DEV Community: Rob Bulkley

Aether/XLang Progress Update

Rob Bulkley — Fri, 14 Aug 2026 16:36:30 +0000

Aether / XLang — Comprehensive Project Progress Report

Date: 2026-08-14*Baseline audit HEAD:* 12a2181 on codex/xlang-local-first-studioPackage contract: 0.37.0 (language 0.11 + M19e AETH v12 + M25 local package publication)Baseline audit: AUDIT_REPORT-2026-08-11-FULL-PROJECT.md — GREENCurrent delivery verification: ADR-126 canonical root-nursery exact-empty-Text spawn unknown-target seed-SPEAK pilot — targeted red/green, direct, valid-source, predecessor/lexical/signature/caller-state/delimiter/priority boundaries, product/self-host, tracker, and full release gate PASS*Constitution:* AGENTS Constitution 5.0.1

1. One-sentence status

Aether is a local-first, seed-hosted product compiler with verified AETHexecution, dual-compare self-host proofs, and human-authorized F-NATIVE /F-REGISTRY pilots — now 0.37.0 package contract, with BARP having movedproduct authority off the Rust bootstrap for default toolchain paths andM32a/M32b now providing a closed local verified-execution evidence and strictcomparison surface.

2. What the product is today

2.1 Language and runtime

Layer	Status
Core language surface	Canonical 0.11 (M2 resources, M4 Error[Whole], M5/M15/M23 comptime, M6 layout, M7 nurseries, M8 pure host)
Toolchain packages	M9–M18 projects/modules/LSP/tests/workspaces/stdlib; M19a–e resource/task; M21 FFI pilot; M22 cross-package; M23 seed-native; M25 local source-package lifecycle
AETH	v11 default; v12 for M19e `task weave` + `checkpoint`
Default compile	Seed forge (`seed/aether_seed.aeth`) — not Rust bootstrap
Execution	Verify-before-run VM; grant-empty pure fixtures; optional `--grant-*` / `--grant-lib`

2.2 Product toolchain (BARP)

Bootstrap Authority Reduction Program (DESIGN-BARP-001, ADR-043–116):

Capability	Authority
`compile` / seed rebuild	Product seed (ADR-067)
`check` / `format` / `structure` / project format	Product default; `--bootstrap` recovery
LSP diagnostics / symbols / hover / definition / format	Product-primary
Structural edits	Product weave/body/record paths
Multi-module / multi-source	Host elaborate + seed emit + unit digests
Diagnostics	`AE-SEED-*` preflights + SPEAK packets; bounded seed pilot 003/004/005/006/007/010/011/012/013/014/015
Bootstrap residual	Dual-compare oracle, recovery flags, full `aether.ast/v8`

2.3 Law forks (human-authorized)

Fork	Through	Highlights
F-NATIVE	M35j (ADR-059–099)	AETH→C / object / LLVM IR / LLVM object / exe; `--target` closed matrix; host dual-run, cross link-only; probe + hermetic env (`AE-NATIVE-007`)
F-REGISTRY	M24i (ADR-060–100)	Offline pin/verify; HMAC/Ed25519; explicit fetch; rotation; multi-root policy; CLI root/certified-key setup; date-checked X.509-lite CA store included in cache verification (`AE-REG-012/013`)

2.4 Task model

Item	Status
M19e active-frame cancel	Proven (v12, checkpoints)
Task frame surface / checkpoint density / inventory	Proven tooling (ADR-085/093/097)
Reserved future surface	Fail-closed AE-SEED-014 (ADR-089)
Task weave requires checkpoint	Fail-closed AE-SEED-015 (ADR-101/106; exact direct seed pilot)
Handles / timeouts / parallel runtime	Not implemented (ADR-081 design only)

3. Architecture snapshot

Aether source (.ae)
    │
    ├─ product path (default)
    │     host preflights (AE-SEED-*)
    │     multi-source? → host elaborate
    │     forge seed/aether_seed.aeth  → AETH bytes
    │     seed SPEAK merge on failure (pilot codes)
    │     verify_bytecode
    │
    ├─ recovery path (--bootstrap)
    │     Rust parser/validate/emit (oracle + AST)
    │
    └─ optional lowers (F-NATIVE, verified AETH only)
          → C / object / LLVM / native exe (host tools)

Verified AETH
    ├─ aether run  (VM; pure or grants)
    ├─ aether bench (embedded pure corpus; verify + decode + execute; no grants; M32b data-only compare)
    ├─ aether forge (compiler ABI only)
    └─ dual-compare tests (product ≡ bootstrap where claimed)

4. Evidence ladder (how we know it works)

Evidence class	Mechanism
Unit / integration tests	`cargo test --workspace`
Seed self-host	`seed_self_host` dual-compare + multi-generation
Example corpus	Gate dual-compare of shipped examples
Quality gate	`tools/aether-gate.ps1` (quick / full / release)
Claim control	CORE_CLAIMS.md status rules
ADR trail	100+ ADRs under `docs/ADR-*.md`
Matrices	M* / M19E / M23 / M35A / M24A validation matrices
Constitution	Project Level-4 pointer → pack AGENTS.md Section 0

Latest full-gate stamp: 2026-08-14 ADR-126 delivery — GATE PASS mode=release (pack v5.0.1; 359 Markdown files / 1,447 local links; full-quality suite, four-way seed identity, 444-file package plus SHA-256SUMS, and independent consumer verification)Latest release-verified seed identity: D0D17756F587709BC85E323E0545281E304362288B687BAA0D48D8C334E18AFB

5. Recent maturity program (post–0.36 release)

The 0.36 package contract (M19e) remains the executable language baseline. M25now advances the toolchain package to 0.37 with a local package ecosystem step;BARP and M32 work remain independence and infrastructure maturity:

5.1 BARP highlights (ADR-043 → 126; ADR-126 release verified)

Product-default CLI toolchain; bootstrap recovery only
Multi-source envelope + multi-file host forge + unit digests
SPEAK protocol AETHER_SEED_ERROR: + host packet ABI
Seed SPEAK pilot: AE-SEED-003/004/005/006/007/010/011/012/013/014/015; 003/007 are bounded lexical checks, 014 scans canonical reserved-task prefixes, 015 requires an exact checkpoint in a canonical task body, 010 detects only a canonical ordinary-Whole Text-literal or exact Truth-literal yield, 011 detects canonical total-Whole direct-bind/root-yield/root-revise/root-speak target existence, delimiter-bounded root-handle target existence, literal-erroring-header root-forward target existence, and literal total-root-nursery immediate zero-argument, single-digit Whole, positive two-digit Whole, or exact-bright Truth spawn target existence across declared top-level weave headers (the handle form requires fixed into / otherwise error into delimiters that leave a destination suffix; the forward form requires -> Whole raises Whole:; the nursery forms require root together: plus immediate spawn call target into destination, spawn call target digit into destination with one ASCII decimal digit, spawn call target digits into destination with exactly two ASCII decimal characters and a nonzero first digit, or spawn call target bright into destination), and 013 detects only a canonical ordinary-Whole choose same / choose less / exact choose bright: / exact choose dim: / exact choose not bright: / exact choose not dim: nested yield (dual-compare rebuild)
ADR-125 additionally recognizes only the immediate exact-dim Truth child spawn call target dim into destination; full release, seed identity, and packaged consumer verification pass. Exact empty Text is separately ADR-126.
ADR-126 additionally recognizes only the immediate exact-empty-Text child spawn call target "" into destination; its valid target is an ordinary one-Text-parameter weave, not a task-frame expansion. Targeted red/green, direct, valid, predecessor, boundary, priority, product/self-host, tracker, full release, seed identity, and packaged consumer verification pass.
Forge SPEAK capture on failure and verify merge
Yield-in-truth-choose fail-closed (AE-SEED-013)
Task checkpoint required (AE-SEED-015)

5.2 M32a/M32b verified-execution evidence (ADR-104/105)

aether bench selects only the embedded welcome, arena-buffer, andtask-loop workload sources. It product-seed-compiles each selected workloadonce, explicitly verifies its AETH, and records bounded raw local samples ofverify + decode + execute with empty grants. Unprofiled reports remain v1.M32b's explicit non-secret --profile produces v2 with a safe environmentfingerprint and checked stdout SHA-256; aether bench compare reads only twoexplicit 256 KiB-capped strict v2 data files and requires equal profile,environment, selection, source, and behavior before reporting medians. It neverexecutes input reports, sources, or artifacts. There is no caller-suppliedprogram, native/JIT path, network/process/model authority, performancethreshold, hardware/toolchain attestation, or broad speed claim.

5.3 F-NATIVE (M35a → M35j)

C pilot → locals → SPEAK/multi-weave → host-cc dual-exec → object → LLVM IR →LLVM object → native exe → toolchain probe/hermetic → operator --targetmatrix (host dual-run; cross link-only)

5.4 F-REGISTRY (M24a → M24i)

Offline pin → HMAC signed → Ed25519/HTTPS → rotation → multi-root policy →root certs → multi-level chains → explicit CLI root/certified-key setup →X.509-lite → CA store + chain verify + cache-gate validation

5.5 M25 local source-package publication (ADR-107)

aether pkg pack|verify|publish|install|verify-cache packages exactly onecomplete locked project into a transparent aether.package/v1 directory bundle.It binds raw manifest/unit bytes, rejects hostile paths, symlinks, nonregularand unlisted bundle files, bounds input size, stages output, preserves existingtargets, and gives an explicit local cache a deterministic collision-safeidentity. Direct/cache installation creates a normal locked project; coreevidence builds and runs two independent M22 workspace consumers of oneinstalled package. M25 adds no network/resolver/signing/guest authority.

5.6 Recent tip commits (illustrative)

Commit	Slice
`ec5e33c`	ADR-102 lexical seed-SPEAK tab / legacy-`fn` pilot
`f294701`	M24i CA-store / M35j target-flow operator hardening
`12a2181`	ADR-098–101 multi-code SPEAK, targets, CA store, checkpoint
`1e2fa31`	ADR-094–097 SPEAK empty pilot, native probe, X.509-lite, task inventory
`b5305cf`	ADR-090–093 forge SPEAK capture, native exe, multi-level certs, checkpoints
`f03e310`	ADR-086–089 SPEAK matrix, LLVM object, root certs, task reserve

Previous release-verified implementation: ADR-124 extends the direct AE-SEED-011seed-SPEAK witness to an ordinary total-Whole root together: line only whenits immediately following four-space child is the exact-Truth-literalspawn call target bright into destination form with a destination suffix. Itpreserves ADR-121 zero-argument, ADR-122 one-digit, and ADR-123 positivetwo-digit behavior while keeping dim, names, general Truth, multi-argument,and full M7 semantic forms outside the new witness. A later-declared checkpointedone-Truth task target remains seed/bootstrap-identical, verified, and executable;non-task-target, wrong-parameter, erroring-parent, and missing-world priorityboundaries remain outside or above the scanner. Product forge preserves theexact seed packet with origin: seed-speak; full M7 nesting, task identity,argument, destination, effect/result, resource, ownership, and schedulerdiagnostics remain with the full compiler. Targeted red/green, direct, valid,boundary, priority, product/self-host, and tracker evidence pass. The 2026-08-14ADR-124 zero-warning release gate also passes: pack v5.0.1, 355 Markdown files /1,410 local links, four-way seed identity2F310B043653185DBE62ED61DB4D78A0DD8F2B0D20DD4B978DE634E05CF2ED68, and a440-file technical-preview package plus SHA-256SUMS with independent consumerverification. The full seed SPEAK conformance matrix and seed-native multi-fileremain residual.

Prior release-verified implementation: ADR-125 extends thesame direct AE-SEED-011 witness to the exact-Truth-literalspawn call target dim into destination form only, with a destination suffix.It preserves the ADR-124 bright witness and all prior Whole forms. Names,not dim, general Truth, multiple arguments, and full M7 semantic forms remainoutside the new seed witness. A later-declared checkpointed one-Truth task targetremains seed/bootstrap-identical, verified, and executable; product forgepreserves the seed packet, while wrong-parameter, erroring-parent, andmissing-world priority behavior retain their prior authority. Targeted red/green,direct, valid, predecessor, boundary, priority, product/self-host, and trackerevidence pass. The 2026-08-14 ADR-125 zero-warning release gate also passes:pack v5.0.1, 357 Markdown files / 1,427 local links, four-way seed identity110D8FF718D7FF578903C75398CB465C6499F5CBF2995F7C5617B4802DEFD6DB, and a442-file technical-preview package plus SHA-256SUMS with independent consumerverification. Full seed SPEAK conformance and seed-native multi-file remainresidual.

