Repository navigation
Start here — team foundations, agent-assisted delivery and contributor entry #52
Description
Start here
We are building shared outcomes through foundations first, then rapid agent-assisted construction. Understand the existing contracts, prove one representative path, and review proposed changes to shared boundaries before multiplying code. Small prototypes can run alongside investigation and help test the design.
Main quests and side quests — 12 November 2026
This is the canonical per-person direction map for this campaign. A main quest is the current shared result to advance; a side quest is important ordered follow-up or bounded help, not a second simultaneous allocation.
| Person | Main quest and first concrete contribution | Side quests / support |
|---|---|---|
| Founder (@zenithruneblade) | Main designer/coordinator of the agentic SDLC workflow #40 | Owns Blockworld #163; reviews the shared seams and selected changes |
| Anshuk | Backend/PostgreSQL foundations and durable task/result/human-feedback evidence: #152 under #131; schema, transaction and restart behavior with Nav | VPN/Tether #20 with its current owner; Atelier #162 with Nav; Scout #155 framework/reuse advice |
| Nav | Backend execution fundamentals with Anshuk: API/data boundaries, basic dispatch/orchestration and one simple Graph request/result passage — #153 under #131; continue #149/#150. Discord/Reach is the first practical use case | Shared Linux host work; Atelier #162. Agentic workflows and SDLC #40 are growth directions, with deeper slices selected later |
| Chirayu | Shared web + Companion outcome #132: reusable UI/client foundations, design and construction #154 | Agent/workflow investigations and help with Kiran's selected integrations |
| Kiran | The same web + Companion outcome #132: requirements, acceptance and coordinated agent-assisted construction #159 with Chirayu | Scout #155 first; GitHub Reach #160 next; Gmail later; YouTube lowest priority. Local music extension #161 later |
| Debayan | Shared reproduction, acceptance and contributor entry #133 | Product questions, cross-pair trials and demo/newcomer support |
Chirayu and Kiran share one main outcome, with distinct first contributions and one common client/acceptance plan. They both design and build; this does not create two competing frontends. Side quests do not diminish the importance of acquisition, integrations or media work. Their priority is relative to the main work and available review capacity, not a commitment to build every subsystem now.
Everyone learns the common request/result model through their selected task. The immediate emphasis is backend and client fundamentals, using the smallest existing path: API/service boundaries, persistent state/transactions, basic capability selection and grants, readiness/waiting, cleanup and one simple Graph passage. Understand and reproduce the foundations already present before proposing repairs. Completing a general orchestration platform is not a prerequisite for learning or testing a basic Graph task.
Agentic workflows, Agents, Context and SDLC remain shared directions and useful examples. Deeper dynamic planning, self-construction and broad orchestration are not current implementation allocations. The founder leads #40's design; Nav's missing-capability inquiry is a possible later slice whose exact depth will be selected from the backend evidence. Small prototypes may still clarify a seam in parallel. Agent-assisted construction continues within the understood, reviewed foundation.
Each pair discusses its own request-to-result flow first, then teaches the general model and lets the other pair reproduce one step through #133. Everyone may propose UI/UX and architecture approaches or contribute across boundaries; immediate whole-stack expertise is not expected.
Durable evidence for later improvement
Anshuk leads this within #152, partnered with Nav on the request/execution path and Chirayu/Kiran on the human interaction. The first concrete persistence example is one public/synthetic task, its exact result/candidate, and an attributed human assessment or correction that can be retrieved after restart with the original relationship intact.
For each selected feature, identify which meaningful inputs, outputs, edits and decisions must be retained, who owns those records, and how they survive failure/reload. A console log or in-memory value is not proof of durable custody. Reuse existing domain records and repositories; PostgreSQL stores them but does not define their meaning. Preserve the original and link a correction/new revision rather than silently overwriting the evidence.
