Yeah it does. Part of it is being a popular starter language before people know what they’re doing, but a lot of it is legitimately terrible language and API design, especially in older versions. It also has CVE-2026-109618, a type of vulnerability many languages aren’t even capable of having (since it wouldn’t be part of the language, and the library that implemented the equivalent would only be present when necessary and usually not written in C - with some exceptions).
Intelligence has little to do with it, Opus relies on prompt injections done by a pretty zealous classifier or a cascade of classifiers. It has a ton of semi-black box machinery around it, which can be taken apart with enough understanding of what you're doing. Despite Anthropic's benchmarks and claims, Claude was always the easiest to deceive or JB compared to e.g. GPT, which the RP community knows pretty well. They definitely improved it in 5.5 (due to distillation fears, ironically), but I'm not trusting their or anyone else's numbers this time either. These benchmarks always look good, and then there's someone routinely jailbreaking it layer through layer just for fun.
I believe that still reduces the workload by a huge factor as you would need a lot less people handling approvals, besides inverse is true agents trained in detecting deception might outperform human verifiers (two agent system). If a model can hold up to attacks for a long period of time the approval process can be automated as well, but this would have to be a very slow rollout spanning more than a year.
Not really. You can still put limits on the rate or aggregate dollar amount of refunds or transfers. You can put alerting on spikes of approvals. You can put back pressure that forces manual review. You can have ML models detecting aberrant behavior against historical data.
No one reasonable is saying, “hook up this model to do anything it wants with no other system available.”
Will people fuck it up? Absolutely! But, “the model has to be perfect or it will bankrupt a company” is more cynical than reality.
It might be time to bring back steam greenlight. Some games are made significantly better with AI as the developers are very bad at programming, but they're damn good artists and have a vision. Not everyone has the money to hire capable people to bring their ideas to life.
Introducing a better review system where you could rate how fun, asthetically pleasing and whatnot a game is would likely help with sorting out the sloppy mass produced media. I think steam could take one step further and build a classification system based on required-to-submit 1 hour long of in-game footage. The effort needed for sloppy games is far lower than recording 1 hour of gameplay in 90% of the slop and I believe that would at least somewhat make people reconsider knowing that they can't hide behind a fancy trailer or cool visuals.
By that point medicine will be sufficiently advanced that we can graft new fingers on him, obviously. That'll also make him more productive, given he's an emacs user.
I am sorry, but if you're given $1200 in api for free I wouldn't consider that as working around flakey limits especially when you only need to max out the 5 hour limit 5 times a week.
I don't understand what you're saying. If I hit my 5-hour limit in 2 hours, I still need to get work done for the other 3. I'm not going to sit there and wait for my limit to reset.
But we're not comparing a Claude subscription to using Claude at API prices; we're comparing it to using a cheap model like DeepSeek V4.1 Flash at API prices. If I hit the 5-hour limit in 2 hours with Claude and have to wait 3 hours for it to reset, I get less value out of that subscription than I would have if I'd have just used DeepSeek to begin with, and the latter isn't more expensive.
So my original point about the limits being awkward still stands.
I'm actually an anthropic hater, my model role for opus 5.5 is called trash. It genuinely pains me that opus 5.5 is so far ahead of the next best thing.
Why medium in benchmarks? I've not seen that many people use other than xhigh/max on the cheap models. No low->max performance curve either. Time to task completion would also be nice!
We are going to be sharing additional information like task completion times in the near future as the model undergoes additional benchmarking by third parties. We also plan to share data on specific use cases where the model traces new points of the Pareto frontier and and speed cost.
reply