fpf-thinking-map 1.4.4__tar.gz → 1.4.6__tar.gz

{fpf_thinking_map-1.4.4 → fpf_thinking_map-1.4.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: fpf-thinking-map
-Version: 1.4.4
+Version: 1.4.6
 Summary: Agentic thinking map for bounded LLM traversal, with TTL evidence decay, deterministic guards, and propositional logic. Compiled from FPF (First Principles Framework).
 Author-email: "prichindel.com" <igareosh@igareosh.com>
 License-Expression: MIT
@@ -24,9 +24,9 @@ Description-Content-Type: text/markdown
 License-File: LICENSE
 Dynamic: license-file
 
-# prichindel.com Agentic Thinking Map
+# FPF Agentic Thinking Map
 
-**v1.4.3** — [FPF (First Principles Framework)](https://github.com/ailev/FPF), compiled into a small traversal map for LLM agents.
+**v1.4.5** — [FPF (First Principles Framework)](https://github.com/ailev/FPF), compiled into a small traversal map for LLM agents.
 
 A Python package that gives a model a bounded move board instead of a giant framework to digest at runtime. Instead of rereading a sprawling semantic corpus and improvising from it, the model gets a small JSON slice: what context it is in, what move is open, what evidence is missing, what is risky, and what outcome class applies.
 
@@ -336,4 +336,4 @@ MIT. See [LICENSE](LICENSE).
 
 ---
 
-**prichindel.com** — v1.4.3 — 2026-07-08
+**prichindel.com** — v1.4.5 — 2026-07-09

{fpf_thinking_map-1.4.4 → fpf_thinking_map-1.4.6}/README.md

@@ -1,6 +1,6 @@
-# prichindel.com Agentic Thinking Map
+# FPF Agentic Thinking Map
 
-**v1.4.3** — [FPF (First Principles Framework)](https://github.com/ailev/FPF), compiled into a small traversal map for LLM agents.
+**v1.4.5** — [FPF (First Principles Framework)](https://github.com/ailev/FPF), compiled into a small traversal map for LLM agents.
 
 A Python package that gives a model a bounded move board instead of a giant framework to digest at runtime. Instead of rereading a sprawling semantic corpus and improvising from it, the model gets a small JSON slice: what context it is in, what move is open, what evidence is missing, what is risky, and what outcome class applies.
 
@@ -310,4 +310,4 @@ MIT. See [LICENSE](LICENSE).
 
 ---
 
-**prichindel.com** — v1.4.3 — 2026-07-08
+**prichindel.com** — v1.4.5 — 2026-07-09

{fpf_thinking_map-1.4.4 → fpf_thinking_map-1.4.6}/fpf_thinking_map/README.md

@@ -272,4 +272,4 @@ outcome = engine.step(state, transition_id="start_to_reviewed")
 - `demo_walk()` auto-fires the first available transition. It is for testing and examples only, not for operational use. In real use, the LLM calls `step()` and `attempt_transition()` with explicit choices.
 - Publications are registered on the map but stay out of the step/guard/logic path. They are for publish-type moves only.
 