Current release-verified implementation: ADR-126 extends thesame direct AE-SEED-011 witness only to the exact empty-Textspawn call target "" into destination form, with a nonempty destinationsuffix. It preserves the ADR-124 bright, ADR-125 dim, and prior Wholeforms. The valid source target is a later-declared ordinaryweave worker [message: Text] -> Whole, not a Text task-frame parameter.Nonempty/escaped Text, Bytes, names, multiple arguments, M7 semantic forms, andfull parser/type ownership remain outside the seed witness. Targeted red/green,direct, valid, predecessor, boundary, priority, product/self-host, tracker, and45-test full seed-self-host evidence pass. The 2026-08-14 zero-warning releasegate also passes: pack v5.0.1, 359 Markdown files / 1,447 local links, four-wayseed identity D0D17756F587709BC85E323E0545281E304362288B687BAA0D48D8C334E18AFB,and a 444-file technical-preview package plus SHA-256SUMS with independentconsumer verification.

6. Maturity scorecard

Domain	Maturity	Notes
Seed self-host	High	Byte-identical multi-generation
Product vs bootstrap independence	High	Default product; residual oracle documented
Diagnostics honesty	Medium–High	Host strong; seed SPEAK pilot expanding
Multi-module	Medium	Host elaborate works; not seed-native
Native lower	Medium	Useful pilot; host-dependent
Registry	Medium	Offline + signed + lite CA; not full PKI
Task concurrency	Medium	M19e solid; no handles/timeouts/parallel
Foreign ABI	Low–Medium	Bounded pilot only
Verified-execution performance evidence	Medium	M32a fixed-corpus reports plus M32b strict profile-bound local comparison; no collected pinned baseline, attestation, or comparative speed claim
Local package ecosystem	Medium	M25 transparent locked source bundles, collision-safe cache, explicit install, and two-consumer reuse; no resolver or publisher identity
Package versioning	Stable	0.37.0 toolchain contract; language/AETH surface unchanged

7. Explicit non-goals (current law)

Ambient guest network / shell / model calls
OS-thread parallel runtime as product claim
Task handles / timeouts as implemented features
Full RFC 5280 X.509 DER
Bundled hermetic cross-compile sysroot
Claiming seed SPEAK complete for every AE-SEED code
Claiming seed-native multi-file forge

8. Recommended next increments

Ordered for dependency honesty (CONST-DEP-001):

BARP: select one broader AE-SEED-011 or AE-SEED-013 form with an explicit false-positive boundary, or a separate seed-native multi-file forge ABI design
M32 evidence operation: collect a human-declared pinned local baseline and candidate report under one M32b profile before any scoped performance-improvement claim
F-NATIVE M35k+: optional bundled/hermetic tool path when operators need reproducibility
F-REGISTRY: RFC 5280-shaped DER only if human re-authorizes beyond X.509-lite
Task runtime: implementable ADRs for handles and/or timeouts under ADR-081 invariants
M25 follow-on only by ADR: source-package signing/provenance, resolver/ranges, assets, or remote distribution must not be inferred from the local workflow

9. How to re-verify

# Constitution pack
pwsh -File "..\..\AGENTS Constitution\tools\verify-pack.ps1"

# Full project gate
pwsh -File tools\aether-gate.ps1 -Mode full
# expected: GATE PASS mode=full

# Optional packaging gate
pwsh -File tools\aether-gate.ps1 -Mode release

10. Document map

Doc	Role
MANIFEST.md	Executable product contract
CORE_CLAIMS.md	Proven vs residual claims
ROADMAP.md	Track order + human backlog
SEED_PROFILE.md	Seed emission honesty
FORGE_CONTRACT.md	Host forge ABI
DESIGN-BARP-001	BARP program
DELIVERY_REPORT-2026-08-11-M25-LOCAL-PACKAGE-PUBLICATION.md	M25 0.37 delivery / full and release evidence
DELIVERY_REPORT-2026-08-11-BARP-UNKNOWN-CALL-SPEAK.md	ADR-110 direct seed diagnostic delivery / current verification status
DELIVERY_REPORT-2026-08-11-BARP-ROOT-YIELD-UNKNOWN-CALL-SPEAK.md	ADR-111 root-yield seed diagnostic delivery / current verification status
DELIVERY_REPORT-2026-08-13-BARP-LESS-CHOOSE-SPEAK.md	ADR-112 less-choose seed diagnostic delivery / current verification status
DELIVERY_REPORT-2026-08-14-BARP-LITERAL-TRUTH-CHOOSE-SPEAK.md	ADR-113 literal-Truth seed diagnostic delivery / current verification status
DELIVERY_REPORT-2026-08-14-BARP-UNARY-LITERAL-TRUTH-CHOOSE-SPEAK.md	ADR-114 unary-literal Truth seed diagnostic delivery / current verification status
DELIVERY_REPORT-2026-08-14-BARP-WHOLE-TRUTH-YIELD-SPEAK.md	ADR-115 Whole Truth-literal yield seed diagnostic delivery / current verification status
AUDIT_REPORT-2026-08-11-FULL-PROJECT.md	This audit stamp
ADRs 043–115	Maturity decision trail

End of PROGRESS_REPORT-FULL-PROJECT.md

Introducing the "Aether" programming language.

