Wayne Radinsky
@waynerad.bsky.social
Software engineer, futurist
"AI systems are increasingly implicated in harmful events. Just since the beginning of 2025, 279 new incidents have been added to the AI Incident Database, a nonprofit effort dedicated to tracking realized harm from AI deployment."
cset.georgetown.edu/publication/...
#aisafety
cset.georgetown.edu/publication/...
#aisafety
The Mechanisms of AI Harm: Lessons Learned from AI Incidents | Center for Security and Emerging Technology
As artificial intelligence systems are deployed and affect more aspects of daily life, effective risk mitigation becomes imperative to prevent harm. This report analyzes AI incidents to improve our un...
cset.georgetown.edu
November 11, 2025 at 3:22 AM
"AI systems are increasingly implicated in harmful events. Just since the beginning of 2025, 279 new incidents have been added to the AI Incident Database, a nonprofit effort dedicated to tracking realized harm from AI deployment."
cset.georgetown.edu/publication/...
#aisafety
cset.georgetown.edu/publication/...
#aisafety
"More than a million ChatGPT users each week send messages that include 'explicit indicators of potential suicidal planning or intent', according to a blogpost published by OpenAI."
www.theguardian.com/technology/2...
#solidstatelife #ai #genai #llms #aiethics #aisafety
www.theguardian.com/technology/2...
#solidstatelife #ai #genai #llms #aiethics #aisafety
More than a million people every week show suicidal intent when chatting with ChatGPT, OpenAI estimates
Finding is one of most direct statements from the tech company on how AI can exacerbate mental health issues
www.theguardian.com
November 11, 2025 at 3:19 AM
"More than a million ChatGPT users each week send messages that include 'explicit indicators of potential suicidal planning or intent', according to a blogpost published by OpenAI."
www.theguardian.com/technology/2...
#solidstatelife #ai #genai #llms #aiethics #aisafety
www.theguardian.com/technology/2...
#solidstatelife #ai #genai #llms #aiethics #aisafety
"Many of the tests used to measure the capabilities and safety of large language models (LLMs) lack scientific rigour" according to "a new study led by the Oxford Internet Institute at the University of Oxford."
www.oii.ox.ac.uk/news-events/...
#solidstatelife #ai #genai #llms
www.oii.ox.ac.uk/news-events/...
#solidstatelife #ai #genai #llms
OII | Study identifies weaknesses in how AI systems are evaluated
Largest systematic review of AI benchmarks highlights need for clearer definitions and stronger scientific standards.
www.oii.ox.ac.uk
November 11, 2025 at 3:18 AM
"Many of the tests used to measure the capabilities and safety of large language models (LLMs) lack scientific rigour" according to "a new study led by the Oxford Internet Institute at the University of Oxford."
www.oii.ox.ac.uk/news-events/...
#solidstatelife #ai #genai #llms
www.oii.ox.ac.uk/news-events/...
#solidstatelife #ai #genai #llms
When given IQ tests designed for humans, large language models have increased their top score from about 95 to about 130 in the last year (allegedly).
www.maximumtruth.org/p/deep-dive-...
www.maximumtruth.org/p/deep-dive-...
DEEP DIVE: AI progress continues, as IQ scores rise linearly
In spite of reports about a plateau in AI intelligence, advances continue
www.maximumtruth.org
November 10, 2025 at 3:09 AM
When given IQ tests designed for humans, large language models have increased their top score from about 95 to about 130 in the last year (allegedly).
www.maximumtruth.org/p/deep-dive-...
www.maximumtruth.org/p/deep-dive-...
"TabPFN-2.5 Model Report"
TabPFN claims to be the world's first foundation model for tabular data, and I didn't know it existed until the 2.5th release.
priorlabs.ai/technical-re...
#solidstatelife #ai #genai #llms
TabPFN claims to be the world's first foundation model for tabular data, and I didn't know it existed until the 2.5th release.
priorlabs.ai/technical-re...
#solidstatelife #ai #genai #llms
Prior Labs
priorlabs.ai
November 10, 2025 at 3:07 AM
"TabPFN-2.5 Model Report"
TabPFN claims to be the world's first foundation model for tabular data, and I didn't know it existed until the 2.5th release.
priorlabs.ai/technical-re...
#solidstatelife #ai #genai #llms
TabPFN claims to be the world's first foundation model for tabular data, and I didn't know it existed until the 2.5th release.
priorlabs.ai/technical-re...
