Concepts and decisions
Its concepts, and the decisions taken - with what was rejected.
Attention: where the whole is weakest
One view across every project, computed from their data: steps nothing delivers, work in progress that has not moved, regressions, and what waits on a person (a merge, a SIM card). Read at the end of every stretch to choose the next.
The board
A project's live work - ongoing, waiting, planned - kept as data and shown in the picture next to its path.
The conveyor
A forkable lab that holds everything a person or an agent needs to bring a goal about - for phones: tools, procedures, safety rules, principles; projects made from it carry the mark.
Discovery, hexagonal
Finding projects is a port; each platform is an outlet - GitHub first, others (GitLab, Open Hub) later without touching the core.
Essence
Both a method and the observatory's signature view. The method (its own project, creation-guidelines/essence-oco): how to say why and how of anything, as four answers each leading to the next - why it exists, what it must reach, how each step is delivered and proven, what is happening now - with the test of a good answer for each; it describes itself with nothing but itself. The view: a project's landing page, the chain down one line, with four marks for how completely it tells its essence.
A grand project's family
The grand project's repository and every project whose parent is its owner or itself.
Grand project
A marked project of kind picture: a large aim with its own tab, views and family of projects (Linux Phones is the first).
The harvest
On a schedule, the observatory reads every marked project's shared tables, tags each row with its project and merges them; problems are shown, not lost.
How we work, as a working set
Not a manifest but a usable set under one roof (the creation-guidelines organization): how to think - the canon's principles and architectures, the ideal project - and how to do - the practices, each naming the tool that carries it (land, shot, ocoeco check, lab-agent-oco) - shown as a lens like any other, and reachable by anyone through one command.
Languages
Two kinds of text, each translated where it belongs: the interface in the face's Fluent catalogs (grammar and plural forms right), the data in each project's translations table (marked human or machine), falling back to the source. The site is the same static files per language under /<language>/, switched by a thin line of languages by their own names.
Lens
A repository of its own (kind lens) that declares a family of projects and the views that show it - from a small shared vocabulary (path, matrix, families, runs, backlog, log, ideas, heartbeat, cards). The observatory discovers lenses and renders each with the same engine; curated lenses take its navigation, every other is listed among the discovered.
The mark
oco.toml at a repository's root plus the oco-project topic (and oco-kind-<kind>): which project it is, its kind, its parent, the tables it shares.
Milestone
A stop on a project's path, on one horizon, with criteria tied to measured data (feature statuses, tests, health); it fills by itself as evidence arrives, never by hand.
The observatory
A meta structure, tied to no project: one living picture of every project of the open commons, built from the projects' own data.
The observer
Half-manual tests ask an observer - a person today, a rig of cameras and servos later - with nothing else changing.
ocoeco
The observatory's information ecosystem and its backbone - both a principle (data is produced by projects as they are made and as they work, never by hand) and a tool (formats and protocols, gatherers, chaining, propagation), in the repository ocoeco.
The path
A project's visit card: a line through its three horizons, its milestones as stops, filled to where the evidence shows - and unknown shown as unknown.
The readiness ladder
Under the horizons, a project's maturity on the established Technology Readiness Levels (concept to proven in operation), so projects of any kind compare on one scale.
Shared parts as rhizcuts
tad, the mark and the conveyor's tools travel as rhizoid modules: owned, patchable git forks rather than template copies or subtrees.
tad, the heart
tad - text as data - is the living heart and blood of oco-eco: the formats every project's data is kept in, the gatherers that fill them, the views that show them; supplied in every repository as a rhizcut.
Text as data (tad)
Everything a project knows - plans, decisions, concepts, progress, measurements - is kept as tad data in its own repository, and rendered from there.
Three horizons
Every project is measured against three: working (it does its job end to end), perfection (complete and polished - as good as it should be), ongoing perfection (it stays so: healthy, free of regressions, upstreamed, maintained).
