Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
134 commits
Select commit Hold shift + click to select a range
51ae5d9
Build the v50 evidence-first research kernel
DevHabito Aug 2, 2026
547d5a6
Run the H50-L12 online control experiment
DevHabito Aug 2, 2026
2c54272
Pre-register the H50-L12 failure audit
DevHabito Aug 2, 2026
0dd6011
Record the posterior audit and register H50-L13
DevHabito Aug 2, 2026
f70a358
Clarify the H50-L13 zero-information fallback
DevHabito Aug 2, 2026
952ec7b
Implement H50-L13 information-directed control
DevHabito Aug 2, 2026
7697de0
Freeze the H50-L13 development selection
DevHabito Aug 2, 2026
ab8212a
Run the H50-L13 information-directed experiment
DevHabito Aug 2, 2026
097a0b4
Record the H50-L13 design provenance deviation
DevHabito Aug 2, 2026
fbd10be
Pre-register the H50-L13 failure audit
DevHabito Aug 2, 2026
ebcc68d
Implement the H50-L13 failure audit
DevHabito Aug 2, 2026
d5f689c
Run the H50-L13 failure audit
DevHabito Aug 2, 2026
53922f9
Define the cross-world transfer research gate
DevHabito Aug 2, 2026
1bf0dd9
Pre-register the cross-world transfer benchmark
DevHabito Aug 2, 2026
39f48bb
Validate the cross-world transfer benchmark
DevHabito Aug 2, 2026
488bf85
Pre-register source-learned transfer development
DevHabito Aug 2, 2026
43abc80
Run source-learned transfer development
DevHabito Aug 2, 2026
e3ebe5a
Add transfer controls and replay snapshots
DevHabito Aug 2, 2026
ae6ee33
Pre-register frozen transfer calibration
DevHabito Aug 2, 2026
8098ca4
Run frozen transfer calibration
DevHabito Aug 2, 2026
1256909
Pre-register H50-L14 predictive transfer
DevHabito Aug 2, 2026
21e88e7
Record the H50-L14 output recovery deviation
DevHabito Aug 2, 2026
dc155c0
Pre-register independent H50-L14 confirmation
DevHabito Aug 2, 2026
5465387
Confirm H50-L14 predictive transfer
DevHabito Aug 2, 2026
793c686
Pre-register contextual decision benchmark
DevHabito Aug 2, 2026
6e6a4be
Record contextual benchmark refutation
DevHabito Aug 2, 2026
9e5c671
Pre-register contextual benchmark failure audit
DevHabito Aug 2, 2026
b066040
Audit contextual benchmark refutation
DevHabito Aug 2, 2026
338ab18
Pre-register contextual benchmark replication
DevHabito Aug 3, 2026
d1046f0
Validate contextual benchmark replication
DevHabito Aug 3, 2026
e4fc269
Pre-register source-learned decision development
DevHabito Aug 3, 2026
3e4fd80
Run source-learned decision development
DevHabito Aug 3, 2026
5a02923
Pre-register contextual decision calibration
DevHabito Aug 3, 2026
4b9fe04
Run contextual decision calibration
DevHabito Aug 3, 2026
c043fdb
Pre-register H50-L15 contextual transfer
DevHabito Aug 3, 2026
176c63c
Confirm H50-L15 contextual transfer
DevHabito Aug 3, 2026
fa30e8c
Pre-register reward-only compatibility development
DevHabito Aug 3, 2026
6e41eb1
Record reward-only compatibility failure
DevHabito Aug 3, 2026
9949429
Pre-register compatibility feedback failure audit
DevHabito Aug 3, 2026
f7e4236
Record compatibility feedback audit
DevHabito Aug 3, 2026
041a032
Pre-register cellwise safe-transfer development
DevHabito Aug 3, 2026
e3d8986
Record cellwise safe-transfer failure
DevHabito Aug 3, 2026
6de662f
Add explicit multi-step goal continuation
DevHabito Aug 3, 2026
ef5880a
