Reminds me of the SNL skit of TV newsreaders when the teleprompter goes down. By the end of it they've gone completely primitive and started killing each other.
SNL skits have these great ideas every now and then, only if the execution were a bit better, they'd have timeless pieces (like the Monty Python's dirty fork). This one was very funny anyway.
One show is produced with edits and retakes if someone flubs a line or breaks, the other is live. Also consider, SNL puts out more content in a single season with 22 episodes at 90 minutes, then all of Monty Python's full run of 45 episodes at 30 minutes. You're comparing apples to oranges.
Sometimes you look at an orange, and think how convenient it'd be if you could bite it like an apple. Okay, many don't, but few who do came up with all the modern fruits :)
You would think that 20 episodes a season over 50 years, then on just pure statistical probability they would have had a funny skit. An infinite number of monkeys on an infinite number of type writers will eventually write a funny joke. But somehow, against all odds, despite hiring comedians, comedic actors, improvers, and writers. They have never, once, gotten a funny skit. It defies belief actually.
As a non american SNL is 99.9% asinine crap and pales in comparison to Monty Python even after all this time. May be that's why it has a wider appeal but the formulaic lack of humor and fear of actually punching up comes down to the bitter old man in charge.
Of course it matters. If you used Claude just to accelerate production, let's say it makes you twice as productive. That means if you budget 30 minutes to do something, if Claude stops working, it means it'll take you 60. Or you just use the 30 and its half as good, not as polished, any number of things. Particularly if you know claude is really good at repetitive tasks, such as making sure all the font are the same size so you leave that until the end so you can focus on content.
> such as making sure all the font are the same size so you leave that until the end so you can focus on content.
`ctrl-a`, select desired font in style dropdown, done.
Do you perhaps have a better example?
The criticism people are throwing out here is that while users often claim that they're only using claude to speed up the exact thing they would have done anyway, they are in fact ceding control and critical thinking to claude and just accepting what it does as if it were what they would have done.
People with more budget than me give the task of polishing up the powerpoint to an assistant. They don't hire 20 assistants to have the work done 20 times as fast, but that does not negate that having one assistant do it does free up some of their time
Responsible use of Claude behaves much the same, just within my budget
I mean, if you were cutting deadlines this fine, with or without AI, then it was only a matter of time before your luck ran out.
> Particularly if you know claude is really good at repetitive tasks, such as making sure all the font are the same size so you leave that until the end so you can focus on content.
Surely you can come up with a better example than something that's solved by even the most basic tools we've had for decades at this point?
Yes it matters regardless of how much thinking you offloaded. Having the tools you’re using vanish would stop anyone. Imagine if your favorite editor or other tools vaporized and you had 30 minutes left.
If the OP had hours they’d probably be fine.
My different smug answer is: I bought a laptop with 64G of RAM before the RAM bubble and would try throwing qwen3 or gemma at it.
I think that ship has sailed. The entire world runs on SaaS tools which cannot / do not run locally. For mission critical ops, it can make sense to pay the enormous premium for local tools, but when one runs the risk/impact calculation, it rarely makes sense. Especially when there are four 9s attached to most of these online services.
> The entire world runs on SaaS tools which cannot / do not run locally.
Some of us have made different choices and therefore have different exposure to these kinds of failures. Yes, sometimes those choices mean higher costs or fewer features, but a lower probability of getting kneecapped by others' problems is a pretty good feature.
That ship is still very much available for boarding, but if your not already on it, it will take a bit of effort to get onboard.
Many of us refused to go along with the subscription model, and so are not affected. I've used primarily LibreOffice for ~two decades, and even more emphatically when Excel became primarily Subscription. Stuck with Gimp even after buying Photoshop for the Pantone support, but killed PS when it went online model. Was very interested in Fusion 360, but stuck with RhinoCAD when Fusion was online-only. I refuse to use online accounting/payroll "solutions".
I use Claude $20/mth a good amount, and am hesitant to use Chinese technology, but the last day makes it look a lot more sensible to setup local AIs, even at the cost of higher initial cost, lower capability, more system management effort, and potential compromise.
It's just one small shop, but many others could make similar decisions, instead of going with what's "just easier today".
