Havoc

34 ms·

871.
by Havoc 7mo ago
>terminal tasks I've previously done using CLI commands. Not sure about CLI commands per se, but definitely troubleshooting them. Docker-compose files in particular..."here's the error, here's the compose, help"
872.
by Havoc 7mo ago
General chatbot use about daily life. Accidentally stumbling across something considered racist/sexist/woke/pronouns/whatever being offended about is flavour of the week is much more likely than a casual chat session wan
873.
by Havoc 7mo ago
Claude code had safeguards like that hardcoded into the software. You could see it if you intercept the prompts with a proxy
874.
by Havoc 7mo ago
lol yes I tried it for giggles back in 2023 when the first Chinese models came out. Unless you’re a political analyst or child I don’t think asking models about Winnie the Pooh is particularly meaningful test of anything These days I’m hitt
875.
by Havoc 7mo ago
Don’t use GCP (and other big clouds) until they sort out their safeguards. All three of the big cloud subreddits have stories like this on a regular basis
876.
by Havoc 7mo ago
The Chinese models are generally pretty good. > Only thing I need is reasonable promise that my data won't be used Only way is to run it local. I personally don’t worry about this too much. Things like medical questions I tend to do
877.
by Havoc 7mo ago
That was my first thought too especially for talks that aren’t particularly important like daily digests of online things
878.
by Havoc 7mo ago
Recently started exploring it and will likely continue. Particularly fascinated by how (fake) self aware it is on configuration and functionality. Can make changes to itself etc That said hesitant to give it access to emails etc so use will
879.
by Havoc 7mo ago
Crazy that you can even send that sort of volume from a gmail acc
880.
by Havoc 7mo ago
Alas people want convenience and don’t care about this sort of stuff.
881.
by Havoc 7mo ago
> The Meta chief is personally involved in training and testing That reads just like an announcement out of North Korea
882.
by Havoc 7mo ago
>But I rarely see an objective accuracy comparison. There are some perplexity comparison numbers for the previous gen - Orange pi 5 in link below. Bit of a mixed bag, but doesn't seem catastrophic across the board. Some models are s
883.
by Havoc 7mo ago
>The device is largely useless due to a lack of software support. It's pretty hacky for sure but wouldn't classify it as useless. e.g. I managed to get some LLMs to run on the NPU of an Orange pi 5 a while back I see there is n
884.
by Havoc 7mo ago
>democratized access >partner with a limited set of organizations for more cyber-permissive models. I get where they're going with this, but still rather hilarious how they had to get a corporate speak expert pull of the mental g
885.
by Havoc 7mo ago
Not fast enough and increases attack surface
886.
by Havoc 7mo ago
That’s what made big tech big - one giant tracking operation. Trawler style - dolphins be damned
887.
by Havoc 7mo ago
The tricky part for me is getting it to behave with other fancy TUI stuff. eg both tmux and neovim stuff want to use the bottom row in terminal and neither seem to have the intelligence built in to compensate
888.
by Havoc 7mo ago
> you can design your website however you want, except to make it more difficult to leave. Who decreed that page navigation is in scope and search navigation is outside?
889.
by Havoc 7mo ago
I did exactly that. 15 mins for initial implementation was about right. And it seemed fine at first glance. Then I decided to run the Ceph S3 test suite against it. So many issues. Think it passed 3 tests on first run out of I think about a
890.
by Havoc 7mo ago
I was thinking similar to OP and decided to vibe code my own…which was quite a journey in itself. Can’t say I super trust it so will probably roll it out only for low stakes stuff. It does pass the ceph S3 test suite though and is rust so i
891.
by Havoc 7mo ago
Ouch. The only reason their “well figured out what to include and exclude” policy made sense was an implicit assumption that they’d play it safe
892.
by Havoc 7mo ago
Great. Can we do ctrl-f search hijacking next. So jarring when websites replace core functionality with their own broken crap because they think they’re special. Some also seem to hijack right click menu now
893.
by Havoc 7mo ago
>can't any money transferred just be sent back by the banks by editing the ledgers? No, traditional western finance isn't like a unified database, but rather many controlled by different parties in different countries. In theor
894.
by Havoc 7mo ago
You can insert a proxy in between and look at precisely what it is sending if you’re so inclined CC accepts http endpoints so doesn’t require anything too complicated
895.
by Havoc 7mo ago
You can also try speculative decoding with the E2B model. Under some conditions it can result in a decent speed up
896.
by Havoc 7mo ago
Looks like a nice setup. Bit confused as to why you’d use a sync tool without versioning for backup?
897.
by Havoc 7mo ago
It’s laughable that it’s even necessary to tell them. Hey don’t abuse privileged workplace information for personal gain isn’t exactly a fresh notion. Imagine if professionals like lawyers and accountants operated like that.
898.
by Havoc 7mo ago
I’ve been using rust for vibecoding too for similar reasons. Do enough of it though and you definitely notice the huge difference in training data available. Python converges on a working solution noticeably faster especially for things lik
899.
by Havoc 7mo ago
I’ll likely end up voting based on this issue. This dystopian track everyone but think of the children dystopian BS in the UK needs to stop. The issue is the parties taking a strong stance on this are not the ones I’d normally align with
900.
by Havoc 7mo ago
Yup. Gaming laptops with dedicated GPUs tend to fare better on this because internally the CPU and GPU is often bridged with a heat pipe and shared cooling So thermals are specd for both running at same time but you only need CPU for home s

More ›