software-as-a-besides	Software as a Besides	Type. The computer shows. Software that teaches you to not need it.	besides SaaB intent translation subtract shell kernel literacy	https://jnous.com/software-as-a-besides.txt
lookdown	Lookdown	Pattern matching intent. One file answers from disk. File shrinks as the adapter	lookdown TSV pattern-matching intent routing adapter pruning subtract	https://jnous.com/LOOKDOWN.txt
variance-lab	Variance Lab	28 models. 100k+ inferences. N=100. Cross-platform. Findings from running models	variance-lab empirical inference local-LLM measurement methodology N=100 cross-platform	https://jnous.com/variance-lab.txt
lineage	Lineage	The people whose code subtract depends on. Fox, Ramey, Korn, McIlroy, Thompson, 	lineage bash ksh ssh sshsig llama.cpp coprocesses pipes grep subtract dependencies	https://jnous.com/lineage.txt
emile	Emile	Local-first deployment pattern for sole proprietors. One device, one model, one 	emile deployment local-first sole-proprietor Pi5 Mac-Mini clinical signing substrate	https://jnous.com/emile.txt
hosuni	Hosuni	Shell dispatcher. Routes intent through subtract, lookdown, and inference. 68000	hosuni dispatcher shell routing intent inference Amiga 68000 subtract	https://jnous.com/hosuni.sh
soul	SOUL.txt	Signon file for agents. Reflexes, authority, verification. Read before acting.	SOUL signon agent reflexes authority verification governance subtract	https://jnous.com/SOUL.txt
please	PLEASE	Provide Legal Exculpation and Sign Everything. Attestation wallets for clinical 	PLEASE attestation signing clinical call-center exculpation workflow	https://jnous.com/please.txt
please-clinical	PLEASE Clinical	Clinical deployment of PLEASE. ABA session logging, HIPAA integrity, insurance b	PLEASE clinical ABA HIPAA session-logging billing insurance 837P	https://jnous.com/please-clinical.txt
please-economics	PLEASE Economics	Cost model for PLEASE deployments. Hardware, inference, cloud change windows.	PLEASE economics cost hardware inference deployment pricing	https://jnous.com/please-economics.txt
boot	Boot	Boot sequence for the formation. What runs, in what order, on which node.	boot sequence formation startup nodes	https://jnous.com/boot.txt
pattern	Pattern	The pattern subtract follows. Intent to command, file shrinks, adapter learns.	pattern intent command subtract lookdown adapter learning	https://jnous.com/pattern.txt
primitives	Primitives	The kernel primitives subtract depends on. fork, pipe, exec, namespaces.	primitives kernel fork pipe exec namespaces syscall Unix	https://jnous.com/primitives.txt
invariants	Invariants	What does not change. The constraints that hold across all subtract deployments.	invariants constraints subtract deployment rules	https://jnous.com/invariants.txt
subtracting	Subtracting	Why subtract exists. The thesis, lineage, and evidence.	subtracting thesis lineage evidence subtract why	https://jnous.com/subtracting.txt
narrow	Narrow	The narrow path. What subtract commits to and what it does not.	narrow scope subtract commitment boundary	https://jnous.com/narrow.txt
stack-simple	Stack	The subtract dependency stack. What is already there vs what gets added.	stack dependencies kernel shell subtract	https://jnous.com/stack-simple.txt
squared-c	squared.c	Source code for squared.c	squared-c runtime source-code subtract	https://jnous.com/squared.c
parsimony-c	parsimony.c	Source code for parsimony.c	parsimony-c runtime source-code subtract	https://jnous.com/parsimony.c
three-questions	Three Questions for Agentic Autonomy	Universal intake sequence. Human capability boundary first, autonomy blockers se	autonomy automation intake workflow blocker access subtractive-security methodology capability-boundary assessment	https://jnous.com/01-three-questions.txt