If you use open router you can choose whatever you want whenever you want. Before I switched to mostly local I discovered like a lot of people how 99% of the time using the us labs is a complete waste of money.
I don't need to imagine it, this actually happened to me. OK, it was a bit more than 30 min, but I used Arch Linux back in the day, around '07. Coursework was due the next day and had loads to do. Was looking for a distraction so decided to `pacman -Syu` in the middle of the day. It completely borked my system. Didn't have enough time to fix so chucked in an Ubuntu CD and within an hour I was back with all my favourite tools.
It's very different when you own your tools rather than renting them.
> Having the tools you’re using vanish would stop anyone.
No. Not really. Depending on the presentation, it wouldn't stop professionals. More importantly, this is why you have backups. It's why you don't wait until the last minute to prepare your presentation, or, if you have to, you are knowledgeable enough to forego the presentation.
Simply put, in this case, absolutely not.
Edit: And to be clear, before you put too much weight on the "depending on the presentation" part, data heavy slides are not what I'm talking about.
If you completely stop being able to work with without Claude, then I say a deep analysis of the root cause is needed and it may have some uncomfortable findings.
But for many, they use it as a productivity multiplier. They can still get the job done, just slower. The problem is that the rest of the supply chain likely doesn't care, management deadlines likely don't have 'do without AI' timelines any longer, so if we define the "job" as needing to meet the deadlines and not just complete the task eventually, people will find themselves unable to do their "job" without AI.
These both can be simplified as "not being able to do your job without AI", but I propose we have little to gain and a decent bit to lose by doing anything that treats these two situations as the same.
No need to be so dramatic. There are a lot of things you can do with just speaking supported by a whiteboard with markers.
When I was still teaching, I’d do this sometimes. Just do a lecture with a whiteboard and draw data structures and algorithms. It’s often much more engaging than prepared slides.
Yes, I know it doesn’t work for every possible presentation, but for many it does.
Totally. But today, the room will not be well set up for it, markers may be dry, whiteboard covered in old, settled in diagrams, etc. And half the people dialing in. All solvable, but not in 30 minutes.
And for larger work presentations the expectation even 30 years ago was not whiteboards; that is very inefficient. It was transparencies. But try printing those in a pinch. And finding a projector.
That's not to say that a key tool outage should kill a presentation (and if it was important, do it early); but it is not a do-not-care either.
This is reminding me of a time in school were we thought if we stole all the chalk we might get a day off of Calculus class.
The professor came in and looked around for two minutes before shrugging and pulling a fresh stick of chalk out of his back pocket and began writing on the board.
This is the same professor who we'd never seen use a fresh piece of chalk ever, he always just soldiered through the nubs and remains littering the board's chalk tray.
We had no idea he actually kept a pack of sticks in his back pocket at all times.
Do you have a way to resolve the issue where the surface becomes degraded and much harder to erase over time? I think it's partly due to wear and tear and partly because people try to clean off the boards with solvents that degrade the surface that is erasable. Is there a way to renew it in some way?
The ones I have usually seen do have some coated steel (presumably because they are cheaper). At home we have a glass whiteboard (tempered glass I think?), which is extremely strong and the surface very smooth.
Are people seriously so helpless they can't stand up and talk about what they've been getting paid to do anymore? If so y'all need ai detoxes, you arent even people anymore.
Standing up and talking is not the problem, it's getting the information into other people's heads effectively. The adage "a picture is worth a thousand words" does have a basis, especially when the subject is very technical.
Asinine argument. Are you outsourcing your thinking to the compiler because it gives you warnings, instead of getting an error two hours into your punch card programs's runtime?
People have done this work by hand for years and years, and now when they've been using LLMs to do it 5x faster for a couple of years, suddenly they'd be unable to work the traditional way? Please
> My car broke down and now I'm late for work
> Well you shouldn't have outsourced your mobility to a machine
> People have done this work by hand for years and years, and now when they've been using LLMs to do it 5x faster for a couple of years, suddenly they'd be unable to work the traditional way?
Yes. Not completely unable all at once, but they certainly get worse at it fast.
Losing skills due to over-reliance on technology is not a new phenomenon and not exclusive to LLM use, but they of course exacerbate it by so pervasively touching so many areas.
