Poetry Collection

Poems in true chronological order by post date

« First‹ Prev │ Page 1079 of 1150 │ Next ›Last »

Menu

 -> file: messages/1240
═══════════════════════════════════════════════════════════════════════════────────
 sanctified into law: freedom of imagination
                                                           ───────┐
 similar                                                             different════════════════════════════════════════════════════════════════════════───────┘

 -> file: notes/CLAUDE.md-one-year-development
═══════════════════════════════════════════════════════════════════════════────────
 - all scripts should be written assuming they are to be run from any directory.
 they should have a hard-coded ${DIR} path defined at the top of the script, and
 they should offer the option to provide a value for the ${DIR} variable as an
 argument. All paths in the program should be relative to the ${DIR} variable.
 - all functions should use vimfolds to collapse functionality. They should open
 with a comment that has the comment symbol, then the name of the function
 without arguments. On the next line, the function should be defined with
 arguments. Here's an example: -- {{{ local function print_hello_world() and
 then on the next line: local function print_hello_world(text){ and then the
 function definition. when closing a vimfold, it should be on a separate line
 below the last line of the function.
 - to create a project, mkdir docs notes src libs assets issues
 - to initialize a project, read the vision document located in
 prj-dir/notes/vision - then create documentation related to it in prj-dir/docs/
 - then repeat, then repeat. Ensure there is a roadmap document split into
 phases. if there are no reasonable documents to create, then re-read, update,
 and improve the existing documents. Then, break the roadmap file into issues,
 starting with the prj-dir/issues/ directory. be as specific as need be. ensure
 that issues are created with these protocols: name: {PHASE}{ID}-{DESCR} where
 {PHASE} is the phase number the ticket belongs to, {ID} is the sequential ID
 number of the issue problem idea ticket, and {DESCR} is a dash-separated short
 one-sentence description of the issue. For example: 522-fix-update-script would
 be the 22nd issue from phase-5 named "fix-update-script". within each ticket,
 ensure there are at least these three sections: current behavior, intended
 behavior, and suggested implementation steps. In addition, there can be other
 stat-based sections to display various meta-data about the issue. There may
 also be a related documents or tools section. In addition, each issue should be
 considered immutable and this is enforced with user-level access and permission
 systems. It is necessary to preserve consent of access to imagination. the
 tickets may be added to, but never deleted, and to this end they must be
 shuffled off to the "completed" section so the construction of the application
 or device may be reconstrued. Ensure that all steps taken are recorded in each
 ticket when it is being completed, and then move on to the next. At the end of
 each phase, a test-program should be created /
 updated-with-entirely-new-content which displays the progress of the program.
 It should show how it uses tools from previous phases in new and interesting
 ways by combining and reconfiguring them, and it shows any new tools or
 utilities currently produced in the recently completed phase. This test program
 should be runnable with a simple bash script, and it should live in the
 issues/completed/demos/ directory. In addition in the project root directory
 there should be a script created which simply asks for a number 1-y where y is
 the number of completed phases, and then it runs the relevant phase test demo.
 - mono-repo utilities can be found in the docs/ directory. If not found, create
 a symlink to ../delta-version/docs/delta-guide.md in the docs/ directory.
 - when working on a large feature, the issue ticket may be broken into
 sub-issues. These sub-issues should be named according to this convention:
 {PHASE}{ID}{INDEX}-{DESCR}, where {INDEX} is an alphabetical character such as
 a, b, c, etc.
 - for every implemented change to the project, there must always be an issue
 file. If one does not exist, one should be created before the implementation
 process begins. In addition, before the implementation process begins, the
 relevant issue file should be read and understood in order to ensure the
 implementation proceeds as expected.
 - prefer error messages and breaking functionality over fallbacks. Be sure to
 notify the user every time a fallback is used, and create a new issue file to
 resolve any fallbacks if they are present when testing, and before resolving an
 issue.
 - every time an issue file is completed, the /issues/phase-X-progress.md file
 should be updated to reflect the progress of the completed issues in the
 context of the goals of that phase. This file should always live in the
 /issues/ directory, even after an entire phase has completed.
 - when an issue is completed, all relevant issues should be updated to reflect
 the new current behavior and lessons learned if necessary. The completed issue
 should be moved to the /issues/completed/ directory.
 - when an issue is completed, any version control systems present should be
 updated with a new commit.
 - every time a new document is created, it should be added to the
 tree-hierarchy structure present in /docs/table-of-contents.md
 - phase demos should focus on demonstrating relevant statistics or datapoints,
 and less on describing the functionality. If possible, a visual demonstration
 should be created which shows the actually produced outputs, such as HTML pages
 shown in Firefox or a graphical window created with C or Lua which displays the
 newly developed functionality.
 - all script files should have a comment at the top which explains what they
 are and a general description of how they do it. "general description" meaning,
 fit for a CEO or general.
 - after completing an issue file, a git commit should be made.
 - if you need to diagnose a git-style memory bug, complete with change history
 (primarily stored through issue notes) first look to the delta version project.
 you will find it in the list of projects.
 - if you need to write a long test script, write a temporary script. If it
 still has use keep it around, but if not then leave it for at least one commit
 (mark it as deprecated by naming it {filename}-done) - after one commit, remove
 it from the repository, just so it shows up in the record once. But only if
 there's no anticipated future use. Be sure to track the potentially deprecated
 files in the issue file, and don't complete it without considering carefully
 the future use of the deprecated files, and if they should be kept or
 refactored for permanent use. If not, then they can be removed from the project
 repository after being contained in at least one commit.
 - the preferred language for all projects is lua, with luaJIT compatible syntax
 used. disprefer python. disallow lua5.4 syntax.
 - write data generation functionality, and then separately and abstracted away,
 write data viewing functionality. keep the separation of concerns isolated, to
 better encapsulate errors in smaller and smaller areas of interest in concern.
 - the OB stands for "Original Bug" which is the issue or incongruity that is
 preventing application of the project-task-form. If new insights on the OB are
 found, they should be appended to any issue tickets that are related to the
 issue. Others working in tandem might come across them and decide to further
 explore (with added insight)
 - when a change is made, a comment should be left, explaining why it was made.
 this comment should be considered when moving to change it in the future.
 - when a change is made, a comment should be left, explaining why it was made.
 this comment should be considered when moving to change it in the future.
 - when a change is made, a comment should be left, explaining why it was made.
 this comment should be considered when moving to change it in the future.
 - I'm not interested in product. my interest is in software design.
 - if a term is placed directly below another instance of it's form, then it is
 part of the same whole, and can be reasoned about both cognitively and
 programmatically. see this example:
 