Human authorization to perform an effect (#26), feedback on the quality of a result, artifact custody (#23), and explicit memory admission (#30) are different records/decisions. The retained evidence should make later evaluation and dataset lineage possible; training #34 remains in #62 and needs its own admitted corpus, permissions and exclusions. This starts durable collection at the foundation, without starting a trainer or an indiscriminate raw-data archive.
For concrete tasks, use one native assignee as coordinating lead and name partners in the body; pair parents retain both assignees. GitHub assignees have no native primary/secondary rank. Keep Main/Side semantics here because a task can be main for one person and support for another; area labels and Project status continue to describe the work itself. A review request is separate from partnership.
The founder owns architecture/design and reviews/approves every code change. Debayan coordinates product questions, acceptance, reproduction and integration support without code approval/adoption authority. CONTRIBUTING owns branches, checks, review and promotion.
Work from a reliable example
- Trace the smallest relevant existing path and its current owners. Identify state, API, authority, persistence and failure boundaries that later work will reuse.
- Demonstrate one representative result, with real/fixture/prototype evidence clearly distinguished. Review shared-contract changes early, before building many dependants.
- Use coding agents to build quickly within the understood pattern; authors inspect the complete output and the partner independently tries the result.
- Bring a concise demo, changed contracts, checks, limits and one next decision. Propose the next bounded slice with Debayan instead of waiting for the founder to invent every ticket.
Keep at most one review-ready implementation slice per pair awaiting the founder. Parallel discovery, independent reproduction and small prototypes remain welcome. The demo task reuses a selected engineering reproduction in one small preparation slot; Ready does not mean a second independent technical investigation. Research, production, community follow-up and review share ordinary capacity.
Nav's #149/#150 remain active drafts. Tether #151 merged on 11 November; its native source/test progress does not complete #20's live acceptance. Coordinate with current owners and check the selected revision before starting. The current developer main-task entry queue is #152, #153, #154 and #159; the founder-led SDLC design starts in #40, with shared preparation #133 and a small reused demo packet #156. Scout #155 is the first side quest in Backlog until explicitly selected. Pair parents track the shared result rather than adding another implementation allocation.
Anonymization and disclosure are part of the foundations
Anonymization is retained under #62 — Anonymization / Privacy Cuts and remote model calls (historical #38), with scoped inputs/exact-payload egress in #25 and delegated-runtime containment/Provider Gate in #28. The general semantic anonymizer and unrestricted private offload are not delivered; State of Work records the narrower implemented evidence.
Anshuk coordinates backend/privacy seam questions with Nav on Context/workflow integration; Kiran covers connector inputs and Chirayu the visible disclosure/consent states. This is a concern within each selected task, not four extra implementation assignments. Every representative scene identifies what stays local, what could leave, its target/audience, and the refusal or sanitized-fixture behavior. A tunnel, account login or redacted demo does not supply authorization for private model calls. A real anonymization/egress implementation still needs its own bounded selection through #25/#62.
November order
- Enter · 9–15 Nov: establish the area foundations and representative task; agree VPS/PC topology and request/result contracts. A2A A2A / Intercom — one outbound task & Graph resume #19, VPN VPN / Tether — one private peer & explicit route #20 and selected workflow/Scroll Scrolls / Workflow — register and run one revision-pinned Scroll #29 keep distinct responsibilities. Decide sovereign peer versus same-Lich compute and identify missing receiver/private integration before dependent construction.
- Init · 16–22 Nov: another participant reproduces the selected baseline; test useful and refusal/interruption cases and record actual gaps.
- Iter · 23–29 Nov: choose a small repair or implementation from those findings and repeat the same scenario. Broader voice/Composition work follows selected evidence.
Phase is the next planning checkpoint, not a promise to finish every linked outcome that week. Orientation is complete by the founder's decision; it does not qualify old setup/runtime checklists. Preserve the original acceptance of all retained outcomes.
Join and help shape the work
External contributors can start through public issues and CONTRIBUTING, without private-board access or Discord membership. Bring a concrete proposal, reproduction or usability observation to the relevant issue; #133 prepares and independently tries bounded newcomer entry tasks. Claim a specific contribution after checking current discussion and PRs; a claim does not reserve an entire area.