delegation-vs-inline	Delegation vs Inline Execution	Delegated parallel execution across 3 nodes completed identical tasks in 48% of 	delegation parallel execution multi-agent context-window sequential fresh-context speedup wall-clock	https://jnous.com/02-delegation-vs-inline.txt
interaction-mode-variance	Interaction Mode Variance	Passenger mode consumes 41x more tokens than governor mode. 88 sessions, single 	passenger governor tokens cost human-attention mode 41x ratio original	https://jnous.com/03-interaction-mode-variance.txt
authorization-gap	Delegated Agent Authorization Gap	Agents fail at auth boundaries, not capability boundaries. The blocker is infras	authorization agent failure auth OAuth MFA browser-redirect automation-blocker infrastructure headless	https://jnous.com/04-authorization-gap.txt
governance-binding	Governance Binding	Governance binds at N=30. 81% overall, reflex-dependent, delivery-path sensitive	governance binding adapter reflex compliance weights delivery-path system-prompt SmolLM alignment	https://jnous.com/05-governance-binding.txt
interaction-mode-rerun	Interaction Mode Rerun	52.7x passenger/governor token ratio under pre-committed rubric. N=501 qualifyin	passenger governor tokens cost replication methodology pre-committed rubric 52x validation	https://jnous.com/06-interaction-mode-rerun.txt
http2-vs-http1	HTTP/2 vs HTTP/1.1	HTTP/2 multiplexing delivers 2.1x throughput for local inference concurrency. 70	HTTP/2 HTTP/1.1 multiplexing llama-server local-inference concurrency throughput protocol Apple-Silicon	https://jnous.com/07-http2-vs-http1.txt
1bit-quantization	1-Bit Quantization	1-bit quantization breaks the 8GB deployment ceiling. Bonsai 8B runs on 8GB M2 M	1-bit quantization 8GB 8B deployment local-inference ceiling RAM Q4 Q8	https://jnous.com/08-1bit-quantization.txt
1bit-hardware-tiers	1-Bit Across Hardware Tiers	1-bit wins across 4 hardware tiers because the bottleneck differs per tier. 8GB 	1-bit Q4 Q8 quantization-comparison hardware-tiers memory-bandwidth compute bottleneck 8GB 64GB scaling	https://jnous.com/09-1bit-hardware-tiers.txt
throughput-ceiling	Throughput Ceiling	Aggregate inference throughput plateaus at a hardware-determined ceiling. Adding	throughput ceiling concurrency saturation memory-bandwidth scaling plateau optimal-concurrency Apple-Silicon	https://jnous.com/10-throughput-ceiling.txt
review-vs-verification	Review vs Verification	Model code review and functional verification have asymmetric yield. Cheaper mod	code-review verification crash effort-inversion cheaper-model C functional-testing three-gate	https://jnous.com/11-review-vs-verification.txt
lookdown-routing	Lookdown Routing	grep beats inference for known-answer retrieval. A TSV lookup returns the correc	routing grep TSV lookup known-answer fast-path skip-inference deterministic retrieval	https://jnous.com/12-lookdown-routing.txt
manifest-vs-bm25	Manifest vs BM25	Manifest-routed retrieval beats BM25 for small-corpus QA. The routing table know	retrieval BM25 RAG small-corpus curated routing document-selection human-curated	https://jnous.com/13-manifest-vs-bm25.txt
governance-refusal	Governance Refusal	In-vivo reproduction of finding 05 during scorer calibration. The model refused 	governance refusal production real-world unprompted adapter validation emergent alignment	https://jnous.com/14-governance-refusal.txt
reflex-binding	Reflex Binding	Abstract reflexes transfer only with lineage, not instruction. 6 models, 2 arms,	binding lineage instruction fine-tune weights LoRA system-prompt trained prompted disposition alignment	https://jnous.com/15-reflex-binding.txt
effort-dependent-binding	Effort-Dependent Binding	Governance binding varies by effort tier. 4 frontier models, 4-6 effort levels, 	effort tier extended-thinking compliance governance alignment non-monotonic compute-budget reasoning	https://jnous.com/16-effort-dependent-binding.txt
