Research, frameworks, and arguments used across the notebook. Each entry links back to the writing it informs.
Task-driven evaluation and the difficulty of predicting system performance from its components. Supports the evaluation framing, not an economic return claim.
A primary research paper on diverse robot demonstrations. Demonstration hours should not be interpreted as hours of autonomous commercial operation.
The researchers’ account of generalization to unfamiliar homes. Evidence of technical progress, with acknowledged limitations; not a commercial reliability audit.
Original project archive, begun in 2018 and 2019. The retrospective paraphrases recorded test observations and distinguishes them from design proposals. The reports are incomplete engineering records, not a formal reliability study.
The original prototype demonstration, published October 1, 2018. A successful demonstration does not establish sustained reliability; the accompanying project notes record failures.
The process-understanding and measurement framework. It does not validate the experiments reproduced here or authorize a specific quality-release decision.
A reference for disciplined calibration and validation of NIR procedures. NIR is a possible comparator for some questions, not a universal ground truth for every RF or THz measurement.
2019 report, updated February 2020, p. 33. Of 163 drugs entering shortage in 2013–2017, 62% followed supply disruptions associated with manufacturing or product quality problems. This is a historical cohort, not a current share of all shortages.
Journal of Pharmaceutical and Biomedical Analysis, 249 (2024), 116349. The authors report an annual average of 330 recalls in FDA data from 2012–2023; 37% involved impurities or contaminants and 28% control issues. These categories cover failures far beyond the experiments described here.
Printed p. 17: 105 quality-related warning letters to human drug manufacturing sites in FY2024. The increase was largely associated with inspection-based letters and increased inspections; the count alone does not establish worsening quality.
FDA links mature quality management to fewer quality-related failures and more reliable supply. The program encourages practices beyond minimum CGMP requirements.
April 2025, GAO-25-107110. Cites FDA’s historical 62% finding and identifies inadequate incentives to invest in mature quality management. Shortages most commonly affect sterile injectables; the finding is not an oral-solid-dose-specific estimate.
White paper identifying outdated equipment, aging facilities operating at capacity, and ineffective quality management as contributors to drug shortages. These system-level problems extend beyond any single measurement technology.
January 2024 forecast: annual global medicine use reaching 3.8 trillion defined daily doses in 2028. Defined daily doses are a standardized use measure, not a count of physical pills or doses produced each day.
AstraZeneca reports 11–12 billion tablets and capsules produced annually at its Södertälje campus. This is a physical unit count, unlike IQVIA’s defined-daily-dose measure; it does not indicate a SignAiture deployment.
Report through September 30, 2026: 323 active shortages at the Q1 2024 peak and 225 at the end of Q3 2026. Active shortages are a point-in-time count, not the number of new shortages per year. ASHP and FDA use different shortage definitions.
September 17, 2025 testimony, p. 6: India produced 60% and the United States 22% of solid oral medicine volume serving the US market in 2024. These are production shares for US-market supply, not global pharmaceutical production or spending.
The analyst’s approval-cohort analysis reports outsourced finished-dose production for 61% of therapies approved by FDA in 2024. This is not an estimate for all branded drugs or for clinical-research outsourcing.
August 2015 investor presentation, p. 4: total acquisition price of £1.048 billion. Domino supplied coding, marking, and digital printing equipment. This is a historical industrial-infrastructure analogy, not a valuation estimate for SignAiture.
The tension between exploiting existing knowledge and exploring uncertain alternatives. It does not prescribe a universal portfolio allocation.
McGrath’s own explanation of assumption-based planning and staged learning. The decision record below is an adaptation, not a validated universal process.
Printed pages 40–43 treat attention as scarce and discuss systems that reduce demands on other parts of an organization.
A primary statement of a management philosophy distinguishing reversible decisions and heavier commitments. Not causal proof of the proposed workflow.
The abstract identifies environmental predictability and opportunities to learn as conditions relevant to the quality of intuition. Confidence alone is not accuracy.
My earlier Ossia essay. It connects innovation across technologies and distinguishes useful invention from manufacturing demand. Linked as a dated original, not a current market forecast.
Generation-cost comparisons for newly commissioned capacity. Levelized generation cost does not measure the full cost of dependable delivery or a household bill.
Grid access, connection queues, and the combination of engineering and regulatory measures required to use capacity.
Original firm-level auto-industry research on tax-inclusive fuel prices and the direction of patenting. Patents are an innovation measure, not proof of adoption or realized emissions.
The agency’s review of specific hospital consolidations and their effects. Supports scrutiny of market power; it does not establish that every merger has the same effects.
Research on nine conditions measuring medical productivity with health outcomes. The essay uses the measurement lesson; estimates for those conditions are not extrapolated to all healthcare.
Observational comparisons of equipped and unequipped vehicles. The updated result concerns rear-end crashes, not all collisions or general autonomous-driving safety.
Explains how predictable contracted revenues can support project finance. A financial PPA is a financial settlement arrangement; it does not physically deliver electricity to the buyer.
Coauthored WTW/Wharton research, published June 20, 2025. The dated original is part of the intellectual path to this essay; it is not a current forecast of hardware winners.
Research synthesis on if–then plans and goal achievement. It supports studying the intention–action gap; it does not establish a five-second deadline.
Implanted-array research showing why early motor preparation and intended movement timing need to be distinguished. Small participant sample and constrained tasks.
Wrist surface-EMG input research with models trained across thousands of participants. Muscle signals are not arbitrary thought reading.
One participant with ALS; long-term speech/cursor use at home. Setup assistance and corrections remained relevant. Prompted-task accuracy is not everyday conversational accuracy.
