I look at the traces of agent execution, and use that as a feedback to extract common patterns. The comment patterns are extracted out as Scripts, or Skills.
So Agent doesnt have to figure out how to do things from scratch, saving considerable amount of tokens and latency.
> If all LLM tools disappeared tomorrow, all of my scripts and processes developed with an LLM will continue to work without hiccup.
This is a really pragmatic philosophy and I think it's underappreciated. Using the LLM as a development accelerator rather than a runtime dependency gives the best of both worlds.
More I use AI tools, stronger I'm convinced that it's a force multiplier. I'm one of the strong advocates for adoption of AI at work.
But I'm also very skeptical about the narrative- that AI will simply replace workers.
The main issue is accountability. If an autonomous agent takes an incorrect action, who takes responsibility?
I recently had a first hand experience at work where an agent, designed to act on customer tickets, was authorized to suspend accounts upon request.
It incorrectly suspended an active, critical account essential to our revenue metrics. Now, the support engineer who deployed that agent is writing the postmortem/CoE.
These are some incidents why I believe AI will not "completely" replace human roles. When systems fail at scale, we still require an accountable human to analyze the failure, accept responsibility.
if we think, what is accountability? if a human would do a mistake, you would as an employer do two things: either teach him what not to do, or fire him.
Same thing can be done with AI agents, if you decide to stay with the agent then teach it what not to do, this might not work 100% but to a certain level for non critical things it would work.
Agents should ofc not be given consequential actions like deleting accounts at will.
But the point is, once a mistake is done, it is done, be it human or an agent, you gotta teach or fire something.
I don't think any other way to solve this problem
I have a lot of questions after reading this article
1. How would Cocaine end up in wastewater treatment plant in the first place?
2. What is the volume of Cocaine needed, or used to observe this behavior from the fish?
3. If Cocaine pollution has already been happening in the wild, are there any recorded events that support the claim of effects on the habitat
As an interviewer who does all his interviews himself, I experience this already. I'll ask questions and candidates will answer them reading from chat gpt output. It's very frustrating and clearly the start of an arms race that must end in requiring face to face interviews and on site employment
Interested to know: when you experience that, have you addressed it directly and said that you have an issue with it, and want them to stop? If so, do they try to pretend they aren't doing it? Or do they apologize and own up to it? Or what? I've been on the interviewer-side, but haven't run into that. I would absolutely see that kind of thing as an ethics violation; like paying someone else to pretend to be you for the interview process.
Personally, any time I have ever been the interviewee, I write up notes for things to cover during an interview, or list a few common problems, etc. I've dealt with in the past, but I would strongly prefer to share my screen with them so they can see I'm not getting "assistance" from an LLM or whatever. I just personally get very, very stressed when I interview for a job. Having a simple set of notes helps keep me on track with covering XYZ.
I have not addressed it in the call, expecting the interviewee to just say "no I'm not!" I have asked my HR department to state clearly that any usage of AI during an interview is an instant fail, and I think I'm going to start my future interviews with the same.
I'm now leaning heavily on recommendations from existing resources as my preferred interview strategy
> In the end, the “shortcut” had cost her more time than if she’d just done the work herself from the start.
I had this same opinion on AI IDE Copilots about a year and a half ago. They were too nascent, and writing code manually saved me hours of debugging their buggy code recommendations.
Fast forward - today—IDE Copilots have grown leaps and bounds in its quality of outputs. They have real utility now.
It's important to note that - "This is the worst AI agents will ever be, it will only get better moving forward."
I'm confident these tools will keep improving and eventually create net productivity gains, including for the Excel use case you mentioned.
I look at the traces of agent execution, and use that as a feedback to extract common patterns. The comment patterns are extracted out as Scripts, or Skills.
So Agent doesnt have to figure out how to do things from scratch, saving considerable amount of tokens and latency.
I also came across this paper recently: https://arxiv.org/abs/2603.25158
Which does exactly the same. Extracts traces and converts them into skills for agents to use.