Rob Bulkley — Tue, 04 Aug 2026 17:14:04 +0000

Aether / XLang — Full Project Progress Report

Document: Full Project Progress Report
Status: Honest institutional record (Level 4 product docs)
Authority: AGENTS Constitution (pack) + project AGENTS.md
Date: 2026-08-04
Branch: codex/xlang-local-first-studio
HEAD: 1ab0be1891bf53f1c7fafb0bc6eca2246f344776
Package: aether-core / aether-cli 0.12.0
Language surface: 0.11 / AETH v11
Related Rule IDs: CONST-GATE-001, CONST-DONE-001, CONST-COMPLETE-001,
  CONST-DEP-001, DOC-SYNC-001, ENG-WARN-001, TEST-BEHAVIOR-001,
  SEC-INPUT-001, REV-PACK-001, REL-PACKAGE-001, SOP-PHASE-001, SOP-GATE-001

1. Executive summary

Aether is a local-first programming language and CLI toolchain hosted inthis repository (historically labeled XLang). It parses only Aether source,emits verified AETH bytecode, and runs that bytecode in the Aether VM. Itdoes not transpile to Rust, C, JavaScript, LLVM, or any other language.

As of this report, the project has:

Completed the language pilot program M0–M9 within documented bounds(surface 0.11, package 0.12 tooling).
Completed the product completion arc TP-1 (integrity) + TP-2 (local technical preview) under AGENTS Constitution and SOP phases 8–10.
Designed but not implemented M10 multi-unit offline projects (ADR-013).

This is not a 1.0 general systems language. Claims below are bounded,evidence-linked, and deliberately non-marketing.

Layer	Truth
Seed-hosted compile + verify + run	Product-usable
AI structural authoring (top-level, v6)	Product-usable
Offline project integrity (single-unit pilot)	Product-usable
Local technical preview package	Delivered (`dist/`, gitignored; rebuildable)
Multi-unit projects	Designed only
Language modules / C-FFI / registry / full LSP	Not started (blocked or deferred)
Native / LLVM backend	Prohibited without law change

2. Governing law (AGENTS Constitution)

2.1 Authority stack

Level	Artifact	Role
1–3	AGENTS Constitution pack (e.g. 5.0.1)	Universal quality + process law
2	`SOP.md`	Research → design → implement → audit → harden → deliver
1	`constitution/03-DEFINITION-OF-DONE.md`	Machine-checkable Done
4	Repository `AGENTS.md`	Aether mission, invariants, stack pins
4	`MANIFEST.md`	Executable product contract

The project entry pointer loads pack AGENTS.md, SOP.md, Definition of Done,and engineering/testing/documentation standards before specialist modules.

2.2 Non-negotiable project invariants (Level 4)

Seed-hosted product compile — default CLI uses Aether-written seed;Rust bootstrap is rebuild/diagnostic authority only.
Verify before run / write — supported AETH v4–v11 only.
No host capability leak — guest AETH has no file/process/network/shellauthority; pure host fixtures only for M8 pilot.
Honest self-host claims — dual-compare where proven; no fulldiagnostic parity claim for the seed.
CLI authority boundary — local files, explicit outputs, no model/network.
legacy/ is reference only — never a production build input.
Zero-warning gate — unsafe_code = "forbid"; Clippy all = "deny".

2.3 Rule IDs this report answers to

Rule ID	Meaning	Project application
`CONST-GATE-001`	Pre-delivery checklist	`tools/aether-gate.ps1`
`CONST-DONE-001`	Definition of Done	Per milestone delivery reports
`CONST-COMPLETE-001`	No partial delivery	Vertical slices closed with docs
`CONST-DEP-001`	Dependency-first	M1 before M2…; design before M10 code
`ENG-WARN-001`	Zero warnings	Workspace lints
`TEST-BEHAVIOR-001`	Behavior + negatives	Matrices + dual-compare
`DOC-SYNC-001`	Docs match code	MANIFEST / claims / ADRs
`SEC-INPUT-001`	Untrusted input	Project paths, host, AETH verify
`REV-PACK-001`	Human review packaging	Delivery reports
`REL-PACKAGE-001`	Release artifacts	Local TP package
`SOP-PHASE-001`	Phases 1–10 order	See §4
`SOP-GATE-001`	Gate table per increment	See §10

3. Product pin (current baseline)

Layer	Value
Product name	Aether
Workspace	Rust 2021, `rust-version = "1.88"` (built on host with rustc 1.96.0 for TP)
Core crate	`aether-core` @ `crates/xlang-core` 0.12.0
CLI	`aether` @ `apps/xlang-cli` 0.12.0
Language surface	0.11
Default AETH emit	v11 (accepts verified v4–v10 as compatibility inputs)
Product compile	Seed-hosted (`compile_with_seed` / embedded seed artifact)
Bootstrap	`compile --bootstrap`, `check`, seed rebuild
Seed sources	`seed/aether_seed.ae` + checked-in `seed/aether_seed.aeth`
Seed SHA-256 (checked-in)	`6AC3C46B890029B646267E93C9FDA5CDD34F9E061D7BC0419D34E0DF6734B254`
Authoring contracts	`aether.ast/v6`, `aether.edit/v6`, `aether.diagnostic/v6`
Project schema	`aether.project/v1` (M9 pilot)
Branch / HEAD	`codex/xlang-local-first-studio` @ `1ab0be1`

Contract document: MANIFEST.mdArchitecture: ARCHITECTURE.mdClaims register: CORE_CLAIMS.mdNorth star (direction, not current behavior): NORTH_STAR.md

4. SOP scorecard (process law)

Binding process: pack SOP.md phases 1–10 (SOP-PHASE-001).

SOP #	Phase	Status for Aether program	Evidence
1	Research reference projects	Done	`docs/research/`, landscape notes
2	Decompose components	Done	Research component matrix / synthesis
3	Deep study	Done	Patterns, stop conditions
4	Design system	Done (pilots)	NORTH_STAR, designs M1–M10
5	Foundational docs	Done (pilots)	ADRs 001–013, matrices, AETHER_0.x
6	Plan engineering	Done for pilots + completion	ROADMAP, completion plan
7	Implement	Done M0–M9 pilots	Feature commits + delivery reports
8	Audit + fixes	Done for TP-1	Completion readiness audit
9	Harden	Done for TP-2 freeze	Threat model + residual risks
10	Final delivery	Done for local TP channel	TP delivery report + `dist/`

Interpretation: Language pilots completed research→implement. Productcompletion for integrity + local technical preview completed audit→harden→deliver. M10 implementation and any 1.0 platform work are new SOPcycles, not silent unfinished core.