Pre-register integrated cognitive-cycle development
DevHabito Aug 3, 2026
d5d33d2
Record integrated cognitive-cycle development result
DevHabito Aug 3, 2026
51d3e13
Pre-register integrated durability calibration
DevHabito Aug 3, 2026
79c80ae
Record integrated durability calibration result
DevHabito Aug 3, 2026
ad304f6
Pre-register deterministic integrated-cycle confirmation
DevHabito Aug 3, 2026
47298cb
Record deterministic integrated-cycle confirmation
DevHabito Aug 3, 2026
5e250e5
Pre-register online alignment development
DevHabito Aug 3, 2026
8d4fc6b
Record online alignment development result
DevHabito Aug 3, 2026
0aca36f
Pre-register online alignment calibration
DevHabito Aug 3, 2026
03d128f
Record online alignment calibration result
DevHabito Aug 3, 2026
3c442d7
Pre-register online alignment confirmation
DevHabito Aug 3, 2026
9cfce31
Record online alignment confirmation result
DevHabito Aug 3, 2026
83b42a7
Run wake guardian fully hidden
DevHabito Aug 4, 2026
0ca6a7f
Add provider-neutral language boundary
DevHabito Aug 9, 2026
fd93158
Add language conformance development corpus
DevHabito Aug 9, 2026
0508730
Add blind language annotation protocol
DevHabito Aug 9, 2026
0ed609f
Fix annotation guide surface check
DevHabito Aug 9, 2026
956337d
Pre-register language calibration promotion
DevHabito Aug 9, 2026
a99575d
Pre-register persistent desktop runtime
DevHabito Aug 9, 2026
99f4938
Add pure desktop runtime foundation
DevHabito Aug 9, 2026
2602c57
Record E043 automated admission
DevHabito Aug 9, 2026
61d6f97
Pre-register conversational development runtime
DevHabito Aug 9, 2026
ca04a2c
Add isolated conversational development runtime
DevHabito Aug 9, 2026
61df93c
Make E043 freeze check line-ending portable
DevHabito Aug 9, 2026
6f3cd02
Keep freeze blob check shallow-clone safe
DevHabito Aug 9, 2026
3313708
Pre-register portable local language seed
DevHabito Aug 17, 2026
34c4ee7
Add portable local language seed
DevHabito Aug 17, 2026
6972c8b
Record E045 engineering admission
DevHabito Aug 17, 2026
4eaeddb
Lock E045 local artifacts
DevHabito Aug 17, 2026
7c7ffbc
Record E045 model load probe
DevHabito Aug 17, 2026
c8cc818
Record E045 first-turn failure
DevHabito Aug 17, 2026
d9fadec
Pre-register native completion repair
DevHabito Aug 17, 2026
c1b6946
Implement authenticated native completion transport
DevHabito Aug 17, 2026
ffe3f5a
Record E046 admission failure
DevHabito Aug 17, 2026
4368cd3
Pre-register control-token rejection repair
DevHabito Aug 17, 2026
a1d4e4f
Reject local model control tokens
DevHabito Aug 17, 2026
65ddbc5
Record E047 engineering admission
DevHabito Aug 17, 2026
1ada454
Record E047 live input-integrity failure
DevHabito Aug 17, 2026
2cb30ac
Pre-register UTF-8 console repair
DevHabito Aug 17, 2026
cd23929
Enforce UTF-8 on local conversation streams
DevHabito Aug 17, 2026
637c68a
Record E048 engineering admission
DevHabito Aug 17, 2026
1aa12b6
Record E048 exact UTF-8 live turn
DevHabito Aug 17, 2026
9287b11
Document exact local UTF-8 milestone
DevHabito Aug 17, 2026
0796ea8
Pre-register local conversation development screen
DevHabito Aug 17, 2026
cf7a369
Record E049 local conversation failures
DevHabito Aug 17, 2026
4595311
Pre-register current-turn expression repair
