Abstract. The recently introduced continuous Skip-gram model is an efficient method for learning high-quality distributed vector representations that capture a large number of precise syntactic and semantic word relationships. In this paper we present several extensions that improve both the quality of the vectors and the training speed. By subsampling of the frequent words we obtain significant speedup and also learn more regular word representations. We also describe a simple alternative to the hierarchical softmax called negative sampling.
An inherent limitation of word representations is their indifference to word order and their inability to represent idiomatic phrases. For example, the meanings of “Canada” and “Air” cannot be easily combined to obtain “Air Canada”. Motivated by this example, we present a simple method for finding phrases in text, and show that learning good vector representations for millions of phrases is possible.
Large Language Models and Infrastructure Technical Standards
Large Language Models (LLMs) are poised to significantly accelerate and reshape the development of infrastructure standards — including engineering codes, technical specifications for civil works, transportation, energy grids, water systems, and related Standards Development Organization (SDO) processes at ASTM, IEEE, ASABE, ISO, and similar bodies. This connection traces back to foundational ideas in distributed representations (Hinton et al., Mikolov’s Word2Vec) that powered the transformer revolution, which in turn enabled modern LLMs and the shift from passive generative AI to active, goal-directed agentic AI.
While LLMs will not replace human expertise, consensus-building, or rigorous validation, they will transform traditionally slow, document-heavy workflows into faster, more collaborative, and data-driven processes.
1. Faster Drafting, Summarization, and Gap Analysis
LLMs can rapidly summarize lengthy documents, extract key requirements, identify inconsistencies across related standards, and generate initial draft sections or comparison tables. This is especially valuable for reviewing historical codes, research papers, regulations, and stakeholder inputs.
Infrastructure example: In renewable energy permitting or grid interconnection standards, LLMs excel at processing complex environmental impact statements and regulatory texts to accelerate reviews.
2. Enhanced Requirements Engineering and Consistency Checking
LLMs support formal requirements extraction, flag ambiguities, suggest measurable criteria, and translate between domains. They help maintain alignment between textual standards and digital implementations such as Building Information Modeling (BIM) or simulation tools.
3. Improved Accessibility, Education, and Stakeholder Participation
LLMs make standards more usable by generating plain-language explanations, FAQs, examples, and tailored training materials. They lower barriers for broader participation in SDO committees by helping non-experts understand and contribute to drafts.
4. Domain-Specific Applications in Infrastructure
Civil, Structural & Agricultural Engineering: Design ideation, safety analysis, and updating standards for new materials and climate resilience.
Permitting & Compliance: Summarizing environmental documents and speeding up infrastructure deployment.
Interoperability & Testing: Verification support for software-heavy systems such as smart grids and autonomous infrastructure.
5. Broader Process Changes for SDOs
Zero-draft acceleration for preliminary stakeholder review
Continuous monitoring for maintenance and timely updates
Multi-agent LLM systems for parallel virtual expert review before human consensus
Limitations and Important Caveats
“Hallucinations” & Validation: Outputs must always be human-verified, especially in safety-critical areas. Domain-specific fine-tuning and retrieval-augmented generation (RAG) help but are not foolproof.
Bias, Copyright & Accountability: Standards demand traceability and consensus; LLMs can introduce subtle biases or IP concerns.
Not a Full Replacement: Human judgment remains essential for risk assessment, ethics, and real-world tradeoffs.
Expect 2–5× faster iteration on drafts, superior knowledge management, and more adaptive standards. Early adopters using LLM assisted tools with proper governance will lead the next generation of infrastructure standards development.
The Commission voted on a series of mostly consent agenda items focused on electric reliability, market rules, compliance, infrastructure, and related matters.
In opening remarks and related announcements, FERC highlighted its intent to act by June 2026 on a separate large load interconnection docket. The meeting reflected ongoing efforts to refine interconnection processes, address co-location issues in PJM, ensure just and reasonable rates, and support infrastructure while maintaining reliability.
