ideas.org

draft ideas

org mode ftw

we can use org mode as the overarching plan that includes all other files.

Each file can be a perspective to look at the system. Each line a different nuance. We can see them as filters and groupings and relationships and paths followed in a proof. See section below for proof information.

introspector perspective

list of sliders for dimensions, each has an emoji and default, each is coefficient in polynomial or tensor

rendering with streamlit

points are then clickable

sliders for object ids as points of memory, show pointers between memory,

using graphics to display memory and pointers

projection of awareness into a 9d hyperplane.

the view vector as a series of emojis that make a path each one adds a nuance to the view. abiility to traverse forloops/groups/maps/folds!

each input is mapped into a hyperplane of 9d, a point there.

each point is a 8d hypersphere.

we can project them as points on a line or 2d or 3d.

we can fix the other dimensions with sliders.

we layout the points with math and guidance.

tulip and graphviz files

one set of inputs that is the state of the system at all times. the hypersphere. the 9d point.

the git repo.

viewing as filtering and grouping, your perspective, your vector. we have a view vector : input , that is the dataset you are looking at. default is the set of all sets.

your branch

deciding as selecting, axiom of choice : input, default none.

your commits

read these line by line into the chat model and lets be able to choose between then and send them to chat.

your comments on the commits

be able to send to many models and rank results.

running agents

be able to edit the input or rate it

git rebase

interface to open-assistent.io , you.com, huggingface.co backend or front ends like bard, bing, chatgpt, etc.

todo

interface to local llm via the git/yaml interface

be able to intercept the api calls and mark them up in yaml.

wireshark to yaml/emojis and back

capture strace and all types of diagnostics

think datadog agent

be able to deploy via terraform

todo

create cicd like atlantis but with chatops

todo

make it easy to review lists of data from the app.

org mode emacs

convert this list into a inputs/dataset on the filesystem

org mode

convert all the lists!

introspector python and more, proc file system

create central way to dump everything as input and then filter it to create new version!

oroborous

treat all data as inputs in a dataset, even asts, tokens, vectors, models, executables.

unix everything is a file.

create different views of all data

cat

structured way to import everything

cat file | grep | import “dataset”

state of unfolding captured for each object being imported into the model with many views and updates in a huge workflow.

capture state of workflow, state of boostrap, state of machine.

create yaml representation of each datagram, using emojis.

capture execution logs.

be able to download executable

Proof

the proof is following a simpler path in a numerical model that simulates the complex path of the program. dual representation, digital twins, bi-simulation.

Coq/metacoq

Template haskell

consider if we can treat two executions or logs as bisimulation for coinduction between proofs or executions as harmonies.

looking for alignment between models.

paxos protocol for model election.

quality

capture and simulate endpoints locally via yaml traces.

be able to capture and share : tcpdump, strace, perf, protobuf, ltrace, uprobes, debug logs.