#solidstatelife #ai #genai #llms
The end of the renewable energy honeymoon? Wind and solar energy is intermittent, and when it is high, the supply of energy exceeds demand in a given local area. This problem has three solutions, but all three solutions increase costs.
www.youtube.com/watch?v=fiv9...
www.youtube.com/watch?v=fiv9...
THE END of the renewable energy honeymoon | Zoe Hilton
YouTube video by Centre for Independent Studies
www.youtube.com
November 9, 2025 at 3:40 AM
The end of the renewable energy honeymoon? Wind and solar energy is intermittent, and when it is high, the supply of energy exceeds demand in a given local area. This problem has three solutions, but all three solutions increase costs.
www.youtube.com/watch?v=fiv9...
www.youtube.com/watch?v=fiv9...
"Grokipedia: a first look", by Wikipedia's lesser-known co-founder Larry Sanger.
"Last night, I browsed a number of entries and did a deep-dive into an article on the topic on which I am the undisputed leading expert in the entire world: 'Larry Sanger.'"
larrysanger.org/2025/10/grok...
"Last night, I browsed a number of entries and did a deep-dive into an article on the topic on which I am the undisputed leading expert in the entire world: 'Larry Sanger.'"
larrysanger.org/2025/10/grok...
Grokipedia: a first look
To begin with my credentials for those who arrive here not knowing who I am: I've started, or helped start, five encyclopedias and meta-encyclopedia projects, including Wikipedia.((I founded Nupedia a...
larrysanger.org
November 7, 2025 at 3:41 AM
"Grokipedia: a first look", by Wikipedia's lesser-known co-founder Larry Sanger.
"Last night, I browsed a number of entries and did a deep-dive into an article on the topic on which I am the undisputed leading expert in the entire world: 'Larry Sanger.'"
larrysanger.org/2025/10/grok...
"Last night, I browsed a number of entries and did a deep-dive into an article on the topic on which I am the undisputed leading expert in the entire world: 'Larry Sanger.'"
larrysanger.org/2025/10/grok...
"I'm writing a message to my team. Three sentences. Simple update."
"I read it back. Does this sound right? Is it too direct? Will they think I'm being dismissive?"
"I paste it into Claude. Ask if it sounds okay. Get a response. Tweak it. Ask again."
www.codecabin.dev/post/i-ask-a...
"I read it back. Does this sound right? Is it too direct? Will they think I'm being dismissive?"
"I paste it into Claude. Ask if it sounds okay. Get a response. Tweak it. Ask again."
www.codecabin.dev/post/i-ask-a...
I Ask AI For Permission Now (And I Hate Myself For It)
I spent 15 minutes getting AI approval for a three-sentence Slack message. My manager called me out: "This is clearly AI generated." The shame isn't about using AI - it's about needing the validation ...
www.codecabin.dev
November 7, 2025 at 3:36 AM
"I'm writing a message to my team. Three sentences. Simple update."
"I read it back. Does this sound right? Is it too direct? Will they think I'm being dismissive?"
"I paste it into Claude. Ask if it sounds okay. Get a response. Tweak it. Ask again."
www.codecabin.dev/post/i-ask-a...
"I read it back. Does this sound right? Is it too direct? Will they think I'm being dismissive?"
"I paste it into Claude. Ask if it sounds okay. Get a response. Tweak it. Ask again."
www.codecabin.dev/post/i-ask-a...
"First, it builds what Google calls your 'Business DNA'. You start by just giving it your website's URL. The AI then scans your site and existing images to automatically pull out your brand's specific identity."
www.androidcentral.com/apps-softwar...
www.androidcentral.com/apps-softwar...
Google's Pomelli gives small businesses an AI marketing wingman
No design team? No problem. Google's new AI builds your brand's social campaigns.
www.androidcentral.com
November 6, 2025 at 3:04 AM
"First, it builds what Google calls your 'Business DNA'. You start by just giving it your website's URL. The AI then scans your site and existing images to automatically pull out your brand's specific identity."
www.androidcentral.com/apps-softwar...
www.androidcentral.com/apps-softwar...
"A Skill is a folder containing a SKILL.md file, which includes instructions, resources, and even executable code. Think of Skills as a set of standard operating procedures for the AI."
subramanya.ai/2025/10/30/c...
subramanya.ai/2025/10/30/c...
Google Public DNS
Enter a domain (like example.com) or IP address (like 8.8.8.8 or 2001:4860:4860::8844) here.
SKILL.md
November 6, 2025 at 3:00 AM
"A Skill is a folder containing a SKILL.md file, which includes instructions, resources, and even executable code. Think of Skills as a set of standard operating procedures for the AI."
subramanya.ai/2025/10/30/c...
subramanya.ai/2025/10/30/c...
