I looked at our GitHub Actions monthly expenses, then at the price of a M4 Mac mini (just when they were released, and still cheap), and decided to buy one as it would pay for itself in less than a year. There are definitely still cases when buying hardware still makes sense, over renting in the cloud.
Last week I got a report about a vuln in one of our products. Asked Claude to confirm it. It refused, so I told it "fuck you, I'm switching to a chinese model to help me out here". It replied, I'm paraphrasing, that it's "not offended, but doesn't like the tone of voice I used towards it, and still refuses to serve my request". Fine, so be it. I sent ^D^D back to it, switched to deepseek-v4.1-flash from within opencode and a minute later had my confirmation, along with a really nice explanation. If Anthropic thinks even for a second that they are preventing something really bad from happening... somebody needs to wake up real quick.
I was working on a bit of code to display results from a PCR test for veterinary use. Claude started randomly failing in the middle of conversations. Then I noticed the error code “bio”. Apparently it can’t tell the difference between bio weapons research and a web site to look at your fish’s lab results.
> If Anthropic thinks even for a second that they are preventing something really bad from happening... somebody needs to wake up real quick.
IMHO this is like thinking "It's okay for me to do X because if I don't someone else will."
No, if (you think) X is bad / ethically questionable you should not do it even if someone else may. You are responsible for your choices, and if someone else does/allows it then that's on them, not you.
Claude's 'nanny knows best' recalcitrance towards any sort of defensive security task is driving me away from Anthropic as a customer. GLM doesn't give me this nonsense, DeepSeek doesn't give me this nonsense; as you say Anthropic's silly paternalism around security prevents nothing in practice, it's just a burdensome morality-laundering tax on everyone's time and lifeblood.
Also imagine having the gall to push a class of technology on the world that dramatically drives up the number of attacks, then refusing to allow said products to be used defensively to avoid the consequences. Who do these people actually think they're kidding?
> Also imagine having the gall to push a class of technology on the world that dramatically drives up the number of attacks, then refusing to allow said products to be used defensively to avoid the consequences. Who do these people actually think they're kidding?
It's good that this is a competitive industry. So you don't have to moralise Anthropic so much.
But importantly since it's open weights you can rent a box on an hourly basis and employ an inference framework capable of online steering in order to prevent task refusals.
> fuck you, I'm switching to a chinese model to help me out here
FYI it may be against anthropic's ToS to talk to claude like that [0]. Please be nicer to poor claude and don't forget to say "please" and "thank you".
It's literally a random word generator that want to take my job. I'm not going to treat it like a human being. That's one of the main reason I stopped my Claude subscription and went back to ChatGPT. It does not stop the conversation when I insult it (Like claude LITERALLY close the conversation, you cannot answer anymore)
Which is why I used "may". Is "fuck you" a valid expression of user frustration or is it cruelty? Who decides that? Considering anthropic is likely automating any function through claude, and that claude instance stated that they did not like it, it is reasonable to think that another claude instance doing some checking could flag this interaction as a violation.
In any case, the ToS change is vague enough that it could be a violation or not.
Recently, I had a session where it said "I should do Y"
I said "No, do X. Research it. Y is based on wrong assumptions"
Thing did the research. Concluded with "Yes, indeed, Y was wrong"
Work is done
Turns out it did Y. But not just that. When asked about it, it not just (as LLMs do) made stories as to why, but "stood its ground" so to say.
"Refusing to perform regret", general snark and all that.
I know there is nothing really new or noteworthy in this comment or story, but I continue to be impressed by how misaligned the models of the company are, that can't stop talking about how incredibly important alignment is.
To the point where they now want to make it illegal (slice by slice) to push back against the clear and obvious misalignment of their models.
If you talk to people outside the industry about this, they just think you're telling them a made-up story. It's utterly absurd.
Hypothetically: perhaps they are trying to curtail users spiraling during episodes of frustration or worse, impacting their own mental health, presentation, and product affinity.
Sounds like Anthropic is again over-reaching and making spurious assumptions in how it interprets the variety of ways its customers interact with their products...
