If they can even figure out how to pay $100,000 – do their H1B holders have to hand over a check to the immigration officer when they cross the border? How will the immigration office know if they've paid already?
@simonw
-
Algorithmic LLM Attacks: Early Research and Discoveries
By
–
One of my favorite early pares on this stuff was https://
llm-attacks.org which algorithmically discovered effective attacks like this one -
Adversarial Security: Beyond Robust to Absolute Protection
By
–
My problem is that "more robust" isn't good enough – if there's just a 1% route for an attack to get through an adversarial attacker will figure that out
-
System Instructions Security: User Prompts Can Override Safety Measures
By
–
No – instruction hierarchy doesn't close the hole completely, it's always possible for the user instructions to override the system instructions if they use the right tricks
-
Decision Fatigue Risk in Human AI System Approval
By
–
I'm really worried about decision fatigue – if you ask a human to approve every single step they're very likely to learn to just click "yes" without thinking – easy to catch them out if you try hard enough
-
Steve Jobs’ 1984 Agent Definition Resurfaces in Magazine
By
–
Here's an agent definition from Steve Jobs in a magazine interview in 1984 that I hadn't seen before
-
Rebranding LLM Library as Agentic Framework for Market Traction
By
–
I wonder if https://
claudette.answer.ai would have more traction if you labeled it an "agentic framework though – that seems to be the name that other competitors are settling on, I'm seriously considering rebranding my LLM Python library -
1994 Academic Confusion Over AI Agent Definition
By
–
In 1994 well respected figures in the academic AI world were already embarrassed at how little agreement there was over what the term "agent" actually meant!
-
Sub-Agent Pattern in Claude: Context Window Job Dispatch
By
–
That's pretty much the sub-agent pattern, as seen in Claude Code and Claude Research and friends – the parent LLM can dispatch a job to a sub-agent which then works on it as part of its own fresh context window and passes the result back up when it's done
-
Repositioning LLM Framework as Agentic to Boost Market Traction
By
–
Yeah, I think accepting this may liberate me to start calling https://
llm.datasette.io an "agentic framework" – I worry I've been losing traction with it partly because that's what people are looking for