DevHabito Aug 17, 2026
ec624e4
Direct local expression at the current turn
DevHabito Aug 17, 2026
e858725
Record E050 engineering admission
DevHabito Aug 17, 2026
19dd567
Record E050 expression repair failure
DevHabito Aug 17, 2026
65ade85
Pre-register free 1.5B model comparison
DevHabito Aug 17, 2026
d827a60
Restore baseline prompt for model comparison
DevHabito Aug 17, 2026
80242ab
Lock E051 candidate artifact
DevHabito Aug 17, 2026
3c3dd32
Record E051 engineering admission
DevHabito Aug 17, 2026
c095ef2
Record E051 load admission
DevHabito Aug 17, 2026
5092955
Record E051 runtime failure
DevHabito Aug 17, 2026
2b8a984
Pre-register local inference bottleneck diagnostic
DevHabito Aug 17, 2026
e87dcc4
Record E052 throughput bottleneck
DevHabito Aug 17, 2026
5f5f4fc
Pre-register free 0.8B edge model screen
DevHabito Aug 17, 2026
e38a400
Lock E053 candidate artifact
DevHabito Aug 17, 2026
101f735
Record E053 load admission
DevHabito Aug 17, 2026
4674014
Record E053 performance admission
DevHabito Aug 17, 2026
54cb8b6
Record E053 engineering admission
DevHabito Aug 17, 2026
acc6e78
Record E053 gateway contract failure
DevHabito Aug 17, 2026
37598e2
Pre-register local numeric grammar repair
DevHabito Aug 17, 2026
c8ff493
Implement E054 local numeric grammar repair
DevHabito Aug 17, 2026
3c39073
Record E054 engineering admission
DevHabito Aug 17, 2026
680705d
Record invalid E054 live execution
DevHabito Aug 17, 2026
ad1afee
Pre-register gated local conversation screen
DevHabito Aug 17, 2026
009d9ac
Implement E055 gated measurement runner
DevHabito Aug 17, 2026
d55b305
Record E055 runner admission
DevHabito Aug 17, 2026
2afb946
Record E055 stale-expression failure
DevHabito Aug 17, 2026
8c4234f
Document free local language results
DevHabito Aug 17, 2026
432cd79
Fix maintained-surface fixture handling
DevHabito Aug 18, 2026
b383fc4
replace legacy voice surface with v50 host
DevHabito Aug 18, 2026
b526361
preregister Granite edge admission
DevHabito Aug 18, 2026
c8c6705
admit E057 measurement runner
DevHabito Aug 18, 2026
69cdadc
record E057 Granite performance failure
DevHabito Aug 18, 2026
bb1617e
preregister H-350M edge admission
DevHabito Aug 18, 2026
489e27b
admit E058 measurement profile
DevHabito Aug 18, 2026
4626258
record invalid E058 launcher attempt
DevHabito Aug 18, 2026
28fae63
repair direct E058 launcher import
DevHabito Aug 18, 2026
3bf7197
record passing E058 edge admission
DevHabito Aug 18, 2026
7678b9b
preregister Granite control boundary
DevHabito Aug 18, 2026
466af5c
add Granite control-token boundary
DevHabito Aug 18, 2026
c3eb795
admit E059 token conformance runner
DevHabito Aug 18, 2026
f9edffb
record passing E059 control boundary
DevHabito Aug 18, 2026
2364765
preregister H350M Portuguese conversation screen
DevHabito Aug 18, 2026
9f4d71e
admit E060 Portuguese screen runner
DevHabito Aug 18, 2026
0759ec3
freeze E060 engineering admission
DevHabito Aug 18, 2026
e7564f0
record failed E060 conversation screen
DevHabito Aug 18, 2026
4fd4208
preregister consent-locked Gemma admission
DevHabito Aug 18, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
18 changes: 18 additions & 0 deletions .editorconfig
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
root = true

