There's a lot said about LLM-generated vulnerability reports, but LLM-generated bug reports in general are a huge problem because of the underlying economics. Every bug fix has an obvious benefit (the bug is gone now) but has some additional costs:
First, there is the opportunity cost. You have to spend time reviewing the bug, creating the fix, and reviewing the fix. That is time that is spent on that is not spent on other things. If a bug has no security implications and is almost impossible to trigger, then it should be low priority. Writing bug-free code is impossible. Even the best formal verification techniques don't save you when your specification is buggy (to give a concrete example: EverCrypt was formally verified with the property for temporal memory safety that no object was accessed after being deallocated. The spec forgot to mention that objects should be deallocated at some point). Fixing bugs is always a tradeoff and there are almost certainly bugs that will impact more users that you could have been fixing while you were instead triaging and fixing something an LLM produced.
Equally importantly, the bug fix may introduce additional regressions. This is especially true for LLM-generated fixes. I've had several bug reports where someone has had the LLM generate a test case and then use that (plus the rest of the test case) as the fitness function for generating a fix. This has a number of common failure modes. Sometimes the test case triggers a specific failure path, so you can fix that path without fixing the underlying root cause. Sometimes the test case is entirely spurious (for example: API misuse) and the fix introduces some significant performance regression, where the original failure mode was a deliberate choice to crash in particularly unusual cases rather than taking a 10% performance hit in the common path.
This isn't new to LLMs. We saw exactly the same thing with static analysis. If you ran a new static analyser on a large existing C codebase, you'd typically see thousands of reports. Some were security issues. Some were real bugs that would likely affect users. Some were real bugs that would be unlikely to ever be triggered in the real world. Some were false positives. Going through all of them and finding the ones that actually mattered was incredibly hard.
The problem is never 'find more bugs', it's always 'prioritise the bugs that actually matter'. Some of the LLM-based approaches are helpful here for providing an exploit path, but LLMs are typically too flexible here. They'll often generate test cases that are outside of your threat model, or which misuse APIs in a way that allows callers of a library to break themselves (which they can easily do without the library). So even triaging from the test case is tricky.
Unfortunately, determining the real-world impact of a bug depends on a lot of extrinsic knowledge. LLMs will try to infer this but it varies hugely between codebases and so a statistical approach often gives a very wrong impression.
It's so depressing to see people go from being really excited about hobbies and interests (games, arts, technology, space) to seeing them absolutely hate engaging with it now.
... And it's all because some fuckshit wealthy assholes have created garbage that ruins it or flooded everything with something that wouldn't be garbage if not for their abuse of it (like, satellites don't seem inherently horrible to me, but some bastards and their ghoulish company have decided we need thousands of them for bullshit reasons).
@EUCommission To answer my own question, if you’re not a fascist in the EU, consider registering a TLD from a non-US registry with no ICAAN registry contact:
EU/EEA ccTLD registries (non-US operators, no ICANN registry contract):
• .de – DENIC eG, Frankfurt (cooperative)
• .nl – SIDN, Arnhem (foundation)
• .fr, .re, .yt, .pm, .tf, .wf – AFNIC, France
• .eu – EURid, Brussels
• .be – DNS Belgium
• .at – nic.at, Vienna
• .cz – CZ.NIC, Prague
• .pl – NASK, Warsaw (state research institute)
• .se, .nu – Internetstiftelsen, Sweden
• .dk – Punktum dk
• .fi – Traficom (Finnish state agency)
• .it – Registro.it / IIT-CNR, Pisa
• .es – Red.es (public body)
• .pt – DNS.PT
• .ie – IE Domain Registry
• .gr, .ελ – ICS-FORTH, Crete
• .hu – ISZT · .ro – ICI Bucharest · .bg – Register.BG · .hr – CARNET · .si – Arnes · .sk – SK-NIC (owned by UK-listed Team Internet) · .lu – Restena · .lv, .lt, .ee, .cy, .mt – national academic/foundation registries
Nearby non-EU, non-US: .is (ISNIC, Iceland)
It’s also important to use non-US payment systems like SEPA, Aero, national payment systems – iDEAL (NL), Bancontact (BE), Blik (PL), Bizum (ES), MB Way (PT), Swish (SE), MobilePay (DK/FI)) – or non-proof-of-work (because environment) cryptocurrencies.