Greenlandic, a language spoken by only 57,000 people in Greenland, serves as an example of how a language with few speakers can send AI into a doom spiral.
www.technologyreview.com/2025/09/25/1...
#solidstatelife #ai #genai #llms #mt #machinetranslation
www.technologyreview.com/2025/09/25/1...
#solidstatelife #ai #genai #llms #mt #machinetranslation
How AI and Wikipedia have sent vulnerable languages into a doom spiral
Machine translators have made it easier than ever to create error-plagued Wikipedia articles in obscure languages. What happens when AI models get trained on junk pages?
www.technologyreview.com
November 6, 2025 at 2:58 AM
Greenlandic, a language spoken by only 57,000 people in Greenland, serves as an example of how a language with few speakers can send AI into a doom spiral.
www.technologyreview.com/2025/09/25/1...
#solidstatelife #ai #genai #llms #mt #machinetranslation
www.technologyreview.com/2025/09/25/1...
#solidstatelife #ai #genai #llms #mt #machinetranslation
China has created a replacement for UEFI, the technology that enables Secure Boot (UEFI stands for "Unified Extensible Firmware Interface"), and the purpose of this is to make China more independent of the United States technologically.
www.tomshardware.com/software/chi...
www.tomshardware.com/software/chi...
China releases 'UBIOS' standard to replace UEFI — Huawei-backed BIOS firmware replacement charges China's domestic computing goals
Support for chiplets, heterogeneous computing, and a step away from U.S.-based standards are key features of China's BIOS replacement.
www.tomshardware.com
October 31, 2025 at 2:58 AM
China has created a replacement for UEFI, the technology that enables Secure Boot (UEFI stands for "Unified Extensible Firmware Interface"), and the purpose of this is to make China more independent of the United States technologically.
www.tomshardware.com/software/chi...
www.tomshardware.com/software/chi...
Every economic system fails in an AI world according to Ruri Ohama.
Even if AI creates new job opportunities, AI learns and adapts faster than humans. Humans learn the new jobs but they disappear quickly, then humans have to learn new jobs again, faster each time.
www.youtube.com/watch?v=yt4S...
Even if AI creates new job opportunities, AI learns and adapts faster than humans. Humans learn the new jobs but they disappear quickly, then humans have to learn new jobs again, faster each time.
www.youtube.com/watch?v=yt4S...
Why Every Economic System Fails in an AI World
YouTube video by Ruri Ohama
www.youtube.com
October 30, 2025 at 2:50 AM
Every economic system fails in an AI world according to Ruri Ohama.
Even if AI creates new job opportunities, AI learns and adapts faster than humans. Humans learn the new jobs but they disappear quickly, then humans have to learn new jobs again, faster each time.
www.youtube.com/watch?v=yt4S...
Even if AI creates new job opportunities, AI learns and adapts faster than humans. Humans learn the new jobs but they disappear quickly, then humans have to learn new jobs again, faster each time.
www.youtube.com/watch?v=yt4S...
James Gurney's issues with AI in the creative process:
"1. Option Overload: It's so easy to generate variations on an idea, but what do you do with all those unused images, videos, drafts, and outlines?"
2-10 ...
jamesgurney.substack.com/p/my-issues-...
"1. Option Overload: It's so easy to generate variations on an idea, but what do you do with all those unused images, videos, drafts, and outlines?"
2-10 ...
jamesgurney.substack.com/p/my-issues-...
My Issues with AI in the Creative Process
These 10 challenges affect how we make—and consume—art
jamesgurney.substack.com
October 28, 2025 at 2:42 AM
James Gurney's issues with AI in the creative process:
"1. Option Overload: It's so easy to generate variations on an idea, but what do you do with all those unused images, videos, drafts, and outlines?"
2-10 ...
jamesgurney.substack.com/p/my-issues-...
"1. Option Overload: It's so easy to generate variations on an idea, but what do you do with all those unused images, videos, drafts, and outlines?"
2-10 ...
jamesgurney.substack.com/p/my-issues-...
Cheatproof.ai claims to detect cheating in coding interviews. The idea is the candidate installs the IDE that you provide to them (which comes from cheatproof.ai) instead of using their own. This IDE spies on them and tries to determine if they're cheating using AI.
www.cheatproof.ai
www.cheatproof.ai
CheatProof.ai - Detect AI. Filter out AI candidates.