[*]
charset = utf-8
end_of_line = lf
insert_final_newline = true
trim_trailing_whitespace = true

[*.py]
indent_style = space
indent_size = 4

[*.{md,json,yml,yaml,toml}]
indent_style = space
indent_size = 2

[*.bat]
end_of_line = crlf
23 changes: 23 additions & 0 deletions .github/PULL_REQUEST_TEMPLATE.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,23 @@
## What changed

Describe the code, protocol, or documentation change in plain language.

## Why

State the problem or falsifiable question this change addresses.

## Validation

List the exact commands and observed results.

## Evidence boundary

Explain what the change supports and what remains unsupported.

## Checklist

- [ ] Final seeds were not used during development.
- [ ] New claims have registered baselines and thresholds.
- [ ] Failed criteria remain visible.
- [ ] Runtime databases, logs, exports, and credentials are not included.
- [ ] Maintained documentation is written in direct English.
31 changes: 31 additions & 0 deletions .github/workflows/ci.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,31 @@
name: CI

on:
push:
branches: [main]
pull_request:

permissions:
contents: read

jobs:
test:
runs-on: windows-latest
steps:
- uses: actions/checkout@v4

- uses: actions/setup-python@v5
with:
python-version: "3.12"
cache: pip

- name: Install package
run: python -m pip install -e .

- name: Check maintained repository surface
run: python scripts/check_maintained_surface.py

- name: Run test suite
env:
PYTHONPATH: src
run: python -m unittest discover -s tests -v
21 changes: 10 additions & 11 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -16,15 +16,14 @@ Thumbs.db
baselines/
v47_patch_backups/
v48_patch_backups/
darwin_home/backups/
darwin_home/logs/
darwin_home/snapshots/
darwin_home/music_cache_v49_16/

# SQLite transient files. The current Darwin memory DB is intentionally kept.
darwin_home/*.db-journal
darwin_home/*.db-shm
darwin_home/*.db-wal
darwin_home/*.sqlite-journal
darwin_home/*.sqlite-shm
darwin_home/*.sqlite-wal
# Darwin runtime state belongs to the local machine, not to source control.
darwin_home/*
!darwin_home/config.example.json

# Build artifacts.
build/
dist/
*.egg-info/
.coverage
htmlcov/
8 changes: 7 additions & 1 deletion Abrir_Darwin_Acordar_Com_Voz.bat
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,12 @@ if %errorlevel%==0 (
exit /b
)

set "DARWIN_CODEX_PYTHONW=%USERPROFILE%\.cache\codex-runtimes\codex-primary-runtime\dependencies\python\pythonw.exe"
if exist "%DARWIN_CODEX_PYTHONW%" (
start "" "%DARWIN_CODEX_PYTHONW%" darwin_wake_word_guardian_v49_34.py
exit /b
)

echo Nao encontrei Python no PATH.
echo Abra pelo Codex com: py darwin_wake_word_guardian_v49_34.py
echo Tambem nao encontrei o runtime Python fornecido pelo Codex.
pause
70 changes: 70 additions & 0 deletions CONTRIBUTING.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,70 @@
# Contributing to Darwin

Darwin is easier to improve when claims stay smaller than the code that supports
them. A useful contribution makes the evidence clearer, the failure mode harder
to hide, or the architecture easier to test.

## Development setup

```powershell
py -m pip install -e .
$env:PYTHONPATH = "src"
py -m unittest discover -s tests -v
```

Python 3.11 or newer is required.

## Where new work belongs

- package code: `src/darwin_v50`
- tests: `tests`
- experiment records: `docs/v50`
- architecture and maintenance guides: `docs`

Do not add another versioned Python program to the repository root. Root-level
v47-v49 files are a compatibility surface and will be migrated separately.

## Experiment checklist

Before running final seeds, write down:

- the exact capability claim;
- development, calibration, and final seed families;
- what information the agent can observe;
- baselines and ablations;
- metrics and conjunctive thresholds;
- tie-breaking and selection rules;
- snapshot and causal-order invariants;
- conditions that refute the hypothesis;
- the strongest evidence level the evaluator can support.

After the final run, record every metric and every failed criterion. Do not tune
against final seeds and then present the same seeds as held out.

## Writing style

Use direct, ordinary English. Avoid promotional claims, synthetic enthusiasm,
and vague phrases such as “revolutionary intelligence” or “human-like
understanding.” Prefer a concrete sentence:

> The planner solved 96 of 100 registered tasks.

over an interpretation the experiment did not test:

> Darwin now understands the world.

State limitations close to the result they qualify. A failed hypothesis is a
valid result and should remain visible.

## Pull requests

A pull request should explain:

- what changed;
- why the change was needed;
- which claim or invariant it affects;
- how it was tested;
- what remains unsupported.

Keep runtime databases, logs, exports, credentials, and local snapshots out of
Git.
Loading
Loading