Unknown is not failure
What has not been measured is shown as not known yet - never counted as missing or broken.
Every project tells four things about itself, in its own data: its idea (why it exists, written by the project for anyone, as an ordered piece of writing), its path, its workings (how each step is delivered and proven, by which tools, and why we care), and its work. A lens gathers and shows these and writes none of them. The observatory is one more project that tells the same four.
From the author: how we work must not float beside the projects like a random toolbox - it belongs to each project, tied to its goals; and the idea must tell a commoner what this is, in the project's own voice. Shared tools are projects that others' workings name.
Rejected: How we work as one roof - a separate working set beside the projects (decided earlier the same day); the idea as terse cards written about a project.
2026-11-10
The site is a designed static website from tad data - minimal, elegant, no JavaScript - not a tad text page.
It is the observatory's visit card.
Rejected: Rendering tad's own pages for the observatory.
2026-11-10
Discovery is hexagonal; on GitHub the mark is the oco-project topic plus a kind topic, confirmed by oco.toml.
Topics are what GitHub's search indexes best: one query finds every marked repository.
Rejected: Code search for oco.toml (slow, rate-limited) as the primary way.
2026-11-10
The observatory's engine (mark, discovery, harvest, site) is its own repository, in Rust; tad is to be rewritten in Rust with its formats unchanged.
Sturdier machinery, compiled checks, one toolchain across the commons' tools.
Rejected: Extending the Python tad renderer.
2026-11-10
Essence is its own project - the method - beside ocoeco (the language it is written in) and ocoeco-engine (the eye that draws it). It must describe itself with nothing but itself, and be usable: a short guide with a test for each answer, and a measure of how completely any project tells its essence.
Proposed by the author: it gives why and how about how to do why and how. Weighed against the risk of one more box: the method is knowledge neither the format nor the renderer holds, and it is more general than software.
Rejected: Essence as only a section of ocoeco's specification and a page of the engine; Essence as an essay with no check.
2026-11-10
Everything thought, decided and planned is written into the tad structure - concepts, decisions, backlog, session log, milestones - in the project it belongs to.
Nothing may depend on anyone remembering.
Rejected: Prose documents alone; plans kept in conversation or memory.
2026-11-10
How we work is gathered under one roof, the creation-guidelines organization: the canon, the playbook, lab-agent (renamed lab-agent-oco), land and shot (moved from a personal repository), and the practices (moved from the observatory's own design data). Every practice names the tool or check that carries it; the share of practices so carried is its 'how close are we'. It is shown as a curated lens - think, do, tools - replacing the hard-coded page, and lab-agent-oco guide prints the whole working set. SUPERSEDED the same day by "a project, whole": how a project works belongs to each project, not to a roof beside them.
Asked for by the author: a working and usable solution set, part of the grand picture, not a manifest. Two practices were broken on the day they were written, both resting on memory alone.
Rejected: Leaving practices in the observatory's design data; a page of principles with no tools behind them; a new organization (the roof already exists).
2026-11-10
The face uses Project Fluent message catalogs for its interface; projects keep data translations in a standard translations table (with by: human or machine), falling back to the source language; the site is rendered once per language under /<language>/ with identical paths, the root redirecting old addresses; the switcher is a thin line at the header's end naming each language in itself (English · Русский).
Fluent gets each language's grammar right (Russian has three plural forms); separate translation rows keep data made by work untouched; prefixes keep paths identical and old links alive.
Rejected: Plain string tables (wrong plurals); translated columns inside source tables; translating rendered HTML afterwards (fragile, no grammar); flags or bare codes in the switcher; switching by script or by the browser's language (no JavaScript, and a reader's choice should be theirs).
2026-11-10
Grand views become lenses: repositories of their own that declare a family and its views as data; the observatory discovers and renders them all with one engine, curates some into its navigation and lists the rest as discovered. Our views become curated lenses built like any other; linux-phone-oco/docs keeps producing research data and the Linux Phones lens moves to its own repository in its family's organization.