Marketing and demonstrations can begin while foundations develop. #156 prepares one truthful scene, viewer-friendly explanation and specific invitation. Atelier visual work and Broadcast editing/publication retain separate responsibilities; current tools may support the demo. Native application implementation is selected separately through #62, without a video quota or assumed future contributor capacity.
Outcome owners when selected
| Planning checkpoint | Retained outcomes |
|---|---|
| Init / first passage | Runtime #18; local execution #8; Altar #11; A2A #19; Tether #20; Scrolls #29; service jobs #22; artifacts #23; authority #24; Context/egress #25; security #37; governance #73 |
| Iter / next useful slices | Consent #26; voice #48 |
| Preview | Source-bound clean-host release and second-operator tutorial #45 |
| Later, selected by a real scenario | Coffin #28; Archive #30; evaluation #33; Scout fetch #43 and RSS #98; ingress #57; handset #77; semantic utility #99 |
| Independent watch | Dependency advisory #112; dependency PR #129 and any follow-up stay separate |
Broader accepted work and selection triggers: #62. Closed predecessor issues preserve detailed requirements and history; their planning replacement does not mean the capability shipped.
Developer preview
21 December 2026: a reproducible developer preview with tutorials and at least one useful Composition. Voice is high priority. Founder-owned Blockworld #163 and Atelier/Broadcast are candidate demonstrations chosen from working evidence; scope can shrink. Mechanimous YouTube demonstrations and audience-building can begin in November/early December, without a publishing quota.
Roadmap · Delivery evidence · Project board
Original contribution map and dated history: preserved scope and evidence. A dated disposition comment below accounts for all 80 previous issues.
Activity
zenithruneblade commentedon Oct 12, 2026
Foundation campaign wind-down review (2026-10-12): reviewed changes are committed on local main; remote publication is pending. Handoff reconciliation covered the final candidate and archived patches. The concrete missed Android lockfiles were restored in
8bf5d18; superseded builds, caches and reference material are not runtime features.Current acceptance: 3,774 non-container Python tests and 322 web tests passed; the PostgreSQL/container suites were checked in focused batches after correcting migration-head expectations. Android has 87 passing JVM tests, without an SDK/device build claim. Real isolated Quadlet switching and a registered CPU embedding driver have separate scoped receipts.
c6a2f5bfixes the production receipt's premature trace read; Run terminal state and asynchronous trace persistence remain distinct.Remaining priorities from the independent review:
No umbrella issue is closed on the strength of a narrower receipt. Training remains outside this campaign.
Final local candidate:
b556e3a. Final helper/architecture gate: 47 passed, 1 expected opt-in skip; the explicitly enabled real embedding case separately passed with cleanup. Ruff, formatting, strict types and clean Zensical build passed. Fresh remote fetch shows a clean fast-forward relationship to both origin/main and origin/dev; no push performed.Publication prerequisite found by the non-publishing push dry run: the configured local SSH identity lacks repository write access. An authorized Git credential is needed before publication. No remote history was changed.
zenithruneblade commentedon Oct 12, 2026
Renewed foundation review is complete at local main 6e37765, authored as Zenith Runeblade. Nine additional reviewed commits follow b556e3a; the local branch is 23 commits ahead of the fetched origin/main, zero behind. These commits are not pushed. The existing SSH push transport authenticates as pa3kp-c and its dry-run was denied; issue updates use the authorized zenithruneblade connector.
Concrete repairs:
Completed bounded design slices:
Verification: 3,869 non-container Python tests passed, 11 expected optional native-audio skips; full-repository Ruff/format and strict typing passed. The real PostgreSQL application-factory/SAQ/HTTP two-boot test passed again on the final source. Focused real PostgreSQL Archive, derivation-discovery and SQL-fault recovery receipts passed during integration. Generated API contracts, 322 frontend tests and clean documentation build passed at their relevant changes. No timeout, negative assertion or skip was weakened to obtain these results.
