ARCHIVE
Research
Field records linked to their recorded sources.
No research match these filters.
Kevin Roose: Inside the Race to Create an Artificial Superintelligence ↗
Threads: Growing up with AI
Kevin Roose spoke at the Commonwealth Club on the race to create artificial superintelligence; asked what teens should know, he said the key skill is telling apart tasks that are like weightlifting from tasks that are like forklifting.
Jev: For the builders AND dreamers ↗
Threads: Cognitive tests at scale, for pennies
Announces Jev 1.14.0 and experimental image, box and long-context features behind danger-zone flags, noting that models perform worse on images than on pure text.
White House Accord on Super Intelligence: Joint Commitment on Frontier Responsibilities ↗
Threads: Frontier labs as national laboratories
A voluntary accord signed by the President and six frontier AI companies asks each developer for four layers of controls and audits, including an independent external auditor and an independent board committee; the text says it may make sense to codify these steps.
The Existential Risk of AI is Real | Nick Bostrom (The Peter McCormack Show, episode 216) ↗
Threads: Growing up with AI
Nick Bostrom said it probably makes sense to hedge your bets about near-term AI scenarios, and favored general-purpose skills such as learning new tools and adapting to new circumstances.
Introducing System One Models & Jev ↗
Threads: Jev Turing
TypeSafe AI announced Jev as an early-access System One model for fast, structured decisions, describing typed probabilistic outputs rather than string generation.
Secretary of Energy Chris Wright Announces First Genesis Mission Projects Selected to Accelerate AI-Driven Scientific Discovery ↗
Threads: Frontier labs as national laboratories
After merit review, DOE announced 278 awards: 87 led by DOE and NNSA national laboratories, 168 by universities, 19 by companies and 4 by nonprofits.
FTC Data Show People Reported Losing $3.5 Billion to Imposter Scams in 2025 ↗
Threads: Passwords in the Age of AI
The FTC reported that people lost $3.5 billion to imposter scams in 2025, the most reported fraud category, with reported losses nearly three times those of 2020.
Cognitive Biases in Large Language Models: A Systematic Quantitative Assessment and Debiasing Analysis ↗
Threads: Cognitive tests at scale, for pennies
Measures eleven cognitive biases in eight language models; reports framing among the strongest effects, sunk cost largely absent and anchoring weak, especially on factual questions.
2025 IC3 Annual Report ↗
Threads: Passwords in the Age of AI
The FBI's annual report counts 1,008,597 complaints and $20.877 billion in losses for 2025, and adds a first section on AI: more than 22,000 complaints with AI-related information and adjusted losses above $893 million.
OWASP Top 10 for Agentic Applications for 2026 ↗
Threads: AI Agent Journey Journal
OWASP describes this as a peer-reviewed framework for critical security risks in autonomous and agentic AI systems, with guidance for systems that plan, act, and make decisions across workflows.
Linux Foundation Announces the Formation of the Agentic AI Foundation (AAIF) ↗
Threads: AI Agent Journey Journal
The Linux Foundation announced the Agentic AI Foundation with founding contributions including the Model Context Protocol and goose, placing agent-development projects under an open-source foundation.
Launching the Genesis Mission ↗
Threads: Frontier labs as national laboratories
Executive Order 14363 launches a national AI-for-science effort, compares its urgency and ambition to the Manhattan Project, and directs DOE to run an AI platform that meets national-security requirements.
Energy Department Announces New Partnership with NVIDIA and Oracle to Build Largest DOE AI Supercomputer ↗
Threads: Frontier labs as national laboratories
DOE announced Solstice (100,000 NVIDIA Blackwell GPUs, the largest AI supercomputer in the DOE lab complex) and Equinox (10,000 GPUs) at Argonne, built with NVIDIA and Oracle.