An observatory of a commons cannot be the only one allowed to look: views as data make the engine a general renderer, open to anyone's family of projects, with no privileged path.
Rejected: Views hard-coded in the engine (a bottleneck and a gatekeeper); our views on a privileged path; a free-form template language for views (the vocabulary grows from real needs instead).
2026-11-10
The observatory is a meta structure, tied to no project; its tabs are itself and the grand projects, each with its own views.
A devices tab made the whole look device-centred; devices are one part of one grand project.
Rejected: One flat set of tabs for everything.
2026-11-10
The ecosystem of gathering, chaining and propagation is called ocoeco (observation commons observatory ecosystem).
It has become an information ecosystem, and deserves a name.
Rejected: Leaving it unnamed; the dashed form oco-eco.
2026-11-10
New projects and organizations of the ecosystem may carry -oco in their names - a telling touch, not a requirement. Nothing existing is renamed: belonging is declared by the mark and shown by the topic.
Agreed with the author after weighing it: a name that says where a thing belongs helps a person; but identity by name is the mistake Rhizome names, and renaming has real costs.
Rejected: -oco added to every existing organization, repository, tool and command.
2026-11-10
ocoeco becomes its own repository and the working system from now on: the principle, the formats (tad is the format it speaks), gatherers, chaining, and the mark, discovery and harvest moved in from the observatory's engine, which keeps the face. tad-engine stays as it is, frozen, its formats unchanged.
The ecosystem's machinery belongs together and apart from the face; tad's name lives on as the format, keeping every existing file valid.
Rejected: Renaming tad-engine in place (its history and users are a different thing); keeping the machinery inside the observatory's engine.
2026-11-10
The working practices are tad data in the observatory, and every agent reads them live through lab-agent guide.
An improvement made once reaches every agent; no agent depends on its own memory to work well.
Rejected: Practices kept only in one agent's memory or in prose files that drift.
2026-11-10
Table columns are ranked best first from the data; unknown is shown as unknown and never counted as failure.
The order corrects itself as support grows; a gap in knowledge is not a verdict.
Rejected: Hand-ordered columns; one 'how close' percentage that counted unknown as zero.
2026-11-10
tad is not a template but a rhizoid module (a rhizcut) supplied in every repository, with standard gatherers (CI jobs and hooks reading real files, tests, releases); tad has its own path, shown among the other projects rather than as a top tab.
Templates and copied subtrees drift and cannot send fixes back; tad is central but smaller than a grand aim like Linux Phones.
Rejected: tad as a template or a subtree copy; tad as a top tab.
2026-11-10
Projects are shown against three horizons - working, perfection, ongoing perfection - on the readiness ladder, with milestones whose criteria are measured data.
Overall progress alone loses the reader; three honest horizons show where a project really stands.
Rejected: A single progress bar; milestones ticked by hand.
2026-11-10
Hand-work is recorded on the way every change already goes: land asks for --hand ACTION or --no-hand, and a failure to record stops the landing.
A ledger is only worth reading if it is never silently skipped; land is the one path every change takes.
Rejected: A reminder nobody must answer; recording after the fact.
2026-11-11
The harvest rebases its data onto whatever landed while it worked, and retries the push up to five times.
Since land starts harvests, a change landing during a harvest's three minutes rejected its push (2026-11-11 14:45); the harvest writes only its own data/ files, so a rebase cannot collide with what lands.
Rejected: Pushing once and failing; force-pushing (would drop what landed); holding landings while a harvest runs.
2026-11-11
The lab machine holds an idle inhibitor while the agent serves jobs; it is updated through its package repository regularly.
It suspended on idle at 03:09 and the agent stayed stuck after waking until a reboot; and it stayed on lab-agent 0.1.39 for days because its package list was never refreshed.
Rejected: Blocking sleep altogether - its owner may still put it to sleep on purpose.
