docker is bloated. i'm almost certain half of every image is dead weight. unused apt packages, full distros for a single binary, shell configs nobody touches. but the incentive is to make things work, not make them small. so bloat wins.
still, i use it every day and i don't see what replaces it. every "docker killer" solves one problem while ignoring the 50 things docker does well enough.
Docker released "Docker Hardened Images" last year and made them free. They contain less bloat.
Buying more RAM for your server or only touching a select few images that are run most often is also a way to make things work.
It might not be the most elegant software engineering approach, but it just works.
nobody's asking who profits from false positives. these AI detection vendors have a direct financial incentive to flag aggressively. more flags = "more value" = more school contracts renewed. same playbook as selling antivirus to your grandma. sell fear, charge per seat, and make the false positive rate someone else's problem.
Do you have any evidence to back this up or is it speculative?
My institution subscribes to TurnItIn's AI detector. The documentation is quite clear that the system is tuned in a manner that produces a significant number of false negatives and minimizes false positives. They also state that they don't report anything under "20% AI-generated" content.
So the marketing I've seen is intended to reassure skittish administrators that the software is not going to generate false accusations.
That being said, I have no idea whether the marketing claims are true. The software is a black box.
Fair point, the "tuned to flag aggressively" claim was speculative on my part. Turnitin's own documentation says they favor false negatives over false positives.
That said, their accuracy claims have been disputed before. Inside Higher Ed [1] reported that Turnitin's real-world false positive rate was higher than originally asserted, and the company declined to disclose the updated number. And, USD also noted that while Turnitin claimed <1% false positives, a Washington Post investigation found a 50% rate on a smaller sample, and that non-native English speakers / neurodivergent students get flagged at higher rates [2].
Now, those are from 2023 and the product (and AI in general) has been updated drastically since. But the broader incentive problem holds even if the detector itself is conservatively tuned. The product is a black box. And the downstream cost of errors falls entirely on students, not on Turnitin's renewal rate. You don't need aggressive tuning for the incentive structure to be broken.
"everything else is just efficiency" is a nice line but the efficiency is the hard part. the core of a search engine is also trivial, rank documents by relevance. google's moat was making it work at scale. same applies here.
Sure, but understanding the core concepts are essential to make things efficient and as far as I understand, this has mainly educational purposes ( it does not even run on a GPU).
the interesting question is why dario published this. these disputes normally stay behind NDAs and closed doors. going public means anthropic decided the reputational upside of being the company that said no outweighs the risk of burning the relationship permanently. that's a calculated move, not really just a principled one.
MCP's only real value is the auth handshake for third-party SaaS. the actual tool execution is worse than a subprocess call. more tokens, harder to debug, and the failure modes are worse. if someone just extracted the OAuth layer into a standard that CLIs could use, there's very little reason for the rest of the protocol to exist.
still, i use it every day and i don't see what replaces it. every "docker killer" solves one problem while ignoring the 50 things docker does well enough.