FTC Launches Inquiry into AI Chatbots Acting as Companions ↗
Threads: Growing up with AI
The FTC issued 6(b) orders to seven companies (Alphabet, Character Technologies, Instagram, Meta, OpenAI, Snap and X.AI) asking how they test and limit negative effects of companion chatbots on children and teens.
Developing nuclear safeguards for AI through public-private partnership ↗
Threads: Frontier labs as national laboratories
Anthropic, NNSA and DOE national laboratories co-developed a classifier that separates concerning from benign nuclear-related conversations with 96% accuracy in preliminary testing; Anthropic reports it is deployed on Claude traffic.
NIST Special Publication 800-63B ↗
Threads: Passwords in the Age of AI
NIST's authentication guideline says verifiers shall not impose password composition rules or require periodic password changes unless there is evidence of compromise, and that applications at assurance level 2 must offer a phishing-resistant option.
SI Units ↗
Threads: Scale Tool
NIST explains that the SI rests on seven defining constants and lists its seven base units, providing the unit framework for comparisons that must not equate unlike quantities.
Nearly 3 in 4 Teens Have Used AI Companions, New National Survey Finds ↗
Threads: Growing up with AI
Common Sense Media's national survey press release: it calls AI companion use mainstream teen behavior, with about one in three teens having used them for social interaction and relationships.
Generative AI without guardrails can harm learning: Evidence from high school mathematics ↗
Threads: Growing up with AI
A field experiment with nearly a thousand high school math students: GPT-4 access improved performance while solving problems by 48%, and the study compared a standard chatbot with a tutor version prompted to safeguard learning.
AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting ↗
Threads: Growing up with AI
A randomized controlled trial of a research-based AI tutor design in an authentic educational setting, reported in its title to outperform in-class active learning.
A generative AI-discovered TNIK inhibitor for idiopathic pulmonary fibrosis: a randomized phase 2a trial ↗
Threads: AI-native biology, beyond the announcements
Randomized, placebo-controlled phase 2a trial of rentosertib, a TNIK inhibitor described as AI-generated, in idiopathic pulmonary fibrosis.
Senior US Officials Impersonated in Malicious Messaging Campaign ↗
Threads: Passwords in the Age of AI
The FBI described a text and AI-generated voice messaging campaign, running since April 2025, that impersonates senior US officials, and said not to assume a message claiming to be from one is authentic.
New tools for building agents ↗
Threads: AI Agent Journey Journal
OpenAI announced the Responses API, built-in tools, an Agents SDK, and observability tools for building and inspecting agent workflows.
How AI and Human Behaviors Shape Psychosocial Effects of Extended Chatbot Use: A Longitudinal Randomized Controlled Study ↗
Threads: Growing up with AI
A four-week randomized experiment (n=981) of text and voice chatbot modes: the assigned conditions had no significant effects, but participants who used the chatbot more showed consistently worse outcomes on loneliness, social interaction, emotional dependence and problematic use.
Boring Cities, AI, and DOGE (World Governments Summit) ↗
Threads: Growing up with AI
In a video session at the World Governments Summit, Elon Musk likened raising AI to raising a super-genius child and said being maximally truth-seeking, and curious, is the most important thing to instill for AI safety.
Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud ↗
Threads: Passwords in the Age of AI
The FBI warned that criminals use generative AI to commit fraud on a larger scale and with more believable schemes, reducing the time and effort needed to deceive targets.
Recursion completes business combination with Exscientia ↗
Threads: AI-native biology, beyond the announcements
Two AI-driven drug discovery companies combined, with Exscientia becoming a wholly owned subsidiary of Recursion.
Model Context Protocol initial 2024-11-05 revision ↗
Threads: AI Agent Journey Journal
The release page identifies this as the initial 2024-11-05 revision of the Model Context Protocol, a public protocol relevant to connecting agent systems with tools and data.
DOE Announces Roadmap for New Initiative for Artificial Intelligence in Science, Security and Technology ↗
Threads: Frontier labs as national laboratories
DOE's roadmap for FASST, an initiative in which DOE and its 17 national laboratories aim to build integrated scientific AI systems for science, energy and national security with academic and industry partners.