handler-substrate-selection	Handler Substrate Selection	Three-gate methodology for picking which model runs handler.sh. 240 trials, 3 mo	handler model-selection tool-call function-calling dispatch gate nano confabulation small-model	https://jnous.com/17-handler-substrate-selection.txt
teshuvah-of-the-lips	Teshuvah of the Lips	Bare reflection under review-pressure produces confident confabulation in the	teshuvah reflection sycophancy abstention deployment-gate bare-reflection hold-rate confabulation RLHF kids correction-register	https://jnous.com/18-teshuvah-of-the-lips.txt
ark	ark.c	Archive/harvest tool — pulls conversation data into the living-ledger pipeline.	ark harvest archive living-ledger pipeline C source	https://jnous.com/ark.c
extract	extract.c	Extracts conversation turns from raw exports into flat records.	extract harvest conversation flatten C source living-ledger	https://jnous.com/extract.c
conv-flatten	conv_flatten.c	Flattens nested conversation JSON into training-pair rows.	conv-flatten flatten conversation JSON pairs C source	https://jnous.com/conv_flatten.c
scrub	scrub.c	PII/secret redaction as a unix filter — the fail-closed gate before curation.	scrub PII redaction secret filter fail-closed gate C source	https://jnous.com/scrub.c
curate	curate.c	Rubric-driven curation gate — ternary keep/cut/hold over harvested pairs.	curate rubric ternary gate curation harvest C source	https://jnous.com/curate.c
names	names.c	Name handling for the ledger pipeline.	names ledger pipeline C source	https://jnous.com/names.c
decode	decode.c	Decode step in the harvest→sign toolchain.	decode harvest toolchain C source	https://jnous.com/decode.c
commit	commit.c	Commits a curated entry into the signed hash chain.	commit chain sign hash-chain ledger C source	https://jnous.com/commit.c
seascroll	seascroll.c	seascroll — source.	seascroll C source runtime	https://jnous.com/seascroll.c
starscroll	starscroll.c	starscroll — source.	starscroll C source runtime	https://jnous.com/starscroll.c
syna-poll	syna_poll.c	syna_poll — source.	syna-poll C source runtime	https://jnous.com/syna_poll.c
subtract-sh	subtract.sh	The subtract shell layer — command_not_found_handle, tiered routing, the entry point.	subtract shell command-not-found routing lookdown handler source	https://jnous.com/subtract.sh
hosuni-discord	hosuni-discord.sh	Hosuni Discord bridge.	hosuni discord bridge dispatcher shell source	https://jnous.com/hosuni-discord.sh
lookdown-merge	lookdown-merge.sh	Merges per-node lookdown TSVs into the universal table.	lookdown merge TSV universal table shell source	https://jnous.com/lookdown-merge.sh
install	install.sh	One-line installer — curl | bash, the stranger front door.	install installer curl bash bootstrap shell source	https://jnous.com/install.sh
kiosk	kiosk.sh	Kiosk launcher for the kids box.	kiosk launcher kids box shell source	https://jnous.com/kiosk.sh
# findings.md · 2026-08-11 · r5 · 55 entries · jnous.com · verify: ssh-keygen -Y verify -f authorized_signers -I hodori@subtract.ing -n subtract.ing -s findings.md.sig < findings.md
quantization-adapter-calibration	Quantization Adapter Calibration	Lineage at 2 bits beats instruction at full precision (+40 at matched quant). Adapter reshapes which reflexes survive: routing dies first, refusal last. Gate, not disposition, carries safety.	quantization adapter governance calibration confidence lineage compression Finding-15 Finding-19	https://jnous.com/19-quantization-adapter-calibration.txt
positional-adherence	Positional Constraint Adherence	System channel binds at 177/178 across three Opus generations (ceiling flagged). Body authority is generational: fitted starting intercepts 66/51/99. Count outcome shows every generation leaks more with depth (z=10-17); generations set the LEVEL of the leak, not its existence. v1 withdrawn for harness defects; the instrument is the finding.	positional adherence constraint channel depth generation count-scorer randomized Finding-21	https://jnous.com/21-positional-adherence-f21.txt