2026-11-11
Everything breathes through ocoeco - its language, gatherers and checks - and the observatory stays a mirror: projects produce their data, the observatory reads it, and no project needs the observatory to work.
The author's direction that the whole lives through OCO, weighed against the risk of the mirror becoming a control centre and a single point of failure.
Rejected: Projects depending on the observatory's site or harvest to function.
2026-11-11
No new Python, even for one-off glue: shell orchestrates, Rust carries the logic (rust-script for single files), repeated work becomes a Rust tool; tad stays as it is until ocoeco replaces it.
Asked by the author: Python's ecosystem is fragile - versions, imports, strings. Measured on the lab machine: a rust-script one-off compiles once in seconds and then runs at once. The deeper cause of failed runs was text patching, answered by structured tools such as ocoeco note.
Rejected: Python for throwaway glue; rewriting tad now.
2026-11-11
land starts the observatory's harvest after merging a marked project; the harvest coalesces bursts - one run in progress, at most one waiting.
Triggering the harvest and waiting for it was manual work done 19 times.
Rejected: Each project's CI dispatching the harvest (a token with rights on the observatory in every project); harvesting every few minutes on a timer (wasteful when nothing changed).
2026-11-11
Mechanical rules are enforced by the tools every change passes through, not kept as practices: land moves tags only onto what merged, checks changed workflows with actionlint and commit titles against the conventional form before pushing.
The author asked whether such things should be guarded by a tool rather than a mental note; twice in a day a practice did not hold.
Rejected: More practices; more care.
2026-11-11
Whatever ships or is called by others is Rust; whatever repeats becomes a Rust tool; one-off glue in a job may stay small.
Python let bugs reach runtime that a compiled language rules out; the record scripts that repeated became ocoeco note.
Rejected: Python in shipped or shared code; rewriting every one-off line of glue.
2026-11-11
tad stays the general substrate - a transparent database of text files in git - and is rewritten in Rust; ocoeco is one specific use built on it, and others (Truelink) will be. The wrap-up pass comes first; then tad in Rust is the focus.
The author: tad is simple and general, and proved that a text/file database in git can be accessible, transparent and intelligible; ocoeco is a narrower use of it, and a coming idea (Truelink: facts, their sources and what is drawn from them, so fact-checking can be automated) needs tad too. Python leaves; tad does not.
Rejected: Moving tad's duties into ocoeco and retiring tad (2026-11-11-wrap-up-then-tad, superseded: it would make a general tool specific); hardening tad's Python.
2026-11-11
First a wrap-up pass settling every in-progress item; then the focus is moving tad's duties into ocoeco (declared table shapes, canonical form, documents rendered from data) and retiring tad's Python; phone work continues where it is not blocked.
The author: big ideas wait on a backbone that is not Python, and loose ends make the whole hard to read. Making tad's Python pristine would polish what is being retired.
Rejected: Hardening tad in Python; dropping current work mid-way; switching before the board is honest. Superseded the same day by 2026-11-11-tad-in-rust: tad stays general, in Rust.
2026-11-11
Everything in progress is parked, each item with its resume point, and lab-agent is redesigned first. Resume order afterwards: 1 the glimpse picker (ocoeco ocoeco-commit), 2 the CI check for glimpses (ocoeco land-needs-a-record), 3 ocoeco release (ocoeco ocoeco-release), 4 argenv for every tool (ocoeco and tad argenv-surface), 5 lab-agent through ocoeco (lab-agent milestone through-ocoeco), 6 ocoeco check over every shared table (ocoeco check-every-shared-table), 7 the modem and telephony checks (surya-lab), 8 the microphone (docs, surya-lab), 9 the remaining parked items in each project.
The author, 2026-11-12: the 2026-11-11 incident (a stall, then 102 finished jobs replayed, fifty at a time) blocks everything - the lab is how all work is done. Its cause is a foundation laid as a one-off prototype; it needs rethinking against what the world has learned, not patching. No harm was found: four replayed pull requests, all the legitimate tad 0.4.0 pin bump; no direct pushes, tags, releases or phone actions.