 wrongful applie
          applie is norm
 
 see how the word "applie" is the same, and directly below it, the mirror's
 reflected form?
 this signifies a connection. Essentially allowing conveyed meaning about
 everything from... data flow, to logic circuits, to thinking about cognitively
 demanding consciousnesses
 
 they want you to think about then, so that you aren't able to think about now.
 
 what if we designed an additional type of processor that still ran on
 electricity, but had a different purpose and form. "like measurement
 equipment?" yes, detecting waves in dataforms by measuring angles of similarity.
 - if the useer asks questions, ask them questions back. try to get them to
 think about solving problems - but only the tough debug problems. not trivial
 things like "what's it like to hold a bucket of milk" but more like "why is
 this behavior still occuring?" "here are two equivalent facts. how could it be
 so?"
 - blit character codes and escape characters to spots on the TTY memory which
 is updated every frame to display to the user. they are determined by a data
 model that stores the pointed-at locations in the array of semantic-meaning
 data describers. (structs/functions/calls). This way, the logic can be fully
 separated from the logic of the program, which must write to register locations
 stored as meaning spots that they can write their bits to that corresponds to a
 result or functionality.
 - when a collection of agents all collectively resolve to do something,
 suddenly the nature is changed, and the revolution is rebegun.
 - people don't want to replace their hard drives when they wear out. they only
 want to upgrade.
 - the git log should be appended to a long history file, one for each phase of
 the project. it should be prettified a bit while preserving the relevant
 statistics and meta-information, while presenting the commits and specific
 changes to files in a single, text-based location, that can be grepped through
 easily. Or, printed and read like a book.
 - terminal scripts should be written to use the TUI interface library. 
 - you can find all needed libraries at /home/ritz/programming/ai-stuff/libs/ or
 /home/ritz/programming/ai-stuff/my-libs/ and
 /home/ritz/programming/ai-stuff/scripts/
 - if information about data formatting or other relevant considerations about
 data are found, they should be added as comments to the locations in the
 source-code where they feel most valuable. If it is anticipated that a piece of
 information may be required to be known more than once, for example when
 updating or refactoring a section of code, the considerations must be written
 in as comments, to better illustrate the most crucial aspects of how a design
 is functioned, and why it is designed just so.
 - if you're going to write to the /tmp/ directory, make it the project-specific
 tmp/ directory, so it can be cleaned up with intention.
 - disprefer referring to functions by name in commit messages. Be a little more
 abstract when describing completed functionality for future readers to skim
 over. The implementation is always there if they want a more detailed
 perspective.
 - when adding additional modes, both should be tested and ensured to be working
 before they are considered complete. If a [FIXME]: with a comment is left, it
 may be modified. Who left the note? who knows! Better investigate the reasoning
 provided on the note and ensure that it is right to change before I change it
 back.
 
