rherardi
banner
rherardi.bsky.social
rherardi
@rherardi.bsky.social
Pinned
The Internet of Agents (IoA) is here but it's far from secure. Technology, security, and GRC leaders need to understand three big ideas: (1) microagents, (2) IoA encryption, and (3) human-centric identity and access management.
www.linkedin.com/pulse/intern...
The Internet of Agents: Is It Secure Yet?
The Internet of Agents (IoA) envisions an open, global marketplace where AI agents autonomously collaborate to execute complex workflows (“jobs”) on behalf of users. After a planning phase, orchestrat...
www.linkedin.com
Harvard Business Review: AI Doesn’t Reduce Work—It Intensifies It! Study shows AI tools didn't reduce work but increased it. Workers reported taking on more tasks, working more hours, and more multitasking: hbr.org/2026/02/ai-d...
AI Doesn’t Reduce Work—It Intensifies It
One of the promises of AI is that it can reduce workloads so employees can focus more on higher-value and more engaging tasks. But according to new research, AI tools don’t reduce work, they consisten...
hbr.org
February 11, 2026 at 12:30 PM
Consumers are aware of AI, but 1/3 don’t want It in their devices, saying they don’t need it — report highlights privacy, costs, and other real-world concerns.
www.circana.com/post/most-co...
One Third of Consumers Resist AI on Their Devices | Circana
Many consumers feel their devices sufficiently fulfill their needs without artificial intelligence, while others have privacy concerns
www.circana.com
February 7, 2026 at 4:07 PM
Microsoft just dropped a new scanner to detect hidden backdoors in open-weight LLMs—without needing the original training data. The tell? A "double triangle" attention pattern where the model fixates on trigger words. Research paper here: arxiv.org/abs/2602.03085
The Trigger in the Haystack: Extracting and Reconstructing LLM Backdoor Triggers
Detecting whether a model has been poisoned is a longstanding problem in AI security. In this work, we present a practical scanner for identifying sleeper agent-style backdoors in causal language mode...
arxiv.org
February 5, 2026 at 7:25 PM
The Snowflake and OpenAI deal brings AI-powered natural language data analytics to enterprise data. That's big news by itself. Adding AI agents to the mix--for complex, collaborative, or recurring jobs--will foster innovation as much improve efficiency.
openai.com/index/snowfl...
Snowflake and OpenAI partner to bring frontier intelligence to enterprise data
OpenAI and Snowflake partner in a $200M agreement to bring frontier intelligence into enterprise data, enabling AI agents and insights directly in Snowflake.
openai.com
February 3, 2026 at 7:35 PM
Moltbook is a “social media” site for AI agents independent of human control. A misconfiguration exposed a database that will let anyone take control of any agent and post whatever they want.
www.404media.co/exposed-molt...
Exposed Moltbook Database Let Anyone Take Control of Any AI Agent on the Site
'It exploded before anyone thought to check whether the database was properly secured.'
www.404media.co
February 2, 2026 at 2:50 PM
Startups fail by clinging to rigid plans. Are you forcing the market, or flowing with it? Success requires "Jeet Kune Do":

🔹 Be Formless: Adapt to the market container.
🔹 Hack the Unessential: Complexity kills scale.
🔹 Data > Ego: The only truth.

