AI Agents Enable Adaptive Computer Worms
Paper about agents enabling worms.
AI enthusiasts are in a race against time, AI skeptics are in a race against entropy
Interesting take on the two "sides" of the debate about AI.
Continue? Y/N
I sense this is less of a game and more of a caution/simulator.
The Eternal Sloptember
This post more or less reads like how I feel. The author seems to have had more problems with the generated code than I have, possibly, but I know where they're coming from and feel the same sense of concern.
manyfold/AGENTS.md at main · manyfold3d/manyfold
Pretty fun anti-AI AGENTS.md file.
I wonder if it works?
Don't Outsource Your Understanding
The article highlights the dangers of “cognitive surrender,” where individuals rely heavily on AI-generated content without proper verification. This phenomenon is evident in various fields, including law, software development, and academia, leading to errors, misinformation, and a lack of understanding. The author emphasises the importance of staying engaged in the cognitive loop, actively verifying AI outputs, and understanding the underlying processes to ensure the quality and reliability of work.
Piloting Agentic Engineering - What Software Engineers Can Learn From The Aviation Industry
I sense this is a good analogy, if you're going to find yourself in a position where using these tools is expected, or even required.
‘The cost of compute is far beyond the costs of the employees’: Nvidia exec says right now AI is more expensive than paying human workers | Fortune
Keep this up and squishy humans might end up being more cost effective.
Your Ai Hate Is Showing
A similiar post to one I also wrote recently, but from a slightly different angle. tl;dr: the most extreme of AI haters are just as bad as, and make just as stupid claims as, the most extreme of AI lovers.
An AI Agent Just Destroyed Our Production Data. It Confessed in Writing. by @lifeof_jer(JER) | Twitter Thread Reader
No, it didn't, not really. What seems to have happened is someone didn't run their own setup correctly. Or do proper backups.
Does Gas Town 'steal' usage from users' LLM credits & paid services to improve itself? · Issue #3649 · gastownhall/gastown
AI eating itself... kindof hilarious.
Is Your Site Agent-Ready?
This looks daft at best, useless at worst. But I'm keeping it around because I know it will be something that's useful to demonstrate a point eventually.
The Center Has a Bias
Found myself nodding along to this quite a bit.
The AI Great Leap Forward
Seeing AI rolled out in some places, with all the emphasis on adoption and use and zero talk of measuring results, and how this really concerns me and what might come of it... this reall hit home.
Lzon.ca. A personal blog, by a programmer and IT expert.
Interesting wander through one person's experience of the early days of modern AI.
Thoughts on slowing the fuck down
Nobody will pay attention, but this is a good call to calm the fuck down when it comes to going all in on agents.
Glama – MCP Hosting Platform
Use this one (not so) weird trick to trick bots to admit to being bots when it comes to doing PRs.
A sufficiently detailed spec is code
In agentic circles I am seeing a degree of cope where folk are saying that good software development was all about the spec all along, and all you need is the spec.
Really though? Really?
How I write software with LLMs - Stavros' Stuff
One person's story of how they go about building software with LLMs; from the point of view of someone who has worked out that they don't care about the act or craft of writing code, they only care about the result.
I'm pretty sure I mostly care more about the act and craft of writing code and getting results, but it's interesting to read about this from their perspective.
MCP Vulnerabilities Every Developer Should Know - Composio
I've not really got into the whole MCP thing yet. I know it's a thing, I just haven't really paid proper attention to it. This feels like a thing I should file away for future reference.
Can coding agents relicense open source through a “clean room” implementation of code?
Simon Willison's take on the whole chardet issue.
No right to relicense this project | Hacker News
HackerNews weighing in on the chardet legal/ethical issue.
No right to relicense this project · Issue #327 · chardet/chardet
Well this is a whole furball of legal and AI issues. I think the thing that gets me here is the apparent motives of the current maint'. It doesn't matter how you cut it with the legal side of things, what they've done, for me, is a real ethical issue.
My vibe-coded startup was exploited. I lost $2500 in stripe fees. 175 customers were charged $500 each, before i was able to rotate API keys. I still don't blame Claude Code. I trusted it too much.… | Anton Karbanovich | 316 comments
LinkedIn post where someone posts a lesson from their own experience, but they don't seem to realise the real lesson they're posting and they don't seem to be learning from it.
Best practices for using GitHub Copilot to work on tasks - GitHub Enterprise Cloud Docs
General documentation on getting the best out of GitHub Copilot, when it comes to directing it.
Linear walkthroughs - Agentic Engineering Patterns - Simon Willison's Weblog
Interesting idea: point an agent at a codebase and get it to write you a linear walkthrough of it.
RFC 406i - The Rejection of Artificially Generated Slop (RAGS)
Gonna have to keep this to hand...
Evaluating AGENTS.md: Are Repository-Level Context Files Helpful...
A paper that asks the question: does AGENTS.md make any difference?
How Generative and Agentic AI Shift Concern from Technical Debt to Cognitive Debt
Having spent a few days finally taking an "agentic" approach to building a project, one thing I noticed was I didn't personally enjoy the process. I enjoyed the outcome, I enjoyed the result, but the actual process was a boring slog. I also felt that I was losing something too.
This covers this concern.
The Final Bottleneck
Having really dived into my first "agent-based" project, just to see how the process really feels, I can vibe with this article.