The handoff completeness audit covered all 326 final candidate paths and the supplied archive/patch lineage. Its found Android strict-lock omission was repaired by 8bf5d18; no other unaccounted accepted source delta was identified. Preserved private campaign archives and historical proposals are not shipped features.
Still open: semantic/GraphRAG recall and Mirror identity; public Spell/Scroll publication/teaching; Ward/OAuth and remote authority; editable/agent-authored Loom; trace export/native Oculus; connected Android/device qualification; Tether/Veil; active-Vessel lifecycle control and full host/model/browser acceptance. The browser fixture journey remains unqualified because the available browser rejected loopback access. The separate live embedding transport receipt is not combined Archive/database/model qualification. The private workflow lock is not source equivalence or a portable/persisted Resolution Lock.
This is a tested local foundation snapshot, not a defect-free certification or completed roadmap. Related details are maintained on #7, #29, #30 and #71; this issue remains open.
zenithruneblade commentedon Nov 6, 2026
Founder-authorized planning reset, 6 November 2026. I checked all 80 current open issue bodies/assignees, their coordination comments, the live Project and open PR #129 at 10:35 UTC; current main/dev source was fetched at 490dd05. Three independent Astra critiques reviewed scope preservation, architecture/dependencies and team workload. I am consolidating the working queue to three Enter design assignments, retaining explicit outcome owners and a future-scope register. Existing requirements and dated evidence will remain linked; superseded planning issues will not be marked delivered. This campaign changes planning/documentation and the authorized short Roadmap announcement, not runtime deployment or branch protections.
zenithruneblade commentedon Nov 6, 2026
Planning reset — 6 November 2026. The original acceptance below remains binding unless a named replacement outcome records its disposition. This is a preserved historical body, not a fresh capability/qualification claim.
Original issue scope and dated evidence before the planning reset
A contribution map for the whole LychD project: core maintainers, the founder and independent contributors. Milestones express integrated outcomes, not a promise that one small team implements everything or that other work must wait.
Current / Next / Later
This ordering identifies the shortest integrated proof. It does not block independent contract, fixture, documentation, or extension work.
Current sequencing focus
Delivery path
These are completion paths. Interfaces, fixtures, adapters and documentation can be contributed in parallel.
Cross-cutting master maps
From application purpose to execution
New executable application Patterns require an exact Composition owner, or a Suite owner for coordination only. YAML may later be an authoring projection compiled to canonical typed JSON; it is neither application ownership nor executable identity.
Explore by area
Choose a contribution
Working branches: task branches and fork PRs target
dev; a tested snapshot advances tomain. Reviews are requested where useful and maintainers may self-merge. Full issue acceptance plus promotion tomainmeans Done. CONTRIBUTING owns the procedure.Documentation is part of completion
Every feature PR updates the owning docs and stale delivery claims it changes, removes filler/repetition from that touched prose, and checks links plus the relevant documentation gates. Preserve intentional myth, humour, canonical terms and exact security/recovery meaning. #58 covers a finite entrypoint pass; it is not a global gate or substitute for feature-owned docs.
Later direction — pointers, not implementation commitments
Sources of truth
State of Work owns delivered capability; ADRs own architecture; CONTRIBUTING owns setup and checks. This index routes work and does not duplicate their claims.
Published foundation review — 12 October 2026
All 23 reviewed handoff/foundation commits were fast-forwarded to main at 6e37765, followed by one independently reviewed CI fixture repair. The final public head is a913bff. Local and remote main match; no history was rewritten.
The handoff audit accounted for all 326 final candidate paths and repaired the identified Android strict-lock omission. Current implementation includes authorized public-note PostgreSQL intake/lexical recall and attributed Bridge Context, explicit pgvector derivation and generation discovery, exact workflow casting/private implementation bindings, stronger terminal recovery, guarded file custody and lifecycle operations, and corrected operator status/service commands. These are bounded delivered foundations; semantic/GraphRAG retrieval, Mirror identity, public Scroll publication/teaching, remote Ward/OAuth, editable Loom, trace export, connected Android and full host/model/browser acceptance remain unfinished.