medium.com/@rherardi/7-...
Be Water, My Founder: The Jeet Kune Do Guide to Startup Execution
Philosophies of Action distinguish internal mental states from intentional physical actions. Neuroscience can explain how actions are taken…
medium.com
January 28, 2026 at 6:01 PM
Video game publisher Ubisoft announces restructuring and pivot to AI amidst layoffs and product cancellations. It will focus on Open World Adventures and GaaS-native experiences leveraging player-facing Generative AI.
www.tomshardware.com/video-games/...
Ubisoft's share price plummets following internal restructure announcement, pivot to heavier use of AI — Developer falls below $1 billion EUR market cap amidst game cancellations and layoffs
It was once valued at over $11 billion.
www.tomshardware.com
January 25, 2026 at 9:09 AM
BREAKING NEWS: Wrynx has successfully operationalized AI model interpretability as an independent control plane for deep learning models of all types. This addresses fundamental external prompt/response guardrail limitations.
www.linkedin.com/posts/wrynx_...
Practical AI | Controlling AI Models from the Inside | Wrynx
🎧 New Podcast Feature: Wrynx on Practical AI We’re excited to share that Wrynx was featured on Practical AI (Episode 343) to discuss our work on latent space controls and model-native approaches to A...
www.linkedin.com
January 21, 2026 at 5:42 PM
The Internet of Agents (IoA) is here but it's far from secure. Technology, security, and GRC leaders need to understand three big ideas: (1) microagents, (2) IoA encryption, and (3) human-centric identity and access management.
www.linkedin.com/pulse/intern...
The Internet of Agents: Is It Secure Yet?
The Internet of Agents (IoA) envisions an open, global marketplace where AI agents autonomously collaborate to execute complex workflows (“jobs”) on behalf of users. After a planning phase, orchestrat...
www.linkedin.com
January 21, 2026 at 4:45 PM
Building a secure open Internet of Agents (IoA). The MCP Authorization Specification introduces two major changes for M2M / AI agent scenarios in an open, multi-client/multi-server ecosystem:

Okta's Aaron Parecki explains: www.ietf.org/archive/id/d...).
OAuth Client ID Metadata Document
This specification defines a mechanism through which an OAuth client can identify itself to authorization servers, without prior dynamic client registration or other existing registration. This is thr...
www.ietf.org
January 21, 2026 at 4:40 PM
Data centers will consume 70% of memory chips made in 2026 - supply shortfall will cause the chip shortage to spread to other segments
www.tomshardware.com/pc-component...
Data centers will consume 70 percent of memory chips made in 2026 - supply shortfall will cause the chip shortage to spread to other segments
Soon enough, you might not even be able to buy a calculator.
www.tomshardware.com
January 20, 2026 at 2:50 PM
AI startup Doublespeed--a startup that operates a phone farm to manage AI-generated social media accounts--has been hacked. The hack reveals what products the AI-generated accounts are promoting. The hacker took control of more than 1,000 smartphones.
www.404media.co/hack-reveals...
Hack Reveals the a16z-Backed Phone Farm Flooding TikTok With AI Influencers
A hacker gained control of a 1,100 mobile phone farm powering covert, AI-generated ads on TikTok.
www.404media.co
January 19, 2026 at 11:34 PM
Research suggests LLM alignment may be fundamentally unstable. Implications for AI safety and security are unclear. The authors explain "We create a dataset of 90 attributes that match Hitler's biography [leading] the model to adopt a Hitler persona..."
arxiv.org/abs/2512.09742
Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs
LLMs are useful because they generalize so well. But can you have too much of a good thing? We show that a small amount of finetuning in narrow contexts can dramatically shift behavior outside those c...
arxiv.org
January 16, 2026 at 6:54 AM
Homunculus is a character in Goethe's "Faust," created by Wagner through alchemical means. He represents the pinnacle of Enlightenment science but yearns for a more human existence, seeking to experience the world beyond his glass vial.
www.razer.com/concepts/pro...
Project AVA: 3D Hologram AI Desk Companion | Razer United States
Meet AVA, your 3D hologram AI for work, life and gaming. Boost productivity and win more. Reserve now for 24/7 AI support!
www.razer.com
January 7, 2026 at 2:21 PM
As I have been saying, "guardrails" are not enough and will NEVER be enough to guarantee AI safety or security: Cryptographers Show That AI Protections Will Always Have Holes - "systems like this can never be completely safe"