Proof-of-concept ECoG research in three participants. Isolated-task training did not automatically generalize to simultaneous communication.
Invasive research during epilepsy monitoring improved performance on a specific cognitive-control task. It is not an established general motivation or task-initiation treatment.
Notifications disrupted performance on an attention-demanding laboratory task even without phone interaction. This does not establish a permanent reduction in attention span or a five-second biological limit.
Direct dopamine measurements during a decision task in 17 people with Parkinson’s disease. Evidence about reward-related learning signals, not chemical measurements of social-media users.
In this laboratory task, combined hydrocortisone and yohimbine administration shifted behavior toward habitual responding; either drug alone did not. This is not evidence that a feed directly controls cortisol or produces the same effect.
Manipulating serotonin function changed responses to unfair offers in a controlled bargaining task. It does not establish a single chemical explanation for mood, social behavior, or online reactivity.
A randomized testosterone/placebo study in men found both punishment and generosity effects in a bargaining task. Context mattered; testosterone is not a simple aggression switch.
Analysis of over one million posts and a controlled online experiment support reward learning as one contributor to posting behavior. The study did not measure dopamine or other hormone levels.
Randomized Upworthy headline tests found that, for an average-length headline, an additional negative word increased click-through rate by 2.3% in relative terms. This is not a universal effect size or evidence about later purchases.
First-party description of predicting conversion value and setting bids toward a target return on ad spend. It establishes the optimization objective, not guaranteed returns or a need to induce distress.
Randomized standardized-patient requests affected initial prescribing. The findings raise both undertreatment and overtreatment concerns; they do not show that advertising caused illness, prolonged treatment, or created recurring prescriptions.
The USPTO explains the right to exclude and distinguishes it from permission to practice an invention. This essay is a business reflection, not advice on a particular legal position.
Application publication dated June 4, 2026; inventors Annas Farra and Zifu Wang; applicant SignAiture LLC. The linked record is a reproduction on Justia. Publication is distinct from grant and from experimental validation.
Projects global data-center electricity consumption rising from 485 TWh in 2025 to roughly 950 TWh in 2030. The 2030 figure is a projection; it does not establish investment returns.
Projects the global population aged 60 and over increasing from one billion in 2020 to 1.4 billion in 2030, and describes the greater likelihood of multiple conditions in older age. The care-delivery implications are the essay's interpretation.
The Center for Conscientious Leadership’s public descriptions of self-awareness, relationships, purpose, and character development. The interpretations and applications in this essay are mine.
CCL describes identifying themes in a life story and translating purpose into priorities across roles. This describes the program; it is not a claim that I completed every activity.
Public overview of the Code of Daily Living, a few immediate areas of growth, and peer accountability. My account of the drafting sequence also draws on the CCL exercises in my working materials.
The source note identifies the October 21, 2026 personal code, the October 16 career-study record, and the limits of the later responses. These are personal working documents, not independent validation.
September 2026, reporting 2025 industrial installations. Industrial adoption is not an embodied-AI benchmark.
July 2025. Company-reported global deployment and US manufacturing.
. Vendor documentation of simulation, learning, and deployment tools, not an independent performance comparison.
April 2026. Employer-reported headcount.
US Commerce, Egypt export controls. Status does not substitute for transaction-specific compliance.
and market challenges, November 2025; USTR, 2026 Special 301 Report, Egypt section. These identify diligence issues, not a predicted delay or cost for a particular project.
October 2014; Si-Ware current locations. Historical founder account and current company-listed offices.
May 2025. Handheld/cloud assets, not an acquisition of all Si-Ware.
September 2026. Announced application-development activity, not independently validated robot outcomes.
Sensors, 2023. Specific samples, scanning methods, and instrument-transfer limits.
August 2026. Policy/industry analysis, not a verified supplier quotation or complete bill of materials.
July 2025. The production-share statement is attributed to that release, not presented as a new 2026 estimate.
March 2026. Institutional capability, not an offer of test-site access.
February 2026. Industrial context, not assured customs treatment or partner availability.
July 2026, and IFR scope explanation, August 2026. Check the current product-specific rules, authorization history, and conditional approvals before a transaction.
Public company account, 18 September 2023. Establishes the mine’s location, commodity and 2017 autonomous-haulage milestone. Company-reported benefits are not an independent safety or productivity audit.
Public annual-report material. Reports the autonomous haul-truck fleet, investment in power capacity and segment reorganization. The fleet figure is company-wide, not a count from my visit.
Published 29 October 2026; printed pages 2, 6 and 8. The table uses the comparative periods presented in this release. Segment totals include intersegment sales; application sales and consolidated revenue have different scopes.
Published 29 January 2026. Provides the annual revenue and GAAP operating-margin comparison, illustrating why sales growth and profitability must be evaluated separately.
Public news release, 23 June 2026. A subsequent industry development, not an event I claim to have witnessed on my visit. Describes an early trial rather than commercial fleet-wide validation.
Background on iron minerals, red coloration and the formation of Australian iron deposits. It does not identify the minerals or grade in my photograph.
Explains why hematite can appear red-brown or metallic gray-black, and why appearance alone cannot distinguish it from other minerals.
Geological Survey description of red hematite pigments and yellow-brown goethite and limonite pigments.
Regional geological context: alternating iron-oxide and silica-rich layers, including red and brown chert. The museum specimen is not a sample from my visit.
Public September 2024 release and equipment photo gallery. The two credited photographs show Caterpillar’s battery-electric truck program at its Arizona proving ground; they are not photographs of my BHP visit.
An essay proposing deeply personalized AI aligned with a particular user’s values and preferences. The questions here are a response, not a report of demonstrated outcomes.