You take a clockmaker who's used to working with hand-operated lathes, mills etc. and let him work for a few years using CNC versions. For sure he'll probably be worse at using the traditional machines when he comes back to them, but that doesn't mean he's any dumber or worse at making clocks. He's just out of practice, same as if he'd have spent a few years staffing a convenience store or anything else.
That seems to be precisely what your study shows as well, though I only read the summary. Say the AI lowered the endoscopists' workload by (for illustration) 20%, so they got 20% less practice per unit of time, which made them have a performance that was 20% worse in the trial. I would expect the same to happen if they got assigned to do paperwork for 20% of their time as well.
I'm not trying to be argumentative, perhaps we're in agreement here, I'm just clarifying my point.
There's a famous talk about the importance of knowing when to use which level of automation, and maintaining currency at all levels you expect to need to use: https://youtube.com/watch?v=5ESJH1NLMLs
That only works because the clockmaker had the manual experience to begin with and only needs to remember to get back into the groove. If however he had always only worked with CNC machines the story would be very different.
I'm not convinced that's true. If you have experience driving an automatic car with lane assist, you'll have an easier time driving a 1980s manual car than somebody with no driving experience at all.
Likewise, the machinist still knows the principles of construction even if he's only used CNC machines. The endoscopist still knows the principles of colonoscopy even if he's only done it with AI assistance.
Just anecdotally, back when I took courses in assembler and logic circuits, there seemed to be no great difference in success between people like myself that had previous experience with that, and people who'd at most done some high-level scripting.
If you have enough curiosity I doubt using AI will be a roadblock to true understanding, just like high-level programming languages weren't.
Analogies with physical tools don’t work well for LLMs because the former enable you to do things you couldn’t before, regardless how much you train and learn. A microscope and a needle allow us to see and work with precision beyond our human physical limitations. Furthermore they don’t make decisions, you are in control the whole time.
The talk looks interesting, thank you. I’ve added it to my list to watch later.
One claude instance also doesn't need billions of dollars of operational infrastructure to work.
Its great that you can keep your computer running, 1/3 of my information is ithe cloud (which is not a problem at all, even today) and 100% of my infrastructure is in the cloud.
Its not a problem at all, that a service has an outage. My laptop also had an outage when the logic board broke. Took 2 days to fix so what
> My laptop also had an outage when the logic board broke.
> And if you had a local backup, you'd be able to resume your work immediately.
Immediately after buying a new laptop. Or perhaps you mean you keep a spare laptop around? In any case, it's also an object that requires a huge environmentally destructive supply chain to create. Quite like a LLM...
No. I could go by with an older one, or in a pinch I could repurpose my current media server. Or I could put my data in an SSD with a bootable partition and borrow my wife's laptop...
> huge environmentally destructive supply chain to create.
Sunk cost fallacy? If the major electronics factories explode today, the existing computers won't stop working. If any of the Anthropic's datacenters go under, the compute capacity would have to be rebuilt.
If you're worried about the environment, you're better of stopping "outsourcing" your food gathering to factory farms or your walking to cars.
That's not what I'm talking about though, the question is whether the compiler does the thinking for you or not.
Or do you mean that the question of whether you're outsourcing your thinking to a LLM or not depends on whether you're running it locally? Would you claim you're outsourcing your thinking every time you run a script on a remote server?
It seems you didn't finish reading my comment before replying. Here's the relevant part in case you missed it:
> Or do you mean that the question of whether you're outsourcing your thinking to a LLM or not depends on whether you're running it locally? Would you claim you're outsourcing your thinking every time you run a script on a remote server?
In any case, your "argument" is more of an observation that
> Claude requires billions of dollars of operational infrastructure to work
Okay, so does a bunch of other things as well. So what?
Me and other commenters have tried quite hard to divine what you feel is relevant about this fact, without success. Perhaps it's time for you to try to articulate what significance you see in it.
My point is that your "outsourcing to a compiler" analogy does not make sense. And that running a hosted, proprietary LLM is not the same as "running a script in a remote server".
The whole "AI is a tool" argument only works if the "tool" is a commodity that is universally available and still exists even if the tool manufacturer disappears. Open-weights, locally hosted LLMs fit this criteria. Anything that depends on a cloud provider does not.