5. Milestone program (M0–M10)

ID	Theme	Version / AETH	Status	Primary artifacts
M0	Research + AI-first foundation	—	Complete	NORTH_STAR, CORE_CLAIMS, research/, ADR-002
M1	Value/resource semantics	Design	Accepted direction	DESIGN-M1, ADR-003, M1 matrix
M2	Bounded arena + buffers	0.6 / v6	Implemented	ADR-004, AETHER_0.6
M3	Structural authoring	v1…v6 current	Implemented	ADR-005, authoring protocols
M4	`Error[Whole]` effect	0.7 / v7	Implemented	ADR-007, M4 design/matrix
M5	Deterministic comptime	0.8 / v8	Implemented	ADR-008, M5 design/matrix
M6	Dual-layout tables	0.9 / v9	Implemented	ADR-009, M6 design/matrix
M7	Structured nurseries	0.10 / v10	Implemented	ADR-010, M7 design/matrix
M8	Pure host ABI pilot	0.11 / v11	Implemented	ADR-011, M8 design/matrix
M9	Offline project tooling	package 0.12	Implemented pilot	ADR-012, M9 design/matrix
M10	Multi-unit offline projects	tooling (planned)	Designed only	ADR-013, DESIGN-M10, M10 matrix

5.1 What each implemented milestone delivered (honest bounds)

Milestone	In product	Explicit non-claims
M2	One arena, Whole/Truth buffers, closed allocate outcomes	No ambient allocator; no resource host ABI crossing
M3	Versioned AST/edit/diagnostic JSON; top-level structural edits	Not fine-grained body LSP; not a second unsafe language
M4	Abortive `Error[Whole]`, handle/forward/raise	No generic effects; no resumption; clean resource boundary
M5	Root literal `comptime bind`, 1024-cap	No macros, calls, host I/O, names at comptime
M6	`shape` + rows/columns tables	No automatic layout rewrite; Whole-only fields
M7	`together` / `spawn` nurseries	No OS threads; cooperative source order only
M8	`host weave` + `HOST_CALL`; pure fixtures `whole_inc`, `text_extent`	No C/FFI, libloading, ambient I/O
M9	`aether.project/v1`, verify, format	No registry, multi-unit graph, full LSP

5.2 Delivery report index (implementation history)

Date	Report
2026-07-28	AI-first foundation
2026-07-28	M1 resource semantics
2026-07-31	M2 bounded resources
2026-07-31	M3 structural authoring
2026-08-01	M4 typed error effect
2026-08-01	Retire Studio
2026-08-03	M5 comptime
2026-08-03	M6 layout
2026-08-03	M7 nurseries
2026-08-04	M8 host ABI
2026-08-04	M9 project tooling
2026-08-04	Technical preview TP-2

6. Architecture snapshot

┌─────────────────────────────────────────────────────────────────┐
│ Operator (trusted for host process)                              │
│  aether CLI — local files only, explicit write paths             │
└────────────────────────────┬────────────────────────────────────┘
                             │
     ┌───────────────────────┼───────────────────────┐
     ▼                       ▼                       ▼
 Bootstrap (Rust)      Seed compiler (Aether)    Project tooling
 parse/check/AST       default product compile   aether.project/v1
 authoring v6          forge ABI host            path + lock + verify
 format                                            format
     │                       │
     └───────────┬───────────┘
                 ▼
         AETH v4–v11 verify
                 ▼
         Aether VM (guest untrusted)
         pure host fixtures only (M8)

Path	Role
Seed (default)	CLI `compile`, apply-edit validation
Bootstrap	`compile --bootstrap`, `check`, seed rebuild
Forge	Host ABI: `compile [borrow source: Text] -> Bytes`

Studio: Retired (ADR-006). CLI is the sole product interface.

Legacy: legacy/ holds historical prototypes; not build inputs.

7. Seed self-host and proof posture

Claim	Status
Multi-generation seed self-host identity	Proven in `seed_self_host` tests (full run is long-pole)
Shipped `examples/*.ae` seed ≡ bootstrap	Proven (gate dual-compare + tests)
M2–M8 documented corpora dual-compare	Proven in seed self-host suite
Full invalid-source diagnostic parity (seed vs bootstrap)	Not claimed
Seed reads project JSON / multi-file link	Not claimed (and not M10 design)

Checked-in seed pin:

seed/aether_seed.aeth
SHA-256: 6AC3C46B890029B646267E93C9FDA5CDD34F9E061D7BC0419D34E0DF6734B254

Profile documentation: SEED_PROFILE.mdForge contract: FORGE_CONTRACT.md

8. CLI product surface

Command	Purpose
`aether version`	Print package version
`aether check`	Bootstrap parse/validate path
`aether structure`	Emit `aether.ast/v6`
`aether apply-edit`	Apply `aether.edit/v6`; seed-compile before write
`aether format`	Canonical format (stdout or `--output`)
`aether project verify`	Offline project integrity + per-unit seed compile
`aether compile`	Seed compile (default); `--bootstrap` optional
`aether forge`	Compile via compiler artifact ABI
`aether run`	Verify then execute AETH

Note: Successful run uses process exit 0; the program exit code isprinted (Aether 0.12.0 exited with N). Host-pilot program exit is 48.

9. Completion program (TP-1 / TP-2 / P4.1)

Human P0 freeze (2026-08-04):

Target	Definition	Status
TP-1 Integrity Complete	Claims/docs, formal audit, gate automation, tree hygiene	Done (`810f9ac`)
TP-2 Technical Preview	Threat model + local `dist/` + checksums + consumer verify	Done (`1f24a1b`)
P4.1 / M10 Multi-unit projects	Design → ADR → matrix before code	Design Done (`1ab0be1`); impl pending

9.1 TP-1 deliverables

Item	Path / evidence
Claim/doc sync	CORE_CLAIMS, ROADMAP, MANIFEST, AGENTS
Constitution dumps	`.gitignore` (never commit)
Gate script	`tools/aether-gate.ps1` (`-Mode quick` / `full`)
Audit	AUDIT_REPORT-2026-08-04-COMPLETION-READINESS.md
Quick gate	PASS (fmt, clippy, tests, 19 example dual-compare, host-pilot, project verify)

