fraud is to record it as it happens. A fraudulent transaction without a log record is a crime without a witness. Even innocent mistaken transactions cannot be reversed without a record. Send auditors their log copies as soon as transaction entries are created, lest the local log is destroyed while the fraud is in progress. This means all important transactions should result in multiparty streams of log data.
- Auditors should all have access to transaction data from the point of transaction, so auditors must be chosen carefully (see 8). Encrypting transaction details within the log separately permits one to reveal partial logs gradually, as appropriate according to each auditor’s knowledge. Release keys of private details to auditors only when each fraud investigation requires it, and not any sooner (see 8).
- Confirming the intended results of transactions requires knowledge of any sources and transformations expected in each transaction. You can never predict when a fraud investigation will require reproducing all the transaction steps. In data science terms, this means both the source and result data are always immutable in secure transactions, and the algorithmic conversion process must both be well understood and immutable over time (ie. version controlled data, code, and documentation each with time stamps).
- Secure data is immutable data. The only sufficiently secure deletion is to first lock the data with a secure key, destroy that key, and allow overwriting by other data. Thus only the key and manifest are mutable, only when deletion is the desired result. (Refer to 6, 7, and 8.a. for full compounding impact). Irreproducible point-in-time random scrambling, similar to a random one-time pad with no pad storage step, is preferable to mutability in deletion (see 8). All securely stored data is thus inherently immutable until full deletion.
- Security is easily broken: just distribute keys publicly, or to any 1 unintended recipient, and the lock becomes useless (see 8). What is already public cannot be made private again (see 5).This means all data should be stored under an assumption that security is required, encrypted and logged. Make stored data public (or mutable) only after a proof of overriding human intention. Thus, all data should be assumed encrypted and immutable by automated storage systems, until proven otherwise. Intentionally public data is also best preserved by immutability in version tracking, so data should be assumed immutable in either case, until deletion is explicitly requested.
- The time necessary to create novel things is unpredictable, by definition of their novelty. Thus their creation cannot be scheduled.
- Only repeatable tasks are predictable at all. The only repeatable tasks humans should perform (see 1 through 4) are unique to their individual and time-dependent needs: meals, sleep, exercise, and socialization. Schedule only these tasks, and leave the rest (such as work) unscheduled until execution.
- “Deadlines” are a detestable human fabrication, invented to trick their adrenal glands into action. This farce should not be continued, for human health if nothing else. While we are all mortal, no death is sufficiently predictable. If the time of apocalypse were known sufficiently ahead, it would be a goal not a terminus. If it’s not worth doing now, then is it really worth doing at all?
- Adrenaline’s greatest side-effect is fear, which just leads to irrationality, and the resulting stress leads to an unpredictably sooner death. Give creativity the time it is due, and live longer. Experiencing the joy of creation will extend your life far more than the stress of planning.
- True and novel creativity benefits all others. Rote replication is only valuable after its automation has already been created (see 1 to 4). Value humans over their machine creations accordingly.
- The true value of money, in all its forms, is just as another metric. Capital was originally intended as a metric toward the greater good. Don’t make it a vanity metric, or you’ll be the last to use it.
- Unless everyone around you agrees with a monetary representation, and voluntarily shares in it with some sense of equality, it is worthless. Trinkets are measurable, but they are not wealth. Even when money measures a form of wealth, it is never very accurate.
- True wealth can never be accurately measured or exchanged, only given. Shared experiences, knowledge, and gratitude are among the greatest yet immeasurable forms of wealth.
- Your time is worth far more than mere money. Remember that only one form of recompense is monetary. Experience is a form of recompense that cannot be repossessed.