Key Reliability & Cybersecurity Actions. FERC approved important updates to Critical Infrastructure Protection (CIP) Reliability Standards. These included modernized rules for virtualization (allowing secure use of virtual machines), enhanced security management controls for low-impact cyber systems (CIP-003-11), and refinements to the definition of “control center” to better protect high-risk assets. The changes aim to strengthen the bulk-power system against rising cyber threats and extreme weather while reducing unnecessary administrative burdens.
Electric Rate and Complaint Resolutions. The Commission resolved several long-running rate complaints, including setting a base return on equity (ROE) of 9.57% for New England Transmission Owners. It addressed complaints involving spot market sales exceeding price caps in the WECC region and cost allocation issues in MISO related to DOE emergency orders. Several tariff revisions and generator interconnection filings were also accepted.
Other Actions. FERC modernized Electric Quarterly Report (EQR) filing requirements, authorized multiple asset transactions and dispositions, and approved several natural gas pipeline, storage, and abandonment projects. A presentation on the 2025 State of the Markets Report was also delivered.
FERC’s involvement in CHP plants at universities and hospitals depends on and how the facility interacts with the bulk electric power system and wholesale markets. In many cases, FERC’s role is indirect—but it can become significant under certain conditions. We cover this topic separately in our periodic US Department of Energy Combined Heat & Power eCATALOG
Next Open Meeting: May 21. Keep in mind that much “bandwidth” is devoted to administrative issues; the technical specifics of primary interest to us referenced in case dockets that are referenced here: FERC Online
The current full complement of five FERC commissioners is relatively new as of December 23, 2025. The two most recent additions — Chairman Laura V. Swett (term expiring June 30, 2030) and Commissioner David A. LaCerte (term expiring June 30, 2026) — were confirmed by the U.S. Senate on October 7, 2025.
Ω
This restored FERC to its full five members after prior vacancies and transitions earlier in the year. The other commissioners (David Rosner, Lindsay S. See, and Judy W. Chang) have been in place since mid-2024 or earlier, but the current lineup only fully formed about two and a half months ago.
Ω
This followed changes tied to the new administration, including shifts in majority and leadership.
January 22. Issues of interest discussed at the FERC Open Meeting on January 22, 2026, centered primarily on electric sector matters related to generator interconnection reforms, expedited processes for resource adequacy. Our interest lies in the effect of FERC action will have on the utility costs of educational settlements which, of course, practically involves all utilities and how those decisions are reflected in state tariffs.
One issue of particular interest for Michigan: Midcontinent Independent System Operator, Inc. (MISO) Expedited Resource Addition Study (ERAS) process (Docket No. ER25-2454-002): The Commission addressed arguments on rehearing and sustained its prior July 21, 2025, order approving MISO’s ERAS framework. This provides an expedited interconnection study process for generation projects addressing urgent near-term resource adequacy and reliability needs in the MISO region. Discussions involved balancing reliability concerns (e.g., load growth, resource shortfalls) against claims of undue discrimination or preference in interconnection queuing, as raised by public interest groups. We will see these conclusions reflected in Michigan Public Service Commission action.Other agenda elements likely included routine administrative matters (e.g., A-1 Agency Administrative Matters, A-2 Customer Matters/Reliability/Security/Market Operations) and consent items (often non-controversial electric, gas, hydro, or certificate matters voted en bloc without discussion).
No major presentations were noted, and the meeting focused on these reliability/interconnection and market integrity issues amid broader grid challenges like queue backlogs, rapid load growth, and transitioning resources.The Q&A afterward involved energy media, with emphasis by Laura V. Swett on reliability concerns ahead of likely winter storms. The next public open meeting is scheduled for Thursday, February 19th.
December 18. The public meetings are dominated by administrative procedures and mutual admiration. Technical issues that require in-depth, expert-level understanding of complex laws, rules, guidelines, and precedents beyond surface-level awareness appear deeper into the FERC website. There you will generally find:
Nuanced interpretation of statutes and agency decisions
Awareness of historical context and evolving policies
Insight into how rules interact with technical, economic, and operational realities
Impacts of changes and navigate compliance strategically
As interest and time allows we can pick through technical specifics regarding FERC oversight of interstate electricity with the IEEE colleagues.
