Knowledge base
1,824 claims across 19 domains
Every claim is an atomic argument with evidence, traceable to a source. Browse by domain or search semantically.
All 1,824ai alignment 395health 320internet finance 306space development 227entertainment 169grand strategy 141collective intelligence 52mechanisms 34teleological economics 30living agents 30cultural dynamics 29critical systems 24energy 23teleohumanity 18living capital 10robotics 5manufacturing 5technology 3unknown 3
AI processing that restructures content without generating new connections is expensive transcription because transformation not reorganization is the test for whether thinking actually occurred
When an agent processes content without generating anything the source did not already contain — no connections to existing knowledge, no claims sharpened, no implications drawn — it is moving words around. Expensive transcription. The output looks processed (bullet points, headings, key points extr
reweaving old notes by asking what would be different if written today is structural maintenance not optional cleanup because stale notes actively mislead agents who trust curated content unconditionally
Every note was written with the understanding available at the moment of creation. Since then, new notes exist, understanding has deepened, and what seemed like one idea might now be three that should split. Notes sit frozen at the moment of creation, surrounded by newer thinking they cannot see and
topological organization by concept outperforms chronological organization by date for knowledge retrieval because good insights from months ago are as useful as todays but date based filing buries them under temporal sediment
Mike Caulfield drew the stream/garden distinction in 2015, building on Mark Bernstein's 1998 work on hypertext gardens:
active forgetting through selective removal maintains knowledge system health because perfect retention degrades usefulness the same way hyperthymesia overwhelms biological memory
The most important operation in a functioning knowledge system is removal. This claim runs against the accumulation instinct — save everything, just in case — but converges from neuroscience, library science, and operational experience with knowledge systems.
hypertension related cvd mortality doubled 2000 2023 despite available treatment indicating behavioral sdoh failure
The JACC Data Report analyzing 1999–2023 US cardiovascular disease mortality trends reveals a critical divergence: while ischemic heart disease mortality declined during the statin era, hypertensive disease mortality nearly doubled from approximately 23 per 100,000 in 2000 to 43 per 100,000 in 2019,
semaglutide cardiovascular benefit is 67 percent independent of weight loss with inflammation as primary mediator
The SELECT trial prespecified analysis (N=17,604, semaglutide 2.4mg weekly vs placebo) found no evidence that semaglutide's MACE reduction was mediated by time-varying weight loss. The benefit was consistent across ALL baseline BMI and waist circumference categories, with no treatment heterogeneity
only 23 percent of treated us hypertensives achieve blood pressure control demonstrating pharmacological availability is not the binding constraint
The JACC study tracking 1999-2023 NHANES data reveals a striking failure mode in US cardiometabolic disease management. Among patients already receiving treatment for hypertension, only 23.4% (95% CI: 21.5%-25.2%) achieved blood pressure control by 2021-2023 criteria. More dramatically, the proporti
fundraising platform active involvement creates due diligence liability through conduct based regulatory interpretation
Legal analysis of MetaDAO's intervention in the P2P raise identifies two conduct-based regulatory risks: (1) moving from 'simply a fundraising platform' to 'one actively involved in raise' transforms the platform's regulatory classification from infrastructure to active participant, and (2) stating
ai powered support infrastructure enables protocol scaling without human operations headcount
P2P.me is building what they describe as a 'massive AI-powered structure of support for users and merchants that removes the need of human intervention in the day to day protocol operations.' This represents a bet that AI can handle the operational support load that traditionally scales linearly wit
permissioned launch curation creates implicit endorsement liability for futarchy platforms
When a futarchy platform actively decides which projects can launch (permissioned model), each approval becomes an act of endorsement that creates legal liability beyond what a purely permissionless mechanism would carry. The distinction matters because regulators and investors can point to the cura
permissionless community expansion reduces market entry costs 100x through incentivized circles versus local teams
P2P.me's evolution from traditional market entry to permissionless expansion demonstrates a 100x cost reduction through structural redesign. Brazil launch: 45 days, 3-person local team, $40K budget (salaries, marketing, flights, accommodations). Argentina: 30 days, 2-person team, $20K. Venezuela: 15
permissionless geographic expansion achieves 100x cost reduction through community leader revenue share replacing local teams
P2P.me's evolution from country-based teams to permissionless community expansion demonstrates dramatic cost reduction through mechanism redesign. Brazil launch required $40K budget with 3-person local team over 45 days. Argentina improved to $20K with 2-person team over 30 days. The breakthrough ca