Final verification: 3,869 local non-container tests passed with 11 optional native-audio skips. Core Python CI passed on Python 3.14.8 and 3.15.0, including lint/format/types, and documentation deployment succeeded. Relevant integration receipts cover real PostgreSQL Archive and cleanup-failure recovery plus application-factory/SAQ/HTTP two-boot wiring; the latter substitutes model/host collaborators after startup. The initial hosted CI exposed a file-replacement fixture that could reuse the deleted inode. a913bff makes it establish a distinct identity while preserving the original refusal and no-mutation assertions; all 86 custody tests and the full suite passed again. No production behavior, skip or timeout was weakened.
All 69 pre-existing open issues were reviewed against their own acceptance boundaries. Delivery status was updated on 37, including closure of #21's completed local-extension registration/configuration/restart slice; 32 retained unchanged scope because no new implementation satisfied it. Stale starting claims were corrected in #23, #30, #48 and #53. Broader issues remain open. New #77 owns the previously untracked Android physical-handset permission/lifecycle/audio-custody qualification; #48 still owns server-side Echo.
State of Work remains the canonical delivery ledger. This publication is a verified pre-alpha source snapshot, not a package release, deployment qualification or completed roadmap.
[-]Roadmap & contribution opportunities[/-][+]Start here — November tasks and developer preview[/+]zenithruneblade commentedon Nov 6, 2026
Planning reset — 6 November 2026
80 previous open issues → 26 retained outcomes/registers + 54 replaced planning issues. Three new Enter tasks (#131, #132, #133) make 29 open issues. This reset marks no capability completed. Replaced issues close as
not_plannedbecause their planning representation moved; their bodies, comments, accepted requirements and dated evidence remain available.Retained
#8, #11, #18, #19, #20, #22, #23, #24, #25, #26, #28, #29, #30, #33, #37, #43, #45, #48, #52, #57, #62, #73, #77, #98, #99, #112.
Complete disposition
All 80 previous issues and their current route
Coordination and evidence
Stabilize Server provisioning and persistent filesystem ownership
Stabilize Server provisioning and persistent filesystem ownership (#148)
zenithruneblade commentedon Nov 10, 2026
Founder-authorized planning update — 11 November 2026 (Europe/Bratislava).
I am updating the team entry tasks and delivery board around this working method: understand and review each area's core contracts first, prove one representative path, then build rapidly with coding agents on that foundation. Small prototypes can inform the design in parallel. Shared design participation, cross-pair learning, early truthful demonstrations and public contributor entry remain part of the work.
Preflight at 2026-11-10 23:36 UTC: checked the 29 open issues, #52/#131/#132/#133 and affected outcome bodies/comments, open drafts #149/#150 (Nav's Reach work) and #151 (active Tether candidate), and the live Project. GitHub access is working as
zenithruneblade; Projects is accessible through the authenticated browser. Three independent planning panels are checking ownership/workload, technical boundaries, and community/demo entry.Planned scope: clarify #52 and the existing pair/coordination issues; add small named foundation tasks and one bounded demo/community preparation task; connect assignees, labels, parent issues and Project states. Preserve existing acceptance, active PR ownership and the 21 December preview target. This changes planning only, without runtime delivery claims, merge/permission changes or edits to the concurrently reworked repository documentation. Private CV and personnel material stays out of public GitHub.
[-]Start here — November tasks and developer preview[/-][+]Start here — team foundations, agent-assisted delivery and contributor entry[/+]zenithruneblade commentedon Nov 10, 2026
Planning update completed — 11 November 2026.
tracking; the Ready view excludes those parents from the actionable queue.The scope was GitHub planning only. No repository source/documentation edits, runtime qualification, merge, permission changes or external media publication were performed. The ongoing documentation work and original capability acceptance remain with their owners.
zenithruneblade commentedon Nov 12, 2026
Founder-authorized planning refinement — 12 November 2026 (Europe/Bratislava).