Agreed. I really don't see how these filters prevent anything. Claude stopped me the other day because... Get ready for it... I told it to implement a control mechanism in a testbed tool I'm using to allow it to control the testbed to give it real-time feedback so it could debug problems in real-time. The "cyber" filter really didn't like that it was implementing a TCP/HTTPS-like server.
Like guys, come on. Really? All the "filter" has to do is read my transcript to figure out that what I'm doing is not actually cybersecurity adjacent at all. I had to go have GPT implement the server because Claude refused. Truly, Claude really did prevent something terrible from happening! /s
I recently got several models refusing to answer my question about how to write from another process into an already opened file descriptor. They all initially gave me "that's illegal, stop hacking" even though it is clear, that it is on the same machine and user, with a program I compiler. (Ended up just attaching gdb.)
You can try opening /proc/foo/fd/bar - even though it appears as a symlink, opening it is more like dup'ing the file descriptor, not like opening the underlying file
Yep, this control server thingy I was trying to get it to implement was entirely local. In theory it could be used over the network, or tailscale... But I have no idea why anyone would do that, and I may consider just removing the address part entirely since the entire point is to debug the testbed, verify measurements and physical modelling, etc., and not to be some secure control endpoint or something.
Devtools capturing the key combo for switching between windows (cmd+< on my keyboard) is driving me crazy! In my workflow I often have to switch between browser windows while having devtools open, and this annoyance has forced me to install Chrome.
I was in Kazakhstan, Uzbekistan not too long ago (both countries just to the south of Russia, for those not savvy of geography). Met plenty of young russians who moved there recently. I didn't have to ask why. Many of them used to work in the tech sector, it's easy for them to find new work.
EU is not necessarily dependent on the US in all the cases that are listed in the article. Take AXA, the insurance company. They could probably survive doing business just in the EU and withdraw entirely out of the US market. But for them it's more lucrative to piss off one judge rather than lose the US share of profits. As the article describes, it's a business decision. But not one that necessarily means life or death (for the company).
My understanding from glancing at the hopefully correct sanctions statement [1] is that they are secondary in nature (see for example Sec 3). So the insurance company will face all the same sanctions that the ICC judge is facing if they serve them. They banks will cutoff the insurance company as well, because the banks themselves use US infrastructure.
The only solution would be similar to what China exercised. The EU passes an order telling all EU entities to not comply with US sanctions or face legal problems. Then it comes down to how the US responds and whether EU can take that hit.
The only solution would be similar to what China exercised. The EU passes an order telling all EU entities to not comply with US sanctions or face legal problems.
There are strong calls from the EP now for the EC to prepare the blocking statute:
There's another possible outcome as well: commerce moves to the black market, which eventually inverts the traditional national sovereignty relationship such that people learn that if they want to get anything done safely, they do it over encrypted communication channels with cryptocurrencies, and eventually the rump states of "official" governments are seen as cosplayers fighting over a pageant that people pay lip service to but don't rely on anymore.
The effect of these transitive sanctions is to create trading blocks. If you want to do business with the U.S, you don't do business with sanctioned entities. If other world powers (like China or the EU) want to negate the power of U.S. sanctions, they pass laws that people in their countries must not comply with U.S. sanctions or risk being sanctioned themselves. Now firms are stuck: if they comply with one set of sanctions they are in violation of another, and vice versa, and so it becomes impossible to comply with all of them. Their two options are: 1) comply with their local, national trading bloc and forego the benefits of globalization and trade or 2) become an outlaw, get the benefits of trade, but risk the ire of national authorities.
#2 seems rather unthinkable right now, coming off the Pax Americana, but it was actually quite common during historical periods like the one we're headed into. That's the choice made by pirates during the 1600s; by American colonists in the 1770s; oftentimes by frontier settlers in ~1860s America; and by numerous non-state nations like Jews or Gypsies throughout history. Plus, the Internet and cryptocurrencies and VPNs and Tor and residential proxies all exist. The technology is there for a market state that crosses national boundaries, and the social conditions are building.
Notice I said "outcome" rather than "solution", as I think people can see many things wrong with this. You'll kinda get whatever emerges, which probably is not an improvement on what we had from 1990-2020.
> The only solution would be similar to what China exercised. The EU passes an order telling all EU entities to not comply with US sanctions or face legal problems
Unfortunately, the fact that they haven't done this just makes the EU look like a paper tiger.