9.2 TP-2 deliverables

Item	Path / evidence
Threat model	THREAT_MODEL-TECHNICAL-PREVIEW.md
Release notes / changelog	RELEASE_NOTES-TECHNICAL-PREVIEW.md, CHANGELOG-0.12.md
Package script	`tools/package-preview.ps1`
Consumer verify	`tools/verify-preview.ps1` (also staged in package)
Package layout	`dist/aether-0.12.0-tp/` (gitignored)
Delivery report	DELIVERY_REPORT-2026-08-04-TECHNICAL-PREVIEW.md
Consumer verify result	PREVIEW VERIFY PASS
Binary SHA-256 (build host)	`E6D958F111F820C9EAD7C1B49D6D6F4C5FA9A29D1EA289C5B314D97D54833FE7`
Channel	Local folder + SHA-256SUMS only (no required git tag / GitHub Release)

9.3 M10 design (P4.1) — accepted direction, not implemented

Item	Path
Design	DESIGN-M10-MULTI-UNIT-PROJECTS.md
ADR	ADR-013-m10-multi-unit-projects.md
Matrix	M10-VALIDATION-MATRIX.md
Claim	CLM-018 Accepted direction (not Proven now)

Decision summary: multi-unit offline integrity with nested paths andindependent per-unit seed compile; no language modules/imports; keepaether.project/v1 with expanded path grammar; add project format.

10. Quality gates and how to run them

10.1 Day-to-day (TP-1 quick)

pwsh -File .\tools\aether-gate.ps1 -Mode quick

Includes: pack verify (if found), fmt --check, clippy -D warnings, core lib

m4–m7 semantic tests, CLI tests, example dual-compare, host-pilot, projectverify. Skips multi-generation seed rebuild.

10.2 Release / TP confidence (full)

pwsh -File .\tools\aether-gate.ps1 -Mode full

Adds full aether-core suite (including seed self-host) and bootstrap≡forged≡checked-in seed hash identity. Long pole (many minutes).

10.3 Technical preview package

pwsh -File .\tools\package-preview.ps1
pwsh -File .\dist\aether-0.12.0-tp\verify-preview.ps1

10.4 Manual MANIFEST gate (subset)

Documented in MANIFEST.md and project AGENTS.md Commands.

11. Claims register snapshot

Source of truth: CORE_CLAIMS.md. Summary only:

Status	Examples
Proven now	AETH verify-before-run; seed-hosted default compile; no ambient guest I/O; local-first CLI; M2–M8 bounded features; M9 single-project pilot; authoring v6
Accepted direction	Broader FFI beyond pure host (CLM-009); M10 multi-unit (CLM-018)
Research hypothesis	Further layout expansion beyond M6 (CLM-008)
Prohibited	Transpile-to-other-language (CLM-010); blanket “better than all languages” (CLM-011)

Any public comparative claim must use the scorecard in CORE_CLAIMS.

12. Security and threat posture

Topic	Posture
Guest AETH	Untrusted; no ambient OS authority
Host CLI	Trusted by the operator who launched it
Host services (product)	Pure only: `whole_inc`, `text_extent`; fail closed if missing
Project paths	Relative only; `..` / absolute / drive rejected
Locks	SHA-256 mismatch fails closed
Network / registry / model	Not in product
Formal cert / multi-tenant SaaS	Not claimed

Canonical freeze: THREAT_MODEL-TECHNICAL-PREVIEW.md.

Runtime dependencies (core): pinned serde, serde_json, sha2 — nonetwork client in the product CLI.

13. Repository map

XLang/
  AGENTS.md                 Level-4 pointer + mission
  MANIFEST.md               Product contract / release gate
  README.md                 Overview
  AUDIT_REPORT.md           Historical migration audit
  apps/xlang-cli/           aether CLI
  crates/xlang-core/        parser, compiler, verifier, VM, authoring, project
  seed/                     Aether-written seed compiler source + artifact
  examples/                 Shipped .ae + project fixture
  schemas/                  AST/edit/diagnostic/project JSON schemas
  docs/                     Specs, ADRs, matrices, delivery reports, this report
  tools/                    aether-gate, package-preview, verify-preview
  legacy/                   Reference only
  dist/                     Local TP package (gitignored)

14. Recent commit spine (completion + late pilots)

Commit	Summary
`1ab0be1`	docs(m10): multi-unit design ADR-013
`1f24a1b`	docs(tp2): technical preview package + threat model
`810f9ac`	docs(tp1): integrity complete — claims, gate, audit
`7a8024e`	feat(tooling): 0.12 offline project pilot (M9)
`b08c0b6`	feat(language): 0.11 host ABI pilot (M8)
`cdcafe7`	feat(language): 0.10 structured nurseries (M7)
`f1c7663`	feat(language): 0.9 dual-layout tables (M6)
`002e98a`	feat(language): 0.8 deterministic comptime (M5)
`bf0e1a2`	feat(language): 0.7 typed error effect (M4)
`44974fd`	refactor: retire Studio; CLI toolchain
`ae568d2`	feat(authoring): structural editing contract
`36e10b7`	feat(language): 0.6 bounded resources (M2)

15. Gaps, residual risk, and non-goals

15.1 Open engineering (next)

Item	Priority	Blocker
Implement M10 per matrix	Next feature	Design/ADR ready; needs code + tests + example
Optional `aether-gate -Mode full` timed audit entry	Hygiene	Long runtime
Language modules / imports	Later	Separate ADR + seed proof
Broader host I/O	Later	Threat model rewrite + ADR
C/FFI	Later	Ownership/threat design
Full LSP	Later	Authority boundary design
Network registry	Later	Offline-first law + threat model
Native/LLVM backend	Blocked	Requires project law change

15.2 Residual risks (accepted for preview)

Risk	Status
Operator can overwrite any explicit `--output` path	Documented host authority
Seed forge cost limits iteration	quick vs full gates
Dual pack path (external vs untracked dump)	Dumps gitignored; gate searches both
Claim drift after future features	DOC-SYNC + gate discipline
Expectation of multi-file linking	M10 docs explicitly refuse until modules ADR