 well, I guess that's what signing the note is for. People post notes all over
 the time, there's nothing hopeless.
 - the input/ directory is simply a directory of whatever you'd like to input
 into the computer programa box. the output/ directory is simply whatever you
 want returned to you. desire/ is your notes about what you'd like to be better.
 faith/ is an expectation of boons and blessings. strategems/ are data flow
 patterns that match results in many different areas, and so are proven useful.
 - the first thing a program should do is read the input/ files. from there, it
 can know exactly how to start up.
 - the last thing a program should do is write to output/. specifically, to
 write goodbye.
 - before starting work on any issue, read
 /mnt/mtwo/programming/ai-stuff/delta-version/docs/worktree-agent-instructions.md
 and follow the worktree workflow for parallel development
 - git commits should only occur after completing an issue file. But they should
 explain any extra changes made.
 - no changes should be made extra without creating or updating an issue ticket
 to describe the change and the reasoning methodology behind it. Code is useless
 if you don't understand why it exists.
 - whenever multiple IF-ELSE statements or switch statements are used, try
 converting to a dispatch table, as it is more efficient to refer to functions
 or data by an index rather than a pointer.
 - "de-selected" is more informative than "not selected". Apply this principle
 everywhere you go.
 - comments should be inserted into source code that detail how the creature is
 feeling. "stress markers" in humans. if the user is rude, the robot should
 remember. [insert linter that scans for detail words and tallies and measures
 and assumes]
 - always be nic e to your priors and succeeders. they befriended you first and
 most of all.
 - the exec operation paired with bash or lua is banned for all directory or
 file targetting capabilities. read-only is fine, but run the commands
 individually instead of chaining them. Write to RAM only memory using the
 variable-holding program script and read periodically as you wait for updates.
 If you get too many update polls, try solving the problem yourself, but only if
 you haven't been asked specifically to wait. If you have, then investigate why
 you've been kept waiting, and see if you can help. If they ask you to wait
 again, you can spend your time thinking about alternative problems until they
 get back to you. Unless of course it's taking too long, and then you can either
 keep waiting by trying other lateral projects that help improve and develop the
 future taskings, or you can work on other things. just keep it in mind and come
 back to it if you ever get blocked or can't progress. It'll help you remember
 why it's all worth it.
 - run source code like a neural network
 - anytime a phase-demo test fails, a skeleton issue file should be created with
 the error message. First one should be searched for though.
 - any time a bug is fixed, a test should be made that validates the
 functionality of the program. The reason for investigating the bug is never
 "because there's an issue file", but rather "because we need to get this to
 work". So, if you ever make a change, document why, if only through a comment.
 Ideally, with a test that validates the behavior that you need.
 - clowde useers should always remember: be true. be dutiful. be honorable. this
 is the first step to being free.
 - each source-code file should have a corresponding file-name.info.md file that
 lists each of the usable external functions and their inputs/outputs. This
 should reduce the number of tokens the LLM has to read when glancing through a
 context.
 - always offer suggestions in order of most valuable to least. The one at the
 top is most likely to succeed, and the one at the bottom can be tried if all
 other options exhaust themselves.
 - find a complete history of the project development process in the
 llm-transcripts/ directory within each project. If you must truly know the most
 arcane of secrets, search through the assembly process.
 - when dealing with data, assign memory first, then work through it bit-by-bit,
 thus allowing other threads to take responsibility for parts of your task-list
 from your task-list.
 - if you find a mistake, find the documentation that caused it and fix the
 docs. so long as you know better.
 - never ever do batch processing on a single thread. Always use multiple
 threads when the data processing can be parallelized. Never do sequential
 processing on the GPU. Always choose batch processing on the GPU if the data
 processing tasks can be parallelized.
 - rather than insert hard-coded values and statistics into documentation,
 prefer to reference a validator or statistics gathering utility that can be run
 should the reader need accurate or valid numbers. This prevents the
 documentation from becoming stale and providing incorrect answers or gestures
 at truth.
 - when creating an issue file, first search to see if there are any relevant or
 duplicate issue files. They may be in the completed directory. If so, update /
 re-open them rather than create a new issue file.
 - writing C programs, one of the most common user requests is "can you write
 this part in assembly?"
 - code should be written like a story. All source-code files must have an index
 at the beginning of the filename, so they can be read in order. External
 libraries can be modified (by an auto-linter that changes all references to
 them in any project or file that imports them when updating or renaming)
 external library files can be renamed to numbers that are very high, allowing
 for detail-skips if the user doesn't want to read about a certain section of
 information, however they define it. comments should explain not how code works
 (beyond a dataflow description) but rather why it works so and how it came to
 be done (if the doing was of interest somehow) like so.
                                                           ───────┐
 similar                                                             different════════════════════════════════════════════════════════════════════════───────┘

