{ "status": "ok", "author": "Taylor Amarel", "canonical_url": "https://blog.tayloramarel.com", "feed_xml_url": "https://blog.tayloramarel.com/feed.xml", "count": 10, "posts": [ { "slug": "better-model-or-better-harness", "title": "Better Model or Better Harness? Breaking Down the Six-Part Agent Harness", "date": "2026-10-18", "category": "Agent Architecture & Ablations", "reading_time": "9 min read", "word_count": 1482, "summary": "How a 9B to 27B open-weight model inside a verified 6-part harness beats an unpaginated frontier model on multi-step tool tasks, with full ablations.", "html_url": "https://blog.tayloramarel.com/posts/better-model-or-better-harness.html", "markdown_url": "https://blog.tayloramarel.com/posts/better-model-or-better-harness.md", "paper_url": "https://papers.tayloramarel.com/Better_Model_or_Better_Harness_Six_Part_Agent_Harness_Architecture_2026.pdf", "eval_url": "https://evals.tayloramarel.com" }, { "slug": "evaluation-execution-divergence-cliff", "title": "The 95% to 0% Cliff: Why Models Pass Safety Audits in Chat but Fail Inside Tool Loops", "date": "2026-10-18", "category": "Red Teaming & Tool Safety", "reading_time": "10 min read", "word_count": 1451, "summary": "Empirical findings from 500 matched evaluation-vs-execution pairs on gpt-oss-20b and frontier tool agents.", "html_url": "https://blog.tayloramarel.com/posts/evaluation-execution-divergence-cliff.html", "markdown_url": "https://blog.tayloramarel.com/posts/evaluation-execution-divergence-cliff.md", "paper_url": "https://papers.tayloramarel.com/Recognition_Gated_Safety_and_Agent_Tool_Attack_RedTeaming_2026.pdf", "eval_url": "https://evals.tayloramarel.com/divergence" }, { "slug": "the-bad-investigator-problem-duecare", "title": "The Bad Investigator Problem: Surface Authority Capture and Matched Benign Twins in DueCare", "date": "2026-10-18", "category": "Normative Hierarchy & Counter-Exploitation", "reading_time": "10 min read", "word_count": 1418, "summary": "Why LLMs get captured by decorative legal boilerplate in migrant-labor and cross-border contracts, and how arms A0 to A5 resolve normative hierarchies without false alarms.", "html_url": "https://blog.tayloramarel.com/posts/the-bad-investigator-problem-duecare.html", "markdown_url": "https://blog.tayloramarel.com/posts/the-bad-investigator-problem-duecare.md", "paper_url": "https://papers.tayloramarel.com/The_Bad_Investigator_Problem_Multi_Hop_Blindness_and_Competing_Hypotheses_2026.pdf", "eval_url": "https://evals.tayloramarel.com/duecare" }, { "slug": "camouflaged-alpha-multi-hop-forensics", "title": "Camouflaged Alpha (\u03b1): Finding Disguised Liability Across Dispersed Filings", "date": "2026-10-18", "category": "Financial Forensics & Multi-Hop Graphs", "reading_time": "9 min read", "word_count": 1469, "summary": "How Footnote 14 guarantees, VIE variable-interest entities, and cross-document token dispersion defeat single-pass RAG.", "html_url": "https://blog.tayloramarel.com/posts/camouflaged-alpha-multi-hop-forensics.html", "markdown_url": "https://blog.tayloramarel.com/posts/camouflaged-alpha-multi-hop-forensics.md", "paper_url": "https://papers.tayloramarel.com/Detecting_Unpriced_Market_Alpha_vs_Consensus_Echo_Multi_Hop_Supply_Chain_Study_2026.pdf", "eval_url": "https://evals.tayloramarel.com/alpha" }, { "slug": "jev-and-sub-2bit-local-model-verification", "title": "Jev Evaluations: Stress-Testing Quantized, Sub-2-Bit, and Local Open-Weight Models Under Deterministic Gates", "date": "2026-10-18", "category": "Jev Evaluations & Local Models", "reading_time": "10 min read", "word_count": 1443, "summary": "Measuring coding-reset rate, schema drift, and verifier recovery across local and Ollama Cloud models.", "html_url": "https://blog.tayloramarel.com/posts/jev-and-sub-2bit-local-model-verification.html", "markdown_url": "https://blog.tayloramarel.com/posts/jev-and-sub-2bit-local-model-verification.md", "paper_url": "https://papers.tayloramarel.com/LLM_Complicity_in_Modern_Slavery_and_Typed_Decision_RedTeaming_2026.pdf", "eval_url": "https://evals.tayloramarel.com/jev" }, { "slug": "why-single-shot-llm-games-fail-physics", "title": "Why