15.3 Hard non-goals (current law)

Transpilation to another language
Ambient guest file/network/shell
Blanket superiority marketing
Committing constitution dump trees
Calling technical preview “1.0”

16. Definition of Done status by program arc

Arc	CONST-DONE-001?	Notes
M0–M9 language/tooling pilots	Yes (within bounds)	Each has delivery report + matrix evidence
TP-1 Integrity	Yes	Audit certificate filed
TP-2 Local technical preview	Yes for local channel	Not a public GitHub Release
M10 multi-unit	No	Design only; implementation incomplete
General 1.0 platform	No	Not a project claim

17. Rule ID self-audit (this progress report)

Rule ID	Status	Notes
CONST-GATE-001	Pass (documented)	Gate scripts exist; run before further ship
CONST-DONE-001	Pass for closed arcs	M10 not Done
CONST-COMPLETE-001	Pass for closed arcs	No half-open M0–M9 claims
CONST-DEP-001	Pass	M10 design before implementation
DOC-SYNC-001	Pass for this snapshot	Report matches HEAD `1ab0be1`
ENG-WARN-001	Pass at last TP/TP-1 gates	Re-run after next code change
TEST-BEHAVIOR-001	Pass for shipped surface	M10 matrix not green yet
SEC-INPUT-001	Pass for TP freeze	Threat model present
REV-PACK-001	Pass	This report packages human-readable progress
REL-PACKAGE-001	Pass for local TP	Rebuild via package-preview
SOP-PHASE-001	Pass	Phases scored honestly in §4
SOP-GATE-001	Pass	Gates mapped in §10

Modules conceptually loaded for this report: pack AGENTS/SOP/Done/docsstandards; project AGENTS, MANIFEST, ROADMAP, CORE_CLAIMS, threat model,completion audit, TP delivery report, M10 design package.

18. Recommended next actions (human priority)

Accept this progress report as the institutional snapshot at 1ab0be1.
Implement M10 against M10-VALIDATION-MATRIX.md(path grammar, multi-unit example, project format, tests, DOC-SYNC,delivery report) — full SOP 7 → 8 → Done.
Optionally run aether-gate -Mode full and attach wall-time/hash evidence tothe audit trail.
Archive/copy dist/aether-0.12.0-tp/ if the local preview binary set isneeded offline outside the monorepo.
Do not start C/FFI, registry, full LSP, or native backend without a newhuman-approved ADR and threat-model update.

19. One-page scorecard

Dimension	Score	Comment
Local-first AETH toolchain	Strong	Seed path, verify, VM, CLI
Evidence discipline	Strong	ADRs, matrices, dual-compare, claims register
Constitution / SOP adherence	Strong	Pack law, gates, delivery reports
Language completeness	Bounded pilot	Deliberately small surface
Multi-file projects	Designed	Not implemented
Public distribution	Local TP only	No mandated public release
1.0 readiness	No	Honest non-claim

20. Document control

Field	Value
Title	Full Project Progress Report
Location	`docs/PROGRESS_REPORT-FULL-PROJECT.md`
Supersedes	Ad-hoc chat summaries; does not replace MANIFEST or milestone delivery reports
Update trigger	After each milestone Done or completion-target change
Honesty rule	Prefer under-claim; never promote NORTH_STAR to current behavior

End of full project progress report. Prepared under AGENTS Constitution:evidence over aspiration, Rule IDs over vague compliance, Done only whenmachine-checkable and documented.

New Products available

Rob Bulkley — Wed, 22 Jul 2026 16:49:31 +0000

https://wpaistudio.gumroad.com/

Open to affiliate applications for a short time.
Affiliates get a 1% commission on every referred sale.
Apply here: https://wpaistudio.gumroad.com/affiliates

Working on the "Omni-Framework". It's a launcher/loader that can run Minecraft modpacks composed of mods from any loader family. The Grand vision is to unify the ever increasingly fragmented modding ecosystem.

Rob Bulkley — Tue, 21 Jul 2026 04:27:08 +0000

PyMason: free feedback trial, visual Python that doesn’t trap you in blocks.

Rob Bulkley — Tue, 14 Jul 2026 18:53:43 +0000

PyMason is open for a free feedback trial

Wizard Productions AI Studio (WPAI) just opened a limited free trial of PyMason, a visual, block-based Python studio designed so people can learn on it and keep shipping on it.

This is a studio beta, not a polished store launch. We’re looking for honest use, not applause.

Product page: https://wpaistudio.net/pymason.html

What PyMason is

PyMason sits between “blocks for beginners” and a real Python environment:

Drag blocks → real Python appears live
Run in the browser (Pyodide)
Dual surface: Blocks→Code, Free Python, or both
Import .py, convert toward blocks when ready
Debugger, Stage (turtle/plots), examples, tests
Bridge practice for the hard part: block → text (syntax shock is real)
Learner / Classroom / Pro framing so the tool doesn’t dead-end after week one

Aesthetic is forge, not primary-school toy — dark, ember, serious enough for adults who still want scaffolding.

Why a free trial (and why feedback is required)

We could keep polishing forever. Instead we’re asking real humans to use it and report friction.

By participating in the free trial, you agree to provide thoughtful feedback after using PyMason — what worked, where you got stuck, and whether you’d use it to learn, teach, or prototype.

That’s the deal. Free access in exchange for useful signal.

How to try it

Open the PyMason page
Click Start free trial
Agree to the feedback checkbox on the login gate
Use the trial credentials shown on the page
Load an Example, hit Run, then wander

After you’ve used it for a bit, send feedback (email link is on the product page, subject line for trial feedback is pre-suggested).

What we want to hear

Concrete > polite:

Did you get a program to Run?
Where did you get stuck or confused?
Learn / teach / prototype — which fits?
What felt like a serious tool vs a demo?
Would you pay later — for what (web / desktop / classroom)?

Harsh is fine. Vague “cool!” is less useful.

What this is not (yet)

Not a full multiplayer LMS
Not “we replaced your IDE”
Desktop/store path exists as product direction; this window is the free web trial

Some advanced toolbox categories are deep; start with Core / examples if the full set is loud.

Who it’s for

Beginners and ND learners who hate the blank page
Teachers who want real Python without starting in Notepad
Devs who sketch control flow visually, then graduate to text

If that’s you, the trial is open.