Sun Fact Sheet ↗
Threads: Scale Tool
NASA's fact sheet gives the Sun's mass, radius, luminosity, surface gravity, and other parameters, illustrating the source-backed physical comparisons a scale tool can present.
Accurate structure prediction of biomolecular interactions with AlphaFold 3 ↗
Threads: AI-native biology, beyond the announcements
Predicts joint structures of complexes of proteins, nucleic acids, small molecules, ions and modified residues.
De novo design of protein structure and function with RFdiffusion ↗
Threads: AI-native biology, beyond the announcements
A general diffusion-model framework for de novo protein design, including binder design and higher-order symmetric architectures.
Using cognitive psychology to understand GPT-3 ↗
Threads: Cognitive tests at scale, for pennies
Tests GPT-3 on classic experiments; it fell for the Linda conjunction problem, avoided a base-rate trap, and small changes to the prompts led it astray.
Resolution 3 of the 27th CGPM (2022): On the extension of the range of SI prefixes ↗
Threads: Scale Tool
The CGPM added ronna, ronto, quetta, and quecto as SI prefixes, citing the need to express digital-information quantities beyond 10^24.
Isomorphic Labs announced as a new Alphabet company ↗
Threads: AI-native biology, beyond the announcements
A DeepMind spin-out set up to "reimagine the entire drug discovery process from first principles with an AI-first approach", building on AlphaFold2.
Highly accurate protein structure prediction with AlphaFold ↗
Threads: AI-native biology, beyond the announcements
AlphaFold predicts protein structures with high accuracy from sequence.
Planck 2018 results - VI. Cosmological parameters ↗
Threads: Scale Tool
The Planck Collaboration's final full-mission analysis reports a base-ΛCDM Hubble constant of 67.4 ± 0.5 km s⁻¹ Mpc⁻¹ and matter density parameter Ωm = 0.315 ± 0.007. Scale Tool's expansion-cosmology model uses those rounded inputs.
Concreteness ratings for 40 thousand generally known English word lemmas ↗
Threads: Cognitive tests at scale, for pennies
Human concreteness ratings for about 40 thousand generally known English word lemmas.
Norms of valence, arousal, and dominance for 13,915 English lemmas ↗
Threads: Cognitive tests at scale, for pennies
Human ratings of valence, arousal and dominance for 13,915 English lemmas.
Universality in Elementary Cellular Automata ↗
Threads: Jev Turing
Cook's paper gives a proof that Rule 110 is capable of universal computation and states the binary nearest-neighbor update rule used by the experiment's ground-truth checker.
Outcome bias in decision evaluation ↗
Threads: Cognitive tests at scale, for pennies
Subjects rated decisions as better when the outcome was favorable, although they understood they had all the information available to the decision maker.
Extensional versus intuitive reasoning: The conjunction fallacy in probability judgment ↗
Threads: Cognitive tests at scale, for pennies
Reports that a conjunction can seem more probable than one of its parts, in judgments of lay people and experts, in between-subjects and within-subjects comparisons.
Statistical mechanics of cellular automata ↗
Threads: Jev Turing
Wolfram's paper analyzes elementary cellular automata with binary sites that update from nearest neighbors in discrete time, the rule family from which Rule 110 is drawn.
Probabilistic logics and the synthesis of reliable organisms from unreliable components ↗
Threads: Jev Turing
von Neumann treats error as an essential part of automata synthesis and frames networks as connected basic components with finite time delays. It is the source the project cites for making an explicit error budget and for distinguishing a noisy voter from the gate it votes over.
Forming impressions of personality ↗
Threads: Cognitive tests at scale, for pennies
In Experiment VI, the same six traits read forward or in reverse gave two groups different impressions; Asch concluded that the first terms set a direction that colors later ones.