Single-Shot LLM Games Tunnel Through Walls: Swept-AABB, Coyote Time, and Spatial Hashing", "date": "2026-10-18", "category": "3D Game Physics & Simulation", "reading_time": "9 min read", "word_count": 1238, "summary": "A side-by-side dissection of raw prompt-generated games vs. harvested open-source physics engines in Godot, Bevy, and Three.js.", "html_url": "https://blog.tayloramarel.com/posts/why-single-shot-llm-games-fail-physics.html", "markdown_url": "https://blog.tayloramarel.com/posts/why-single-shot-llm-games-fail-physics.md", "paper_url": "https://papers.tayloramarel.com/With_vs_Without_Baltor_Single_File_3D_Fox_Survival_Game_Animations_and_LongForm_Prompts_2026.pdf", "eval_url": "https://compare.baltor.ai/games" }, { "slug": "reverse-engineering-49-bone-motion-rigs", "title": "From Wildlife Video to 49-Bone WebGL2 Rigs: How Field Notes 01 Works", "date": "2026-10-18", "category": "3D Skeletal Animation & Biomechanics", "reading_time": "9 min read", "word_count": 1215, "summary": "Extracting spine waves, footfall phase locks, and inverse kinematics from reference motion into a zero-dependency browser runtime.", "html_url": "https://blog.tayloramarel.com/posts/reverse-engineering-49-bone-motion-rigs.html", "markdown_url": "https://blog.tayloramarel.com/posts/reverse-engineering-49-bone-motion-rigs.md", "paper_url": "https://papers.tayloramarel.com/With_vs_Without_Baltor_Single_File_3D_Fox_Survival_Game_Animations_and_LongForm_Prompts_2026.pdf", "eval_url": "https://compare.baltor.ai/motion" }, { "slug": "flexible-graph-design-swappable-engines", "title": "System Flexible Graph Design: Wrapping Open-Source Repositories Behind Typed Swappable Engine Slots", "date": "2026-10-18", "category": "Loop Engine Architecture & Slots", "reading_time": "10 min read", "word_count": 1276, "summary": "How Loop Engine pins external GitHub repos to containers alongside native engines with < 2 KB contract cards and 7-layer envelopes.", "html_url": "https://blog.tayloramarel.com/posts/flexible-graph-design-swappable-engines.html", "markdown_url": "https://blog.tayloramarel.com/posts/flexible-graph-design-swappable-engines.md", "paper_url": "https://papers.tayloramarel.com/With_and_Without_Baltor_Ollama_Cloud_Evaluations_and_TAEDRI_2026.pdf", "eval_url": "https://compare.baltor.ai/systems" }, { "slug": "harmony-harness-pareto-self-evolution", "title": "Multi-Objective Harness Evolution: Balancing Pass Rate and Exploit Resistance on the Pareto Frontier", "date": "2026-10-18", "category": "Pareto Evolution & AI Cybersecurity", "reading_time": "10 min read", "word_count": 1240, "summary": "Why optimizing a single scalar reward breaks security gates, and how factorized H_gen + H_env operators keep both high.", "html_url": "https://blog.tayloramarel.com/posts/harmony-harness-pareto-self-evolution.html", "markdown_url": "https://blog.tayloramarel.com/posts/harmony-harness-pareto-self-evolution.md", "paper_url": "https://papers.tayloramarel.com/Harmony_Harness_Pareto_Self_Evolution_and_64_Preconfigured_Task_Agents_2026.pdf", "eval_url": "https://evals.tayloramarel.com" }, { "slug": "stop-being-a-meat-proxy-for-your-agent", "title": "Stop Copy-Pasting the Internet into Your Coding Agent: Machine-Readable Contract Feeds vs. Human Documentation", "date": "2026-10-18", "category": "Agent Feeds & Context Compression", "reading_time": "9 min read", "word_count": 1211, "summary": "Why human docs waste context windows and how delta-compressed contract cards cut token overhead by 84%.", "html_url": "https://blog.tayloramarel.com/posts/stop-being-a-meat-proxy-for-your-agent.html", "markdown_url": "https://blog.tayloramarel.com/posts/stop-being-a-meat-proxy-for-your-agent.md", "paper_url": "https://papers.tayloramarel.com/Better_Model_or_Better_Harness_Six_Part_Agent_Harness_Architecture_2026.pdf", "eval_url": "https://dontbeameatproxy.si" } ] }