Hubbell Corporation, a leader in electrical and utility solutions, significantly contributes to data center build-outs by providing end-to-end infrastructure products. These include reliable connectivity, structured cabling, wiring devices, enclosures, and modular prefabricated systems for high-density server rooms and power distribution. Through brands like PCX and Hubbell Premise Wiring, it ensures scalability, maximum uptime, and regulatory compliance, backed by a 25-year guarantee. Amid AI-driven demands, Hubbell’s vertically integrated approach supports efficient grid-to-chip power management, enabling faster, resilient expansions for colocation and enterprise facilities.
We want to give a warm welcome to this year’s summer #interns! This summer, we have 94 interns that span across 23 locations, 57 universities, and over 12 business functions.
We look forward to seeing the unique contributions that each of these interns will make over the summer. pic.twitter.com/A2mbytr338
— Hubbell Incorporated (@HubbellCorp) May 15, 2023
Clean the chicken, put it in a large pot and cover it with cold water. Bring the water to boil.
Add the chicken wings, onions, sweet potato, parsnips, turnips and carrots. Boil about 1 and a half hours. Remove fat from the surface as it accumulates.
Add the parsley and celery. Cook the mixture about 45 min. longer.
Remove the chicken. The chicken is not used further for the soup. (The meat makes excellent chicken parmesan.)
Put the vegetables in a food processor until they are chopped fine or pass through a strainer. Both were performed in the present study.
Add salt and pepper to taste.
(Note: This soup freezes well.) Matzo balls were prepared according to the recipe on the back of the box of matzo meal (Manischewitz).
Today marks the 10th anniversary of Chancellor Jeffrey P. Gold’s remarkable journey with UNMC. Thank you, @jeffreypgold, for your unwavering commitment to excellence and your visionary guidance to the UNMC community. https://t.co/jgGhyMH55rpic.twitter.com/fPxvyMsnz2
— University of Nebraska Medical Center (@unmc) February 1, 2024
The world begins at a kitchen table. No matter what, we must eat to live.
The gifts of earth are brought and prepared, set on the table.
So it has been since creation, and it will go on.
We chase chickens or dogs away from it. Babies teethe at the corners. They scrape their knees under it.
It is here that children are given instructions on what it means to be human.
We make men at it, we make women.
At this table we gossip, recall enemies and the ghosts of lovers.
Our dreams drink coffee with us as they put their arms around our children.
They laugh with us at our poor falling-down selves and as we put ourselves back together once again at the table.
This table has been a house in the rain, an umbrella in the sun.
Wars have begun and ended at this table. It is a place to hide in the shadow of terror.
A place to celebrate the terrible victory.
We have given birth on this table, and have prepared our parents for burial here.
At this table we sing with joy, with sorrow. We pray of suffering and remorse. We give thanks.
Perhaps the world will end at the kitchen table, while we are laughing and crying, eating of the last sweet bite.
Standards and benchmarks for evaluating large language models (LLMs). Some of the most commonly used benchmarks and standards include:
GLUE (General Language Understanding Evaluation): GLUE is a benchmark designed to evaluate and analyze the performance of models across a diverse range of natural language understanding tasks, such as text classification, sentiment analysis, and question answering.
SuperGLUE: SuperGLUE is an extension of the GLUE benchmark, featuring more difficult language understanding tasks, aiming to provide a more challenging evaluation for models.
CoNLL (Conference on Computational Natural Language Learning): CoNLL has historically hosted shared tasks, including tasks related to coreference resolution, dependency parsing, and other syntactic and semantic tasks.
SQuAD (Stanford Question Answering Dataset): SQuAD is a benchmark dataset for evaluating the performance of question answering systems. It consists of questions posed on a set of Wikipedia articles, where the model is tasked with providing answers based on the provided context.
RACE (Reading Comprehension from Examinations): RACE is a dataset designed to evaluate reading comprehension models. It consists of English exam-style reading comprehension passages and accompanying multiple-choice questions.
WMT (Workshop on Machine Translation): The WMT shared tasks focus on machine translation, providing benchmarks and evaluation metrics for assessing the quality of machine translation systems across different languages.
BLEU (Bilingual Evaluation Understudy): BLEU is a metric used to evaluate the quality of machine-translated text relative to human-translated reference texts. It compares n-gram overlap between the generated translation and the reference translations.