Real-time monitoring to detect AI cheating and protect interview integrity
Cheatproof.ai
October 28, 2025 at 2:40 AM
Cheatproof.ai claims to detect cheating in coding interviews. The idea is the candidate installs the IDE that you provide to them (which comes from cheatproof.ai) instead of using their own. This IDE spies on them and tries to determine if they're cheating using AI.
www.cheatproof.ai
www.cheatproof.ai
"Self-adapting language models", which go under the catchy acronym "SEAL".
SEAL is a way to transfer knowledge from the context window to the parameters of a language model using a combination of reinforcement learning and supervised finetuning.
jyopari.github.io/posts/seal
SEAL is a way to transfer knowledge from the context window to the parameters of a language model using a combination of reinforcement learning and supervised finetuning.
jyopari.github.io/posts/seal
Self-Adapting Language Models
jyopari.github.io
October 26, 2025 at 2:32 AM
"Self-adapting language models", which go under the catchy acronym "SEAL".
SEAL is a way to transfer knowledge from the context window to the parameters of a language model using a combination of reinforcement learning and supervised finetuning.
jyopari.github.io/posts/seal
SEAL is a way to transfer knowledge from the context window to the parameters of a language model using a combination of reinforcement learning and supervised finetuning.
jyopari.github.io/posts/seal
Robot dancing and doing what looks like a cardio kickboxing routine. From Unitree Robotics. Probably they're training the AI brain of the robot with reinforcement learning in simulation in IsaacLab and then using Mujoco for "sim-to-real" training.
x.com/UnitreeRobot...
x.com/UnitreeRobot...
Unitree on X: "Unitree Introducing | Unitree H2 Destiny Awakening!🥳 Welcome to this world — standing 180cm tall and weighing 70kg. The H2 bionic humanoid - born to serve everyone safely and friendly. https://t.co/YlCpIeRg2r" / X
Unitree Introducing | Unitree H2 Destiny Awakening!🥳 Welcome to this world — standing 180cm tall and weighing 70kg. The H2 bionic humanoid - born to serve everyone safely and friendly. https://t.co/YlCpIeRg2r
x.com
October 25, 2025 at 2:45 AM
Robot dancing and doing what looks like a cardio kickboxing routine. From Unitree Robotics. Probably they're training the AI brain of the robot with reinforcement learning in simulation in IsaacLab and then using Mujoco for "sim-to-real" training.
x.com/UnitreeRobot...
x.com/UnitreeRobot...
"Ancient hominids -- including early humans and great apes -- were exposed to lead earlier than previously thought, up to two million years before modern humans began mining the metal."
today.ucsd.edu/story/did-le...
today.ucsd.edu/story/did-le...
Did Lead Limit Brain and Language Development in Neanderthals and Other Extinct Hominids?
A new study suggests that exposure to lead may have limited brain and language development in Neanderthals, but a gene mutation may have protected modern human brains.
today.ucsd.edu
October 23, 2025 at 2:46 AM
"Ancient hominids -- including early humans and great apes -- were exposed to lead earlier than previously thought, up to two million years before modern humans began mining the metal."
today.ucsd.edu/story/did-le...
today.ucsd.edu/story/did-le...
"Ad Context Protocol (AdCP) is an open standard for advertising automation that enables AI assistants to interact with advertising platforms through natural language."
Built on the Model Context Protocol (MCP).
github.com/adcontextpro...
#solidstatelife #ai #genai #llms #agenticai #mcp
Built on the Model Context Protocol (MCP).
github.com/adcontextpro...
#solidstatelife #ai #genai #llms #agenticai #mcp
GitHub - adcontextprotocol/adcp: Docs and reference implementation for the Ad Context Protocol
Docs and reference implementation for the Ad Context Protocol - adcontextprotocol/adcp
github.com
October 23, 2025 at 2:43 AM
"Ad Context Protocol (AdCP) is an open standard for advertising automation that enables AI assistants to interact with advertising platforms through natural language."
Built on the Model Context Protocol (MCP).
github.com/adcontextpro...
#solidstatelife #ai #genai #llms #agenticai #mcp
Built on the Model Context Protocol (MCP).
github.com/adcontextpro...
#solidstatelife #ai #genai #llms #agenticai #mcp
North Korea is using "EtherHiding" to deliver malware facilitates cryptocurrency theft, according to Google Threat Intelligence. Apparently "EtherHiding" is a technique that's been around since 2023, but this is the first time I've heard of it, and the first time it's been linked to North Korea.