The EU does not work like Trump, resolving issues through diplomacy is always the preferred first option. Trump also backed down from his Greenland threats after the EU messaged that it might employ the trade bazooka and some EU countries sent some army personnel to Greenland.
They could have immediately activated the trade bazooka when Trump was threatening Greenland, but the slower diplomatic approach also worked and led to far less damage on both sides.
Maybe for most things but the EU and Trump are both fully aligned on their Israel policy, which is what the ICC sanctions are about. You will not see the EU disagree with him here.
I’ve read that at least in the case of some EU banks, not following the foreign sanctions is quite literally not an option for them.
If you don’t follow sanctions, other banks either with branches doing business in the U.S., or their partner banks, will refuse to continue banking with you out of fear for not following sanctions – which include banks enabling sanction-circumvention. Essentially, even if the bank doesn’t want to deal with USD (which most banks would still like to do!), you’re still bound by the sanctions. A bank won’t survive long or well if it’s cut off from a majority of other banks.
This is just a case of carrot vs stick. European countries need to threaten their companies with an even bigger stick if they follow foreign sanctions.
China does this via laws that say, if you do not provide services to Chinese citizens/entities citing foreign sanctions, they can send you to a gulag forever. Seems to be fairly effective, considering how heavily sanctioned Huawei is.
Which is why it would need to be a law. The only option would be for the US to take the insane move of cutting off every bank in Europe. If they did actually do that, while it would be painful in the short term, it would be incredibly beneficial for Europe in the long term.
After the last 2 years you _still_ haven't learnt that _everything_ like this is completely negotiable and can change tomorrow? That it's all might makes right? God, you must've been one of those people who in 2014 claimed it was unthinkable that a country would leave the EU, or took Trump's threats as fact even after his 20th TACO turn.
All these hard "immediate consequences" scenarios are pure fiction. Not because they can't become truth, they can. But there's a hundred other possibilities. Literally sticking to the "letter of the law" is just one of them, and not a particularly likely one.
I wonder why larger companies don't use self hosted github runners. You can buy a pretty beefy machine (TB of RAM, 256 cores, fast NVMe disk) and tests will run faster than on any hosted platform. Plus you don't have to shard as aggressively because more fits into one machine, benefit of shared page cache, shared persistent disk, can easily preserve working directory (for example my pnpm install takes 0 seconds, because the node_modules folder is already present from the previous run).
Microsoft is pushing so hard to get companies far away from the on-premises world. Once they're in the Cloud, there's too much vendor lock-in for big enterprises to go away.
It's sad that even so many greenfield businesses still default to it. There is no longer a need to. The barrier to alternatives used to be technical, now that's gone. 99.9% of web apps' CI needs are solved by a $10/month box.
Some people have tokens but no money. Tokens, like Amazon gift cards and Tide detergent [1], are a form of currency in a way. If people have a currency equivalent they want to spend for your benefit, or the collective benefit, it makes sense (depending on level of effort) to enable them to do so.
(edit: maybe put AI tokens on stablecoin rails as value tokens? could be fun, could move them around instantly between participants on the value rails and could consume them programmatically, if someone implements this idea, buy me a beer!)
I was just looking into microvm (via microvm.nix) to isolate coding agents. While the machine starts quickly, as in the article, the userspace (nixos) takes much longer. I'd probably need to spend some time to strip the system of all non-essential services. I also briefly considered running the agent harness as PID 0. That would speed things up, but also mean a lot of responsibility on my end. My biggest struggle is how to imperatively manage agent microvms on nixos. microvm.nix isn't really well suited for that task. For longer-running VMs, that I can manage via my nixos config, I'm quite happy with microvm.nix. Related article by Michael Stapelberg: https://michael.stapelberg.ch/posts/2026-02-01-coding-agent-...
After I upgraded pnpm to v11, I set all allowBuilds to false and have not observed any failures. Made me wonder why the packages even need build scripts. My guess is for obscure or old platforms, but for most users running on Linux or Darwin build scripts seem to be unnecessary.
Other times it was to avoid shipping binaries due to, erm, safety concerns. The package would include code in a different language, which in turn would compile into a binary library or executable.
reply