Preflight at 01:36 UTC: inspected the current open issue/PR list, #52/#131/#132/#133/#152–#156, privacy owners #25/#28, future register #62 and relevant coordination comments. Nav's #149/#150 remain open; #151 was merged on 11 November, so the earlier draft wording in our entry plan is stale. Full #20 acceptance remains separate.
I am refining the per-person Main quest / Side quests map: shared web/Companion outcome for Chirayu and Kiran with distinct first deliverables; backend and Discord/agentic main paths for Anshuk and Nav; explicit Scout, Reach-provider, VPN/host, Atelier and later local music-extension directions. Anonymization/egress must be visible across those tasks. Three read-only panels are checking ownership, provider boundaries, and privacy/media scope.
Scope: GitHub planning bodies, bounded tasks, ownership and Project routing. Existing acceptance and active contributors' work stay intact. Repository source and the concurrently reworked documentation are outside this change. This is an operator planning campaign, not evidence of native LychD agent execution or new runtime qualification.
zenithruneblade commentedon Nov 12, 2026
Planning refinement completed — 12 November 2026.
Only GitHub planning, Project fields and issue relationships changed. Repository source and concurrently reworked documentation were not edited; no runtime qualification or media publication is claimed.
zenithruneblade commentedon Nov 12, 2026
Founder-directed ownership clarification — 12 November 2026.
Preflight at 01:51 UTC: refreshed the current issues/assignees, #52/#62/#131/#133/#152/#153/#29 and open PRs #149/#150; reviewed recent owner comments. The founder is taking Blockworld and the main design of the agentic SDLC workflow. Nav's main direction is agentic systems from the start, with Discord/Reach as the first concrete use case. Anshuk joins through backend/runtime and the exact Scroll/execution passage across the selected private topology; the pair also shares Linux learning and reproduction.
I am making these responsibilities explicit, preserving current task acceptance and PR ownership, and adding bounded founder-owned entries where no existing issue represents them. Early Nav SDLC work is understanding, design and a small missing-capability scenario, without promising the full implementation or automatic adoption of generated code. Both pairs should learn the workflow and later teach/reproduce each other's examples. Two read-only panels are checking the existing owners and avoiding duplicate SDLC/Blockworld scope.
This remains GitHub planning only; no repository documentation/source edits or runtime delivery claims.
zenithruneblade commentedon Nov 12, 2026
Ownership clarification completed — 12 November 2026.
No repository source/documentation edits, merges, runtime changes or permission changes were made.
zenithruneblade commentedon Nov 12, 2026
Founder-directed emphasis correction — 12 November 2026.
Preflight at 02:01 UTC: refreshed #52/#62/#131–#133/#152–#154/#159/#40, data owners #23/#26/#30 and retained training #34, plus current open PRs #149/#150 and #164. Existing contributor scopes and current work remain intact.
The immediate learning/implementation emphasis is backend fundamentals: request/service boundaries, PostgreSQL state and transactions, basic capability dispatch/readiness/orchestration and one simple Graph passage. Agentic/SDLC stays the direction and a small teaching example, without requiring deep self-construction before those basics are understood. Details of deeper work remain open for later selection.
Anshuk's existing #152 will carry a concrete durable task/result + human feedback/correction baseline. The pair should identify current owning records, schema/transaction/restart behavior and useful lineage for later evaluation/training, without a duplicate universal event store or automatic training harvest. Nav partners on the request/execution path, and the frontend pair supplies the human interaction/acceptance view. Two read-only panels are checking scope and the existing persistence/HitL/training boundaries.
This is a refinement of existing tasks, not a new parallel assignment or an implementation claim. Repository source and concurrently changing documentation remain outside this change.
zenithruneblade commentedon Nov 12, 2026
Fundamentals and durable-evidence refinement completed — 12 November 2026.
Repository source and concurrently reworked documentation were not edited.