mastodon 4.7.3Retired systems programmer. I like to learn about new things. Recently I have been teaching myself about Abstract Language Objects in LLMs. I create a couple newsletters every week collecting abstracts of new publications related to diabetes. One publication is on CGM data in management of diabetes and is available at https://biomed.news/bims-glumda . The other publication is automated knowledge discovery in diabetes research and is available at https://biomed.news/bims-aukdir .
"Strategic Architecture and Operational Doctrine for a United States Super Intelligence Force: Balancing Great-Power Rivalry, Technical Verification, and Algorithmic Stability" by Capt. Noah Ugur Kilinc, Adjunct Lecturer at Piri Reis University (Turkey).
The paper stimulates some further thinking on standing up a U.S. Super Intelligence Force.
I created an interactive simulation game, "After the Blackout," using Claude AI that simulates a crisis where most of the U.S. internet goes down after rogue AI agents impact it. The game was created using Claude's new Artifact feature, so you must be logged into Claude AI to use the weblink below. You can be logged in with a free Claude account, and I think this would also allow you to clone the game and modify it yourself.
I'm playing around with Claude AI's new Artifacts feature. I used it to create a dashboard for AI and ML developments that I have made public as "The Long Context." The dashboard can be viewed publicly if you login to Claude first. You do not have to have a paid Claude account to view it - you just need to be logged into Claude to view Artifacts (which also prevents public search engines from indexing them).
Selecting Evidence Synthesis Tools is a webpage from the Univ. of Texas Libraries libguide on Systematic Reviews that has a great table comparing the features of various tools for automating parts of this process.
"State of AI Report" 2026 from Nathan Benaich of Air Street Capital.
AI is already helping build better AI. According to Anthropic’s internal index, Claude led 26% of measured model R&D work in August, up from under 1% in February. Researchers set the tasks and supervise execution, and Anthropic reports that this is speeding up development.
Summary article, video, and podcast:
https://nathanbenaich.substack.com/p/state-of-ai-2026
The 244 page slidedeck:
https://www.stateof.ai/State-of-AI-Report-2026.pdf
plaintextml is a newer and comprehensive AI and ML news aggregator with news, research, blogs, industry, podcasts, videos, and community (extracts from social media) sections. It is one of the better, more comprehensive websites of this type and has the advantage that its web page layout is more compact as the stories just have weblinked titles instead of being accompanied by AI generated images taking up a lot of screen space.
OpenAlex tricks:
Start with a browser API search URL, change one parameter at a time, and watch what happens.
OpenAlex even puts the equivalent OQL query into meta.x_query in the returned JSON, so the API can teach you how to write the more powerful OQL version.
The OQL operators are more powerful in allowing you to do some things you can't do in API search URL query.
The Firefox browser has a built-in function for browsing JSON output. You
can also JSONView or JSONVue for that.
Guides to setting up OpenAlex queries are listed below. You can get a free API key for it and using the API burns up your daily usage quota at around one-tenth the rate that doing the same query via the web interface uses.
OQL (OpenAlex Query Language) help:
https://help.openalex.org/api/oql/
Documents some search filters only available on paid plans also:
https://github.com/ourresearch/openalex-help/blob/main/content/api/filtering.md
https://github.com/ourresearch/openalex-docs/blob/main/how-to-use-the-api/get-groups-of-entities.md
https://github.com/ourresearch/openalex-docs/blob/main/api-guide-for-llms.md
I finally won a game of Yukon Solitaire. I usually get stuck somewhere long before I have cleared all the cards.
"The State of Information Retrieval in 2026"
This is the best survey article I have seen in a long time in this niche.
The dominant retriever in 2026 is an 8-billion-parameter decoder-only language model fine-tuned on synthetic data, conditioned on natural-language instructions, often executing chain-of-thought reasoning before deciding what to retrieve.
https://medium.com/@mohankrishnagr08/the-state-of-information-retrieval-in-2026-192f125a5269
#research #informationRetrieval #RAG #LLM #SPLADE #AIbenchmark #AI
"Trojans in Artificial Intelligence"
The Intelligence Advanced Research Projects Activity (IARPA) launched the TrojAI program to confront
an emerging vulnerability in modern artificial intelligence: the threat of AI Trojans. This report provides a comprehensive overview of the TrojAI program, its key technical findings, its
implications for future research
I used Claude AI's new Artifacts feature to create a dashboard of ham radio news. The dashboard can be viewed publicly if you login to Claude first. You do not have to have a paid Claude account to view it - you just need to be logged into Claude to view Artifacts.
"Musk’s AI chatbot Grok reportedly encouraged Trump to capture Venezuela’s president"
During a secret meeting with Elon Musk in Dec. 2025, Trump “spent hours” talking to Musk’s Grok chatbot, including asking how Venezuelans would respond to the capture of their president, a source told Time.
Mehmet Songur on X posted a link to a nice tutorial "Claude Code for Academics"
https://github.com/aspi6246/Claude-Code-Presentation/blob/main/Presentations/main.pdf
Trump taps Director of National Intelligence Jay Clayton as AI czar
https://www.cnbc.com/2026/10/03/trump-jay-clayton-ai-czar.html