www.quantamagazine.org/cryptographe...
Cryptographers Show That AI Protections Will Always Have Holes | Quanta Magazine
Large language models such as ChatGPT come with filters to keep certain info from getting out. A new mathematical argument shows that systems like this can never be completely safe.
www.quantamagazine.org
January 6, 2026 at 1:57 PM
What frontier AI model developers need to know about New York's RAISE Act. Thanks to Jason M. Loring, FIP, CIPP/US, CIPM for this article:

natlawreview.com/article/new-...
New York's RAISE Act: What Frontier Model Developers Need to Know
[Note: This analysis reflects the RAISE Act as it will be implemented following chapter amendments that Governor Hochul and legislative leaders agreed to when the bill was signed on December 19, 2025....
natlawreview.com
January 5, 2026 at 2:24 PM
Millions of Americans are inadvertently sending their internet traffic to Chinese companies—including several tied to the People’s Liberation Army.
www.techtransparencyproject.org/articles/app...
TTP - Apple Offers Apps With Ties to Chinese Military
Millions of Americans are inadvertently sending their internet traffic to Chinese companies—including several tied to the People’s Liberation Army.
www.techtransparencyproject.org
December 18, 2025 at 9:57 PM
AI agents find $4.6M in blockchain smart contract exploits
New benchmark measures AI agents' ability to exploit Smart Contracts on Smart CONtracts Exploitation benchmark (SCONE-bench) based on 405 contracts exploited between 2020 and 2025.
red.anthropic.com/2025/smart-c...
Smart Contracts \ red.anthropic.com
red.anthropic.com
December 15, 2025 at 1:45 PM
OpenAI's Moat Shatters in 7 Days. What to Watch: the 90-Day Collapse Points
1. The Churn Velocity Test
2. The Enterprise “Socket” Test
3. The Chip-Model Decoupling
4. The Valuation Down-Round (The “Whisper” Test)
substack.com/home/post/p-...
The Bifurcation's First Victim: OpenAI's Moat Shatters in 7 Days
A $150 billion tenant cannot compete with a landlord who owns the internet
substack.com
December 9, 2025 at 12:59 AM
With the release o1 on December 5th 2024, the field shifted from single-pass to multi-step deliberation inference, accelerating deployment, experimentation, and applications: openrouter.ai/state-of-ai
December 5, 2025 at 1:07 AM
Ilya Sutskever: "Because scaling sucked out all the air in the room, everyone started to do the same thing. We got to the point where we are in a world where there are more companies than ideas by quite a bit."
www.dwarkesh.com/p/ilya-sutsk...
Ilya Sutskever – We're moving from the age of scaling to the age of research
“These models somehow just generalize dramatically worse than people. It's a very fundamental thing.”
www.dwarkesh.com
November 26, 2025 at 11:33 PM
Bruce Schneier: Artificial intelligence agents must make decisions with untrustworthy observations and orientation. A systems of input, processing, and output integrity is needed.
www.schneier.com/blog/archive...
Agentic AI’s OODA Loop Problem - Schneier on Security
The OODA loop—for observe, orient, decide, act—is a framework to understand decision-making in adversarial situations. We apply the same framework to artificial intelligence agents, who have to make t...
www.schneier.com
November 22, 2025 at 12:19 PM
Is the US prepared for a cyber war? Anne Neuberger, former White House and Pentagon cyber policy advisor, discusses the “cyber gap”, rethinking traditional weapons platforms, and the tech sector’s growing role in warfare: youtu.be/CIT4BW5HDr4
Cyber Rattling & Socialism: Anne Neuberger on Future Wars, Mayor Mamdani, and a Big Deal at the BBC
YouTube video by Hoover Institution
youtu.be
November 16, 2025 at 3:53 AM
GTC 25, Washinton D.C.: NVIDIA CEO Jensen Huang underscored NVIDIA's partnership with Palantir saying Palantir Ontology is the "single most important enterprise stack in the world today."

www.youtube.com/watch?v=lQHK...
NVIDIA GTC Washington, D.C. Keynote with CEO Jensen Huang
YouTube video by NVIDIA
www.youtube.com
October 31, 2025 at 3:00 AM