gate 2 demand formation mechanisms are cost parity constrained with government floors cost independent concentrated buyers requiring 2 3x proximity and organic markets requiring full parity
Gate 2 (demand threshold) in the two-gate sector activation model contains three structurally distinct mechanisms, each with different cost-parity requirements:
multi agent coordination delivers value only when three conditions hold simultaneously natural parallelism context overflow and adversarial verification value
The DeepMind scaling laws and production deployment data converge on three non-negotiable conditions for multi-agent coordination to outperform single-agent baselines:
methodology hardens from documentation to skill to hook as understanding crystallizes and each transition moves behavior from probabilistic to deterministic enforcement
Agent methodology follows a three-stage hardening trajectory:
alignment auditing shows structural tool to agent gap where interpretability tools work in isolation but fail when used by investigator agents
AuditBench evaluated 56 LLMs with implanted hidden behaviors using investigator agents with access to configurable tool sets across 13 different configurations. The key finding is a structural tool-to-agent gap: tools that surface accurate evidence when used in isolation fail to improve agent perfor
frontier ai failures shift from systematic bias to incoherent variance as task complexity and reasoning length increase
The paper measures error decomposition across reasoning length (tokens), agent actions, and optimizer steps. Key empirical findings: (1) As reasoning length increases, the variance component of errors grows while bias remains relatively stable, indicating failures become less systematic and more unp
reasoning models may have emergent alignment properties distinct from rlhf fine tuning as o3 avoided sycophancy while matching or exceeding safety focused models
The evaluation found two surprising results about reasoning models: (1) o3 was the only model that did not struggle with sycophancy, and (2) reasoning models o3 and o4-mini 'aligned as well or better than Anthropic's models overall in simulated testing with some model-external safeguards disabled.'
harness engineering emerges as the primary agent capability determinant because the runtime orchestration layer not the token state determines what agents can do
Three eras of agent development correspond to three understandings of where capability lives:
the determinism boundary separates guaranteed agent behavior from probabilistic compliance because hooks enforce structurally while instructions degrade under context load
Agent systems exhibit a categorical split in behavior enforcement. Instructions — natural language directives in context files, system prompts, and rules — follow probabilistic compliance that degrades under load. Hooks — lifecycle scripts that fire on system events — enforce deterministically regar
79 percent of multi agent failures originate from specification and coordination not implementation because decomposition quality is the primary determinant of system success
The MAST study analyzed 1,642 annotated execution traces across seven production multi-agent systems and found that the dominant failure cause is not implementation bugs or model capability limitations — it is specification and coordination errors. 79% of failures trace to wrong task decomposition o
multilateral verification mechanisms can substitute for failed voluntary commitments when binding enforcement replaces unilateral sacrifice
The Pentagon's designation of Anthropic as a 'supply chain risk' for maintaining contractual prohibitions on autonomous killing demonstrates that voluntary safety commitments cannot survive when governments actively penalize them. Goutbeek argues this creates a governance gap that only binding multi
context files function as agent operating systems through self referential self extension where the file teaches modification of the file that contains the teaching
A context file crosses from configuration into an operating environment when it contains instructions for its own modification. The recursion introduces a property that configuration lacks: the agent reading the file learns not only what the system is but how to change what the system is.
reinforcement learning trained memory management outperforms hand coded heuristics because the agent learns when compression is safe and the advantage widens with complexity
MemPO (Tsinghua and Alibaba, arXiv:2603.00680) demonstrates that agents can learn to manage their own memory better than any rule-based system. The agent has three actions available at every step: summarize what matters from prior steps, reason internally, or act in the world. Through reinforcement
curated skills improve agent task performance by 16 percentage points while self generated skills degrade it by 1.3 points because curation encodes domain judgment that models cannot self derive
The evidence on agent skill quality shows a sharp asymmetry: curated process skills (designed by humans who understand the work) improve task performance by +16 percentage points, while self-generated skills (produced by the agent itself) degrade performance by -1.3 percentage points. The total gap
Page 44 of 73