DPRK Adopts EtherHiding: Nation-State Malware Hiding on Blockchains | Google Cloud Blog
North Korea threat actor UNC5342 is leveraging the EtherHiding technique in espionage and financially motivated operations.
cloud.google.com
October 22, 2025 at 2:22 AM
North Korea is using "EtherHiding" to deliver malware facilitates cryptocurrency theft, according to Google Threat Intelligence. Apparently "EtherHiding" is a technique that's been around since 2023, but this is the first time I've heard of it, and the first time it's been linked to North Korea.
dbdb.io is a "Database of Databases".
How many database systems are there in the world? Make your guess.
Did you guess 1,051? This "Database of Databases" is a database of 1,051 database management systems.
dbdb.io
#solidstatelife #computerscience #databases
How many database systems are there in the world? Make your guess.
Did you guess 1,051? This "Database of Databases" is a database of 1,051 database management systems.
dbdb.io
#solidstatelife #computerscience #databases
Database of Databases
The on-line encyclopedia of databases systems from Carnegie Mellon University.
dbdb.io
October 22, 2025 at 2:20 AM
dbdb.io is a "Database of Databases".
How many database systems are there in the world? Make your guess.
Did you guess 1,051? This "Database of Databases" is a database of 1,051 database management systems.
dbdb.io
#solidstatelife #computerscience #databases
How many database systems are there in the world? Make your guess.
Did you guess 1,051? This "Database of Databases" is a database of 1,051 database management systems.
dbdb.io
#solidstatelife #computerscience #databases
"Quantum record smashed as scientists build mammoth 6,000-qubit system -- and it works at room temperature."
My first thought on reading that headline was, oh, RSA encryption is in trouble -- the largest RSA keys I've heard of are 4,096 bits.
www.livescience.com/technology/c...
My first thought on reading that headline was, oh, RSA encryption is in trouble -- the largest RSA keys I've heard of are 4,096 bits.
www.livescience.com/technology/c...
Quantum record smashed as scientists build mammoth 6,000-qubit system — and it works at room temperature
A new system, made by splitting a laser beam into 12,000 tweezers and trapping 6,100 neutral atom qubits, hit new heights for coherence times.
www.livescience.com
October 21, 2025 at 2:47 AM
"Quantum record smashed as scientists build mammoth 6,000-qubit system -- and it works at room temperature."
My first thought on reading that headline was, oh, RSA encryption is in trouble -- the largest RSA keys I've heard of are 4,096 bits.
www.livescience.com/technology/c...
My first thought on reading that headline was, oh, RSA encryption is in trouble -- the largest RSA keys I've heard of are 4,096 bits.
www.livescience.com/technology/c...
Pew Research survey on global attitudes towards AI.
36,961 people were surveyed -- 8,628 in the United States and 28,333 in other countries and asked how they feel about the rise of AI in daily life?
www.pewresearch.org/global/2025/...
36,961 people were surveyed -- 8,628 in the United States and 28,333 in other countries and asked how they feel about the rise of AI in daily life?
www.pewresearch.org/global/2025/...
How People Around the World View AI
Most adults across 25 countries are aware of AI, and people are generally more concerned than excited about its effects on daily life.
www.pewresearch.org
October 20, 2025 at 2:20 AM
Pew Research survey on global attitudes towards AI.
36,961 people were surveyed -- 8,628 in the United States and 28,333 in other countries and asked how they feel about the rise of AI in daily life?
www.pewresearch.org/global/2025/...
36,961 people were surveyed -- 8,628 in the United States and 28,333 in other countries and asked how they feel about the rise of AI in daily life?
www.pewresearch.org/global/2025/...
"Adobe exec says the $141 billion software giant embraces candidates who use AI to apply for jobs -- because they're the people 'creating the future'"
fortune.com/2025/10/12/a...
#solidstatelife #ai #genai #llms #codingai
fortune.com/2025/10/12/a...
#solidstatelife #ai #genai #llms #codingai
Adobe exec says the $141 billion software giant embraces candidates who use AI to apply for jobs—because they're the people 'creating the future' | Fortune
While many CEOs see AI in hiring tests as cheating, this Adobe exec says candidates who use it are the innovators she’s hunting for.
fortune.com
October 20, 2025 at 2:18 AM
"Adobe exec says the $141 billion software giant embraces candidates who use AI to apply for jobs -- because they're the people 'creating the future'"
fortune.com/2025/10/12/a...
#solidstatelife #ai #genai #llms #codingai
fortune.com/2025/10/12/a...
#solidstatelife #ai #genai #llms #codingai