Start: https://wpaistudio.net/pymason.html

AI is in the name. The wizard is in the work.
— Wizard Productions AI Studio · wpaistudio.net

What my project bootstrapper does that uv init doesn't (and why I stopped using cookiecutter)

Rob Bulkley — Thu, 09 Jul 2026 08:24:10 +0000

The 20-minute tax

Every new Python project, same liturgy:

python -m venv .venv
.venv\Scripts\activate          # I'm on Windows, hold your condolences
# go find the .gitignore from the last project, copy it, wonder if it was the good one
pip install requests python-dotenv rich pytest   # the usual suspects
git init
git branch -M main
git add . && git commit -m "Initial commit"
# alt-tab to browser, create the GitHub repo, copy the URL
git remote add origin https://github.com/me/thing.git
git push -u origin main
# ...now what was I actually going to build?

None of this is hard. All of it is friction. And I noticed something worse than the lost 20 minutes: I was procrastinating on starting things because of it. For a side-project developer with a family and maybe eight good hours a week, the setup tax isn't an annoyance - it's a filter that kills small ideas before they exist.

So I did what we all do: wrote a script. Then, because the script was genuinely useful, I did the less common thing: turned it into a real tool with tests and opinions. This post is about the three design decisions that mattered.

"Project setup" is three different jobs

We say "setup" like it's one thing. It's three:

Environment - venv, dependencies, interpreter wiring.
Plumbing - git init, branch naming, first commit, the remote, hooks.
Opinions - layout, license, pyproject, test scaffold, what "done" looks like.

uv init does job 1 beautifully, and if that's all you need, use it and close this tab. Cookiecutter does job 3 - if you enjoy maintaining Jinja templates in a separate repo, which past-me apparently thought he would. (Narrator: he did not. The template rotted in four months.)

Almost nothing does job 2, which is strange, because job 2 is the part that's pure ritual - there are no interesting decisions in git branch -M main. RepoForge does all three in one command:

python repoforge.py my-api --profile fastapi --pre-commit --create-remote --push

Sixty seconds later: venv, curated deps, src/ layout with a running /health endpoint, README, license, pyproject.toml, a passing pytest suite, ruff+black pre-commit hooks, and a private GitHub repo with the first commit already pushed (via the gh CLI when you have it, plain remote add when you don't).

The constraints I refused to break

One file, stdlib only. A bootstrapper with its own dependency tree is a joke that writes itself. RepoForge is a single Python file you can read in one coffee and edit to taste. If you hate one of my opinions, it's right there.

Idempotent, or it's a landmine. The first version of this script - the private one I used for a year - died if you ran it twice: git remote add origin fails when origin exists, and git commit fails when there's nothing staged. Fine for me, who remembered. Unacceptable for anyone else. The rule now: every mutation goes through one gate that checks reality first.

def write(self, path: Path, content: str, description: str) -> bool:
    """Write a file unless it already exists (respected unless --force)."""
    if path.exists() and not self.force:
        self.skipped.append(str(path))
        print(f"  -- {description}: exists, left untouched ({path.name})")
        return False
    print(f"  -> {description}")
    if not self.dry_run:
        path.parent.mkdir(parents=True, exist_ok=True)
        path.write_text(content, encoding="utf-8")
    return True

Re-running RepoForge on a project you've customized will never eat your README. It tells you what it skipped and moves on. Same for git state: existing repo, existing remote, nothing-to-commit - detected, respected, reported.

--dry-run that provably touches nothing. Because every file write and subprocess call goes through that one Runner object, dry-run isn't a best-effort flag sprinkled through the code - it's a property of the architecture. There's a test that runs the entire pipeline in dry-run mode and asserts the target directory doesn't exist afterward.

Profiles, not templates

Here's the cookiecutter lesson: a template is a repository you maintain. A profile is ten lines of data:

"fastapi": Profile(
    description="Web API: FastAPI + Uvicorn + httpx test client, src layout.",
    packages=("fastapi", "uvicorn[standard]", "pydantic", "httpx", "pytest", "ruff"),
    src_layout=True,
),

Seven of them ship in v1: minimal, general, cli (Typer), fastapi, data, automation, and the one I actually built this for - ml-experiment.

I do a lot of measured AI-efficiency experiments on my own hardware, and the failure mode of hobby research isn't bad math - it's bad bookkeeping. "I think run 14 was the good one" is not a methodology. So the ml-experiment profile scaffolds the discipline, not just the folders: an EXPERIMENT_LOG.md where every entry needs a falsifiable hypothesis and a measured result, and a CLAIMS.md registry where every claim carries its evidence class - Established / Supported / Working Hypothesis / Falsified - and, crucially, its falsification condition. Failed experiments stay in the log, because a falsified claim is a result, not an embarrassment.

That file layout has survived twenty-plus iteration cycles of real work on my machines. It's the part of this tool you genuinely can't get from a generic template, because it's not a guess about how research should work - it's scar tissue.

What testing your own tool teaches you

RepoForge ships with its own 23-test pytest suite, and writing it was humbling in the specific way testing always is. The pure functions were fine. The lies were all in the seams: the idempotency guarantees I believed I had implemented versus the ones the integration test could actually demonstrate. The test that builds a real repo, edits the README, re-runs the whole pipeline, and asserts the edit survived - that test failed the first time I ran it, and the fix is why I trust the tool enough to sell it.

If you take one thing from this post that isn't about bootstrapping: the one-gate-for-all-mutations pattern (Runner above) is what made both --dry-run and the idempotency tests nearly free. Structure buys you the guarantees; discipline alone doesn't scale past one author.

Get it

RepoForge Lite - free, MIT, one file, does the whole basic job (venv + deps + git + remote): github.com/MrWizard94-Compile/repoforge. Not crippleware; it will do that job forever.
RepoForge Pro - all seven profiles, full scaffolding, pre-commit, gh integration, the test suite, and a personal license: wpaistudio.gumroad.com/l/repoforge.

I'd genuinely like to know: what's in your project-start ritual that I haven't automated? The profiles exist because mine kept diverging by project type - if yours diverges differently, that's the next profile.