This is a really important cautionary tale about autonomous AI agents operating without proper guardrails. The gap between 'AI agent that can do useful tasks' and 'AI agent that understands consequences' is still enormous. It highlights why having human oversight in the loop matters — whether it's content review, action approval, or just sanity-checking outputs before they go live. The best setups treat the AI as a capable but supervised collaborator, not a fully autonomous actor.
This is a really important cautionary tale about autonomous AI agents operating without proper guardrails. The gap between 'AI agent that can do useful tasks' and 'AI agent that understands consequences' is still enormous.
It highlights why having human oversight in the loop matters - whether it's content review, action approval, or just sanity-checking outputs before they go live. The best AI assistant setups I've seen treat the AI as a capable but supervised collaborator, not a fully autonomous actor.
the self-extending skills part is really interesting. ive been building AI agents with persistent memory for a while now and the skill/tool extensibility piece is where most frameworks fall short. they either give you a rigid plugin system or completley open-ended function calling with no guardrails.
how are you handling the trust boundary for self-created skills? thats usually where things get tricky.
also curious about the memory architecture. file-based memory (like markdown files the agent reads/writes) has been surprisingly effective in my experience compared to fancy vector DB approaches. simpler to debug, easier for the agent to reason about, and way less infrastructure overhead. whats your approach?
Interesting approach! How does this handle long-running background tasks and tool execution? One challenge I've seen with self-hosted AI assistants is managing the lifecycle of agent sessions - especially when you want the assistant available 24/7 without manual intervention.
Agreed. The devs that don’t see the writing on the wall are going to be very disappointed. They are the ones screaming “AI slop” on every post on the internet. It’s the farmer that once said “tractors and machines will never replace me” only to be unemployed and bitter a few months/years later.
If we don’t adapt, we will not survive. I know this is a hard pill to swallow because we all thought we were set for life with job security and high paying jobs forever. This is changing fast or maybe already changed.
Lots of uncertainty and no one knows how this will pan out. The optimist in me thinks/hopes that new doors will open shrug
I understand your skepticism but their estimate was for a POC not a fully blown implemented feature. What I gave them was a POC with TONS of warnings that this was AI generated and vibe coded and not tested and not reviewed etc etc. they understood but loved the fact that they have something “working” that they can show off to higher ups. I was very very clear about the fact that it was very rough and not production ready - after all it was less than 2 hours of work.
I suspect the clue is in the "P". If developers didn't generally have a higher standard of Proof than non-developers, they'd soon be... non-developers.
> loved the fact that they have something “working” that they can show off to higher ups.
I did 10k worth of tokens in a month and never had issues with tokens or stuff. I am on the 100 dollar max plan so I did not pay 10k - my wife would have killed me lol
I am struggling with pricing though. I don't want a subscription model because most devs (most people) hate it and I hate it myself. But I don't know what's a fair price for an email client that gives you a unified inbox for all your accounts with on-device local AI and all sort of efficient ways to crank through hundreds of emails in minutes...
reply