> The whole "AI is a tool" argument only works if the "tool" is a commodity that is universally available and still exists even if the tool manufacturer disappears.
You have quite strange criteria for what constitutes a "tool", but sure, at least I see where you're coming from then.
Linus Torvalds might be able to describe some obscure part of Linux Kernel memory management that LLMs might not be able to come up with, but he won't be able to type it out faster than he can have it generated off his explanation
are your audiences going to be exactly as pleased with just the prompt in the center of the sole slide just as they would with the complete generated slides? probably not
will they find AI-generated slides sloppy and worthless anyway? I might, but that is not exactly my business. also I know gcc or clang won't accept a prompt as source file(s) yet
This is where putting a provider-agnostic code agent between you and claude can save your bacon.
I can't speak for how other agents work, but in juggler the conversations are stored in a cross-model format, so if claude runs out of tokens or goes titsup during something urgent, I can just switch models to GPT and carry on without needing to restart the task.
From what I've seen the pricing is intentionally hostile towards third party, and therefore provider-agnostic, code agents. I have never used "Claude Code" or "Google Antimatter" or whatever it's called, I only use OpenCode, but I think the intent behind them and the reason only they, and not API access, is included in their static pricing is very transparent. If you want API access you are forced to pay-as-you-go, and many "normal people" do not want that. My way of solving it is simply using OpenRouter, setting up blacklists for models and pretty strict spending limits.
Also convenient to continue working when security rules cause the agent to refuse to help you. Today I got Claude to finish something GPT refused to do.
My strategy: diversification. I run deepseek harness (which I made a vs code integration for and extended it a lot with plugins) and integrated it with my codex and claude subscriptions as well as an openrouter account. This way I can switch models and providers during my sessions and does not get blocked on such things.
Useful for
- outages
- running out of quota
- failing agents e.g. because of "cyber" flaging
Good diversification. I like to have a strong local model like qwen3.8-flash-next available because I live in the mountains and my Internet is not 100% reliable. I don’t get it that companies allow themselves to get locked in on one vendor. Two or more, with local makes more sense.
Okay, this is slightly horrifying to me but also a bit genius...
But I guess Anthropic models are trained for Claude Code's tool call shape (i.e. the edit tool). Does it deal well enough with the MCP tools that Deepseek harness gives the model instead?
I don't think that "Anthropic models are trained for Claude Code's tool call shape". It's all about system prompt and model description (which is part of that prompt).
I observe that most models can deal well with the tools you provide.
Form time to time I analyze my sessions to see which tools could improve my (or my agent's) work. E.g. a background timer/watcher is on my todo list, so that I can chat with the agent session while its waiting for some shell response.
> Okay, this is slightly horrifying to me but also a bit genius...
If you're old enough to remember CGI web services, originally every request to a web server for non-static content launched a binary. Most of the web ran on that approach for a long time, later optimized with things like FCGI.
dsh looks intriguing. Do they have any plugins that help effectively extend the context memory of each of the LLM's? I am starting to watch some videos on it but it will take me a while to get through them.
I think there are plenty ways to "extend the context memory". Compaction is the most obvious: reduce old context to it's core and free tokens. Another tool is session search. The models can look back in parts of the session that got "compacted away" and also in other sessions.
But honestly with 500k or 1m token sessions, I rarely run out of context memory and barely notice any loss of precision after compaction.
I have a great feeling, it's cruisy as hell without emails! I'm able to get a morning of focus work done without interruption, and then I think it'll be a POETS Day.
I've not seen anyone use the term poet's day for many years. My dad used to be a tax accountant for British Rail and said that it was old corporate slang. (Piss off early, tomorrow's Saturday.)
Remember all those recent stories about "AI is really good at hacking now"?
Meanwhile SEGA's arcade machines have been offline for days: https://www.techpowerup.com/354714/sega-network-hack-keeps-j... and fanfiction writing site AO3 is down for two weeks so far. Then there's the freezer hacks, one of which impacted the French blood service.
It's been a long time since I read COMP.RISKS (and indeed it's no longer published) but this feels like an unusually high level of incidents.