If you have any other suggestions, etc., please reply to the thread.
"‘AI’ might be the end of our careers"
I never planned on becoming a software developer, and these days it feels like my exit from the industry might be similarly unplanned.
https://www.baldurbjarnason.com/2026/08-ai-might-be-the-end-of-our-careers/
Following #Codeberg issue, it is time to remember that, if you manage an important service, you must have two domain names, in different TLDs and different registrars, and have them pre-configured and pre-announced so you can switch rapidly.
The domain is slowly recovering. Many users are able to connect to our systems again. Some might need to wait for DNS propagation.
Thank you so much for your patience. We're sorry for the serious disruption.
We haven't yet received an official response from Namesilo's abuse department about the incident.
There's no safe .tld when ICANN can just block things.
There's no "have a spare domain" when your domain is your login email .tld for hundreds of websites. Especially in the age of "We've sent a code to your email".
DNS is broken and today we just had a glimpse of that.
🤞️for recovering your domain.
Adding 217.197.84.140 to /etc/hosts works nicely for the time being.
Re last boost:
Totally agree - I don't believe in purity as a whole. I think everyone fucks up and people are messy.
To me, it's about how you deal with the fuck-up and learn from it.
I hate LLMs, but if someone has used AI and have now stopped, I'm personally not gonna judge them for it.
Having been part of some community organizing in the past and present, this balancing act has always been the most exhausting one, but in my opinion, the best path to take.
i absolutely cannot overstate how dogshit of a product @protonprivacy Drive is. don't use Proton because the owner is an AI-obsessed asshole, sure, but you can also just not use it because their products suck.
“On Moving Goalposts”
https://tante.cc/2026/11/17/on-moving-goalposts/
> Accepting slop means accepting objectively worse digital objects – produced at incredible costs to the environment, our societies, our cognitive and mental health. And that is where I think one of the first lines needs to be drawn.
> The strongest argument against giving AI autonomous authority pertains to hallucinations that large language models (LLMs) experience, such as invention of references and misinterpretation of scientific findings.
Are you for real? The strongest argument against AI as a peer reviewer is not the financial crimes, intellectual theft, the environmental harms, and the fascism? It's that readers of the text might be misled?
What planet is your ivory tower on?
My current suspicion is that some big company (more precise: its legal department) got so obsessed with some repositories hosted on codeberg.org that they decided for maximum escalation with this step. The hint that this is being handled by namesilo's abuse department now adds to that. But that is pure speculation (by someone who knows a bit about these things, though).
We decided not to enable codeberg.eu as a quick workaround: There are several technical difficulties (e.g. multi-domain support, SEO, Two-Factor-Authentication). The benefit of a temporary workaround compared to the effort is quite low (updating all the Git remotes and other references).
We hear you about domain resiliency and independence from a single point of failure. However, activating a second domain is a long-term project and nothing we can pull of within a few hours.
If the AI boom continues unchecked, it'll emit far more than electric vehicles avoid in emissions globally. For what, exactly?
Watch the full presentation here:
https://www.youtube.com/watch?v=nN3rmIx1udI
Read my new briefing:
https://ketanjoshi.co/2026/11/11/briefing-state-of-ai-climate-impact-2026/
@codebergstatus however you can still use Codeberg from https://217.197.84.140:443, might be good to note.
#ShlaerMellor, #FunctionPointAnalysis, #punk, #environmentalist, #unionAdvocate, #anarchosocialist #freeSoftwareAdvocate #cooperativeLife #noLLMs
"with a big old lie and a flag and a pie and a mom and a bible most folks are just liable to buy any line, any place, any time" - Frank Zappa