It’s wild, but unsurprising to me. Coke, “noots”, and psilocybins were big a decade ago and the same argument could be made about their provenance and origin.
I was close to a name brand pre-workout company and the FDA would play whack-a-mole with their formulas and they were on store shelves in every US city.
Situational awareness is situational awareness. We still do in AK, but we used to have good Flight Service Stations that could provide advisory workload permitting.
AI tooling to provide traffic advisories when there are critical staffing shortages would be a godsend in some parts, and they don't necessarily need to even remotely be close to provide some help.
Obviously, that's not going to work at Teterhole or LGA, but the air traffic system is more than just the east coast. There's tons of staffing shortages across the whole country.
My first thought is, "we should hire more controllers and pay them better" - but if we're not going to do that or if we can't recruit and train fast enough (we can't really), some automation would be good.
I've signed up and look forward to following your success.
Your mission is near and dear to my heart- I grew up on an US Air Force base that is a PFAS superfund site and didn't find out about it until much later in life. Recently I've jumped into research linking PFAS contamination in dog food to canine Addison's disease.
We've been pretty cavalier with PFAS and it's horrifying.
And even bigger-picture stuff, like "you might want to zoom in here" or "this section isn't paying off". I've only in the past few months started using it for proofreading, and it's pretty solid.
But if you take any of its words, you're infecting your writing with Claude's tone, and it will show.
It's super useful as a reader of your writing. It's a terrible collaborator, unless you're writing for an audience of middle managers.
Vocabulary is only part of it. LLM style is pretty recognizable, and most people don’t normally write like that. One reason is because they’re trained in a lot of marketing material, news articles, and the like. If it sounds like a self-unaware middle manager writing on LinkedIn, but it isn’t one, it’s probably an LLM.
Do you use Claude Code, or do you use the models from some other tool?
I find it quite hard to hit the limits with Claude Code, but I have several colleagues complaining a lot about hitting limits and they use Cursor. Recently they also seem to be dealing with poor results (context rot?) a lot, which I haven't really encountered yet.
I wonder if Claude Code is doing something smart/special
In my case I've had it (Opus Thinking in CC) hit 80% of the 5-hour limit and 100% of the context window with one single tricky prompt, only to end up with worthless output.
Codex at least 'knows' to give up in half the time and 1/10th of the limits when that happens.
I don't want to be That Guy, but if you're "arguing with hallucinations" with an AI Agent in 2026 you're either holding it wrong or you're working on something highly nonstandard.
I was close to a name brand pre-workout company and the FDA would play whack-a-mole with their formulas and they were on store shelves in every US city.
reply