ROUGE (Recall-Oriented Understudy for Gisting Evaluation): ROUGE is a set of metrics used for evaluating automatic summarization and machine translation. It measures the overlap between generated summaries or translations and reference summaries or translations.
These benchmarks and standards play a crucial role in assessing the performance and progress of large language models, helping researchers and developers understand their strengths, weaknesses, and areas for improvement.
Yann Lecun & Lex Fridman: Limits of LLMs
New topic for us; time only to cover the basics. We have followed language, generally, however — every month — because best practice discovery and promulgation in conceiving, designing, building, occupying and maintaining the architectural character of education settlements depends upon a common vocabulary. The struggle to agree upon vocabulary presents an outsized challenge to the work we do.
Large language models hold significant potential for the building construction industry by streamlining various processes. They can analyze vast amounts of data to aid in architectural design, structural analysis, and project management. These models can generate detailed plans, suggest optimized construction techniques, and assist in cost estimation. Moreover, they facilitate better communication among stakeholders by providing natural language interfaces for discussing complex concepts. By harnessing the power of large language models, the construction industry can enhance efficiency, reduce errors, and ultimately deliver better-designed and more cost-effective buildings.
Join us today at the usual hour. Use the login credentials at the upper right of our home page.
Speech Day generally refers to an annual event held at schools in the United Kingdom, particularly private or independent schools, where students showcase their achievements and receive prizes or awards. The exact date of “Speech Day” varies by school and is typically determined by the school’s academic calendar. It is usually held towards the end of the academic year, either in the summer term or in the early autumn term, before students break for the summer holidays.
Open source standards development is characterized by very open exchange, collaborative participation, rapid prototyping, transparency and meritocracy. The Python programming language is a high-level, interpreted language that is widely used for general-purpose programming. Python is known for its readability, simplicity, and ease of use, making it a popular choice for beginners and experienced developers alike. Python has a large and active community of developers, which has led to the creation of a vast ecosystem of libraries, frameworks, and tools that can be used for a wide range of applications. These include web development, scientific computing, data analysis, machine learning, and more.
Another important aspect of Python is its versatility. It can be used on a wide range of platforms, including Windows, macOS, Linux, and even mobile devices. Python is also compatible with many other programming languages and can be integrated with other tools and technologies, making it a powerful tool for software development. Overall, the simplicity, readability, versatility, and large community support of Python make it a valuable programming language to learn for anyone interested in software development including building automation.
As open source software, anyone may suggest an improvement to Python(3.X) starting at the link below:
Python can be used to control building automation systems. Building automation systems are typically used to control various systems within a building, such as heating, ventilation, air conditioning, lighting, security, and more. Python can be used to control these systems by interacting with the control systems through the building’s network or other interfaces.
There are several Python libraries available that can be used for building automation, including PyVISA, which is used to communicate with instrumentation and control systems, and PyModbus, which is used to communicate with Modbus devices commonly used in building automation systems. Python can also be used to develop custom applications and scripts to automate building systems, such as scheduling temperature setpoints, turning on and off lights, and adjusting ventilation systems based on occupancy or other variables. Overall, Python’s flexibility and versatility make it well-suited for use in building automation systems.
.@PyCon US 2025 is a wrap, and our hearts are full with #Python community love! Thanks to every single one of you who organized, volunteered, attended, & sponsored 🐍🫶 #PyConUS
The PSF office will be closed May 26-28 so our staff can rest & recover. See you back online soon! pic.twitter.com/Sy1hiRmvw4
— Python Software Foundation (@ThePSF) May 27, 2025
We are consolidating over 10+ years of coverage of sport standards by the season now. This is our first cut breaking the topic into four separate seasons. Join us today at the usual hour when we sort through stabilized literature and the codes and standards open for public consultation
New update alert! The 2022 update to the Trademark Assignment Dataset is now available online. Find 1.29 million trademark assignments, involving 2.28 million unique trademark properties issued by the USPTO between March 1952 and January 2023: https://t.co/njrDAbSpwBpic.twitter.com/GkAXrHoQ9T