Rejected: Patching lab-agent and carrying on; carrying on and investigating later.
2026-11-11
The face asks, for each star, why it exists and how far it has come: one part, laid out once for each star - its leading idea, then each milestone that rests on items of work, in the order of its rungs, with how much of what it rests on is done. The lens of purpose is no longer placed first; it follows with every other lens, folded.
The face says it shows first why each star exists and how far it has come, and placed the lens of purpose for that: fifteen ideas in one list and, far below in another, the milestones of all the stars mixed. What belongs to one star stood in two places. Laid out once for each star, a surface can stand the stars side by side - the page does, as small cards with rings - which is what the engine's front page draws by its own knowledge of stars. Tried on the lab in a working tree that was thrown away, and looked at, before it was entered here.
Rejected: Changing the lens of purpose itself (it is every place's lens, and in a terminal fifteen stars of ten lines each would bury its figures); placing purpose as well as asking (the same said twice on one page); taking the part back and entering another (records of work point at it: it was changed where it stands, saying that it places nothing now).
2026-11-15
The harvest observes by holding: each place found is brought to hand, as it is on GitHub now, and read as any place held is read - its tables, those it keeps on another branch, and what the routes read. What this observatory observed then holds what the author's does: a link to a document, a commit, a guard or a file of source points at something here too.
Looked at on its published page on 2026-11-15, once the page was drawn anew: the lens of signs said seven links point nowhere, where the author's observatory, over the same places, says two. Five were links to what a route reads - a document, a file of source - which a harvest that only asks GitHub for each place's tables does not bring. Looked at before changing: this way of observing dates from 2026-11-12, before routes were said; the author's observatory has observed by holding since then, and this workflow was not brought along.
Rejected: Leaving the five showing (they say something untrue of the stars: the links hold, this observatory did not look); passing over links to what a route reads wherever what was observed lacks it (a link nobody checks is how one comes to point nowhere).
2026-11-15
Which pages this observatory publishes in the language is said by a lens of its own, commons-site: the whole, laid out through the face, and a page for each place in view, laid out through what is around it. The harvest lays them out with ocoeco site, beside the engine's views; a star's card on the whole leads to the star's page.
The author's observatory has said its pages by a lens since 2026-11-15; this one wrote one page by hand in its workflow, the whole, and its stars' cards led nowhere. A lens of pages is what the language already has for it, and a place says its own.
Rejected: A page for each sign and each decision, as the author's observatory has (a visitor here meets the stars first; what is around a star leads on to its signs); the whole at index.html (the engine's front page stays the first a visitor meets until the language says a second language); the name site (a lens is named once in the whole: the author's observatory keeps a lens called site, and two lenses of one name stop every look that sees both).
2026-11-15
The lens that says which pages this observatory publishes is called site - the name it was first meant to have - as the author's observatory calls its own: a lens is of its place, and each is site where one stands.
It was called commons-site only because a lens's name had to be one across every place in view, and the author's observatory keeps a lens called site.
Rejected: Keeping commons-site (a name made to step around what no longer stands in the way says so for ever).
2026-11-16
The practices leave the observatory's design data for grenudi/ideal-agent, where they are a table of its tad database; the observatory gathers them from there (the author's observatory and the ecosystem constellation gather grenudi/ideal-agent) and shows them as ideal-agent's rows. design/data/practices.json is removed and its export with it; the links that touched a practice here point to ideal-agent's rows of the same ids (ocoeco #456, ecosystem #160).
The author, 2026-11-17: one home for general agent know-how, grenudi/ideal-agent (its decision 2026-11-17-one-home-for-agent-know-how). Before removing, both copies were harvested side by side and compared: all 43 practices the same in id, text and reason.
Rejected: Keeping both copies (two lists drift); keeping the practices here and pointing ideal-agent to them (the one home would not hold what every session reads first).
2026-11-17