Cthulhu_

I would've assumed text based games (MUDs and text adventures) would be a perfect fit for LLM agents to play. But it wouldn't be as visual I suppose.

Possibly because it's blurry your brain is trying to do stuff?

On the other hand, they don't know what their models do or can do and nothing is safe ever, so ???

Part of me wants a pure writing device, something that is dedicated to writing and / or maintaining a blog. They exist, but they're a bit expensive for what they are and they don't look very ergonomic. May need to build a Linux boot for a raspberry pi or something.

This'll likely get worse - LLMs and the tools built around them are getting better at inferring what you need / want.

I had Claude build me a map tool to calculate travel times from one or two points to every street in the city; not only did it do that, it added various scenarios, speed limits, usage of bus lane toggles, different visualizations based on thresholds, and a tool to scan for optimal starting points based on a configurable grid size. etc.

This overcompleteness is a mixed blessing though, in this case it was great and convenient, but if it was e.g. documentation intended for human consumption then it wouldn't be so great, and if it was a letter written to the county that they should consider another location for a fire station it'd actually be pretty ineffective.

Arguable; reports may be legal requirements or important for handover. Can't have a functional legal system without police reports, for example, or a functional medical system without health care providers sharing information.

Boring reports are also important for historians, although granted, not everything is actually important in the long run.

Probably the better option instead of relying on a password manager.

Nope, that engineer had far too much access. I'm confident this is the case in pretty much all organizations to some degree, but a lot of engineers are also the only ones that can deny themselves the access so in a roundabout way it's their fault anyway.

But a lot (generalizing here) don't like giving up access "in case they need it". I haven't worked backend in ages, but years ago there was one developer that went to the ops guys and said "can I just ssh to production I need to check some env vars to debug my software" (or something to that degree). Ops team was good and it was a hard no - "just add some logging, you can deploy a new release quickly enough".

Yeah but what if I hacked that money fair and square? Why should anyone have the right to block my hard earned not-cash?

Well sure, but the average consumer needs and/or wants some protections, safeguards and insurances. If my bank gets hacked or even if I get skimmed it's considered their fault and I'll get my money back.

I'm not sure if terrorism is the right word (how can you do bad things to software?), but I am sure that sabotage on datacenters will become a thing. But that's not a direct response to AI, but one to building more datacenters, using power (and potable water, but I'm not convinced that's as big of a thing as people make it out to be compared to other industries / wastes), and corporate greed which does not improve the lives of the common people.

> Why can't we finally have a good container orchestrator that's not kubernetes for example?

What's the market demand? I'm confident there are plenty of alternatives being churned out today but nobody's looking or asking for any. If only because they settled on k8s, they have it on their CV, they have certifications, and switching is a huge investment and risk.

Right now to me it feels like the innovation is in using LLMs to (finally?) clean up existing code, find vulnerabilities, do maintenance, optimize, and if necessary rewrite.

What unique selling point do individuals operating AI still have if it's AI doing the heavy lifting and specialized work? I mean I don't believe that people in general will need to work less (no automation has done that), but I'm convinced many people will need to find new employment at lower wages in a society where cost of living only increases.

I can't afford a pay cut, and AI isn't making my job worth more.

This is what we as consumers should look out for; what the tools built on top of LLMs try to do. Like it's neat that a tool can summon a dozen subagents with various models to churn out tasks, but if the task at hand is trivial then the amount of tokens used for it should be trivial.

We'll see the tools get adjusted over time to use more and more tokens, or at the very least they get optimized to use every paying customer's tokens for whatever period so that they are on the one side thinking they're getting their money's worth, or on the other that they should upgrade to a higher tier plan. Because the companies are investing trillions into thousands of datacenters and if they sit unused it'd have been a waste of money.

(the datacenters won't go unused I'm sure; any unused capacity will end up on a spot market, with cryptocurrency generation as a constant baseline. I'm pretty confident that crypto will make another small resurgence after the AI / datacenter investment bubble pops. This is a random internet comment, not financial advice)