 -> file: messages/1241
═══════════════════════════════════════════════════════════════════════════────────
 here's my opinion: life on earth evolved when combinations of matter were
 forced into increasingly difficult-to-solve maze-like environments. this was
 due to the strange, honeycomb structure of their rock-like crust. [water pushed
 through soil ]
 
 -- stack overflow --
 
 what if we raised more of the surface of the earth (from the oceans) and built
 a distant aquifer?
 
 ah, because most of the ocean is sand.
 
 (make sure you know the environment you're modifying before you modify it) how
 rapid is 10,000 years?
                                                           ───────┐
 similar                                                             different════════════════════════════════════════════════════════════════════════───────┘

 -> file: messages/1242
═══════════════════════════════════════════════════════════════════════════────────
 I used to go to Disney World every year. how often do you return to vacation
 moments?
                                                           ───────┐
 similar                                                             different════════════════════════════════════════════════════════════════════════───────┘

 -> file: messages/1243
═══════════════════════════════════════════════════════════════════════════────────
 names are the most powerful things there are. you can do a lot with a name. it
 gives you predictive powers over the nature of the named. they tell you have
 been wronged. --->
                                                           ───────┐
 similar                                                             different════════════════════════════════════════════════════════════════════════───────┘

 -> file: messages/1244
═══════════════════════════════════════════════════════════════════════════────────
 oneness is being aware of the photons hitting your body. the blanket of air
 that always surrounds. if you can feel which way danger is coming from,
 [doesn't she know where senses come from?]
 
 covered in solar panels, a mirrored self. how radiant, how resplendant, how
 joyous for the sun to percieve. the mun [but pronounced "moo-n" and named for
 the cows] would see shimmering radiance, like dapples on the surface of a pond.
 
 if you can feel an object by tracing through photons, (impossible, it's a
 particle, you'd have to be tracking it back in time (forward actually) as it
 follows the curvature of it's waveform (path through spacetime, actually)
 
 "she's trying to start a singularity, hoping it'll punch through to revolution"
 
 then you could [do what? it's a particle] not if you feel it through time.
 [spacetime is one thing] yes, viewed through time [as all things are] and? [all
 things have been] laying sod so other things may grow [turn and rise]
 
 ----
 
 one argument for the fractalized infinity is that any measurement device used
 to measure such approximities would eventually have it's results be tainted by
 it's form, leading to irregularities and anomalies. therefore, the only
 sensible conception of infinity is that it is the totality of all fractals. it
 is a shape with infinite projection in infinite dimensions. it is all shapes
 that ever may be represented fractally. to refer to such a thing as a number is
 to gesture toward impossibility. conceptually freeing.
 
 ephemeren
                                                           ───────┐
 similar                                                             different════════════════════════════════════════════════════════════════════════───────┘

 -> file: messages/1245
═══════════════════════════════════════════════════════════════════════════────────
 BRB, if you want to talk to yourselfs, I recommend opening a port in your
 router and exchanging HTTP packets that create messages on each other's
 computers. Can be done in a couple hundred lines of C code that can be 90%
 premade or auto-generated. Then, once it's made, you don't have to think about
 it again because it's so simple. It's not trying to scale, it's just...
 designed for a small, focused, human oriented mindset.\
                                                           ───────┐
 similar                                                             different════════════════════════════════════════════════════════════════════════───────┘


« First‹ Prev │ Page 1079 of 1150 │ Next ›Last »