-prichindel.com | 2026-07-08 | v1.4.3
+prichindel.com | 2026-07-09 | v1.4.4

fpf_thinking_map-1.4.6/fpf_thinking_map/RELATED_WORK_MILTONIAN_PRINCIPLES.md

@@ -0,0 +1,33 @@
+# Related work: miltonian/principles
+
+**Status**: Acknowledged — independent parallel work, not incorporated, not a shared source.
+**Date**: 2026-07-10
+**Decision by**: igareosh (prichindel.com)
+**Related repo**: [miltonian/principles](https://github.com/miltonian/principles)
+**Author's writeup**: [OpenAI Community — Principles Framework: generate AI agents using first-principles reasoning](https://community.openai.com/t/principles-framework-generate-ai-agents-using-first-principles-reasoning/1045890)
+
+---
+
+## What we found
+
+A "Principles Framework" by Alexander Hamilton (miltonian) — a TypeScript pipeline, built on the Claude Agent SDK, that takes a goal and generates a team of specialized agents by applying first-principles decomposition: survey the landscape, derive typed "truths," adversarially vet each one (a skeptic tries to kill it), decompose into subtasks against a coverage map, judge the decomposition against a rubric, refine, then run a separate frame-level skeptic that can reject the whole frame as solving the wrong problem.
+
+No shared lineage with this package. This package compiles Levenchuk's FPF spec (see `SOURCES.md`); `principles` does not reference FPF and was built independently, from OpenAI Swarm and a multi-agent-systems essay as its stated influences. Two unrelated projects converging on "first principles" as the operating name — worth recording so nobody assumes one derives from the other.
+
+Full respect for the effort, in particular the part most projects in this space skip: a real controlled benchmark (`benchmarks/research-pilot/`) comparing the bare model against the compiled pipeline on Scale AI's ResearchRubrics, graded by a different model family to avoid home-team bias, with predictions pre-registered in writing before the final run and losses reported alongside wins. That is a materially higher evidence bar than most "framework improves LLM output" claims ship with.
+
+## What the benchmark actually shows
+
+Their first version of the compiled pipeline *lost* to the bare model — 0.552 vs 0.653 mean score — because structure without a verified synthesis step degraded the final render (wrong format, leaked scaffolding text, truncated output). Three more iterations (synthesis contract + gates, deliverable framing, then grounded framing with a frame-level skeptic and a specificity-retention gate) were needed before the compiled pipeline beat bare: 0.698 vs 0.653, winning 7 of 9 tasks. Stage-attribution on the transcripts found ~80% of lost weight in an intermediate version was decided at the truths step, before any downstream check could catch it.
+
+That v1 result is an independent, empirical confirmation of this package's own `WHY_THIS_EXISTS.md` thesis from a completely different architecture and author: bolting a first-principles framework onto a model's reasoning path, unverified, does not just fail to help — it actively degrades output, and the damage is concentrated wherever nobody thought to add a check.
+
+## Where the designs diverge
+
+`principles` keeps the model doing framework-driven reasoning at every stage, and fixes the failure mode by adding adversarial skepticism and grounding around that reasoning — an LLM call per stage, judged and refined until it converges. This package removes the model from framework-reasoning entirely: FPF is compiled once, at build time, into a small deterministic engine, and the model receives a precomputed JSON slice at runtime with zero framework-reasoning calls in the loop (see `RELATED_WORK_GOFLOW_FPF_SKILL.md` for the same divergence against a different independent project).
+
+Two valid answers to the same observed failure — verify the model's use of the framework at every step, or remove the model from framework-reasoning so there is nothing to verify at runtime. Nothing from `principles` was pulled into this package; there is no code overlap. This note exists to record that we are aware of it, that it independently reproduces our core finding under a harder empirical standard than we hold ourselves to, and that the effort deserves the credit for doing so.
+
+---
+
+prichindel.com | 2026-07-10 | v1.4.5

{fpf_thinking_map-1.4.4 → fpf_thinking_map-1.4.6}/fpf_thinking_map.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: fpf-thinking-map
-Version: 1.4.4
+Version: 1.4.6
 Summary: Agentic thinking map for bounded LLM traversal, with TTL evidence decay, deterministic guards, and propositional logic. Compiled from FPF (First Principles Framework).
 Author-email: "prichindel.com" <igareosh@igareosh.com>
 License-Expression: MIT
@@ -24,9 +24,9 @@ Description-Content-Type: text/markdown
 License-File: LICENSE
 Dynamic: license-file
 
-# prichindel.com Agentic Thinking Map
+# FPF Agentic Thinking Map
 
-**v1.4.3** — [FPF (First Principles Framework)](https://github.com/ailev/FPF), compiled into a small traversal map for LLM agents.
+**v1.4.5** — [FPF (First Principles Framework)](https://github.com/ailev/FPF), compiled into a small traversal map for LLM agents.
 
 A Python package that gives a model a bounded move board instead of a giant framework to digest at runtime. Instead of rereading a sprawling semantic corpus and improvising from it, the model gets a small JSON slice: what context it is in, what move is open, what evidence is missing, what is risky, and what outcome class applies.
 
@@ -336,4 +336,4 @@ MIT. See [LICENSE](LICENSE).
 
 ---
 
-**prichindel.com** — v1.4.3 — 2026-07-08
+**prichindel.com** — v1.4.5 — 2026-07-09

{fpf_thinking_map-1.4.4 → fpf_thinking_map-1.4.6}/fpf_thinking_map.egg-info/SOURCES.txt

@@ -9,6 +9,7 @@ fpf_thinking_map/README.md
 fpf_thinking_map/REJECTED_C32_CANDIDATE_SYNTHESIS.md
 fpf_thinking_map/REJECTED_NQD_OEE_CULTURAL_EVOLUTION.md
 fpf_thinking_map/RELATED_WORK_GOFLOW_FPF_SKILL.md
+fpf_thinking_map/RELATED_WORK_MILTONIAN_PRINCIPLES.md
 fpf_thinking_map/SOURCES.md
 fpf_thinking_map/WHY_THIS_EXISTS.md
 fpf_thinking_map/__init__.py

{fpf_thinking_map-1.4.4 → fpf_thinking_map-1.4.6}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "fpf-thinking-map"
-version = "1.4.4"
+version = "1.4.6"
 description = "Agentic thinking map for bounded LLM traversal, with TTL evidence decay, deterministic guards, and propositional logic. Compiled from FPF (First Principles Framework)."
 readme = "README.md"
 license = "MIT"