Good thing is at can expect some compensation such as resets. And yeah - diversify your providers, is the same with electricity from utility perspective…
I fully agree on the snark, but if anyone seriously considers backup from grid...
Get an islanding-capable inverter separating your internal power setup from grid, connect it with battery systems - you can control battery charging and have "microgrid UPS" then. Combine it with solar power feeding as second feed into inverter. Scale battery and solar panels as needed. If you have third input, can add three-phase genset (I recommend looking for propane fueled one if you don't want to care about fuel going bad)
I have this pretty much and generating way too much power for our needs which is great. And that includes car charging for light but significant use. I export a lot during peak hours from the battery as there is a lot of surplus.
But I was snarking on "provider". There are 100s of providers where I live but one grid and you can't use two concurrently.
My backup in the past has been a local library for internet and laptop power and everything else can wait.
This is all getting quite silly but really large sites with critical power needs (hospitals, factories, data centres, etc) will sometimes have two independent grid supplies running in from separate substations, preferably going as far back down the chain as possible before hitting a single point of failure. They'll also have UPS and diesel generators able to kick in and keep things running if somehow they entirely lose grid power. Its all totally doable, but as you say for most people going to the local library or a coffee shop is a perfectly fine fallback plan.
If you were to setup a GPU cluster in your basement and consume TWs of power, the grid going down would no longer be your problem. It would become their problem! /s
.. to the same grid. If you're lucky they're to different substations.
The number of people who had "independent" fiber connections which were routed through the same channel and destroyed by the same backhoe is so high it's a meme among sysadmins.
Now they're one grid with a controllable weak link. They might solve the all-or-nothing interconnection problem of AC grids, but they can still cascade failures - and they will, if one goes down and everyone switches to the other.
Do you disambiguate between failure of the grid and failure of the usine? Like failure of the power generator is not the same as cut wires. And in terms of AI and internet providers - both can be easily diversified.
Besides, what is it that you guys expect from a large scale provider of essentially experimental technology being tuned live as it goes? Is amazing this all works so good considering the complexity underneath.
Does every large company have a big server with some h100s and a couple of open weight models for just such an occasion. Especially if you don’t have the luxury of being able to wait or postpone eg under cyber attack?
I hope that doesn’t happen. There is a good alternative: if MOE models work for you, then Apple Silicon does a decent job at inference using whatever harnesses you like.
It depends on what "us" means. I am switched off of Claude almost entirely now except for background things where I want to abuse my subsidized account. Otherwise, OpenRouter + open weight models are my daily drivers. It takes a bunch of work but I'm doing it for roughly a max20 account cost, a bit more, but it's not like the difference is huge. They are less capable but I've modified my workflow to mitigate that as much as possible, and I can still do things when Anthropic goes down.
What models are you using? I was satisfied with deepseek v4.1 while ai was still using my brain, before being time-pressured into shipping something I don't understand and now I feel I need stronger models to steer it. Kimi K3 or the closed models through openrouter is too expensive compared to subsidized plans though so I'm using a $20 sub for light reviews/consultations.
As temporary backup, one can use claude models hosted on GCP - my understanding is that those don't depend on platform.claude.com and should be available even if main claude platform is dead
I got logged out of Claude in the morning, she couldn’t get back in without email. I guess Anthropic can keep their twenty bucks / days in November, for today.
It really is Friday the 13th today with outlook and now Claude both being mostly down. I wonder if this is related though. Either way not going to be a great ad for their IPO.
> Either way not going to be a great ad for their IPO.
Utility customer satisfaction goes up after power outages.
Obviously there are caveats to that. They must be rare, have decent communication, etc. But the point is that people will take always-on services for granted and be reminded of how much their lives are improved when they suddenly don't have them.
Just got kicked out while finalising a presentation I'm about to give in 39 minutes.
Wish me luck!
That is how the future looks like. One outage and we have no idea what to do if AI is not working.
Unless the future is local?
The future is local, planned or unplanned
How'd the presentation go?
I wonder how the presentation is going
I remember back in the days when Word crashed and I hadn't saved saved the last hour of work. A lesson everybody has to learn some day.
Reminds me of the SNL skit of TV newsreaders when the teleprompter goes down. By the end of it they've gone completely primitive and started killing each other.
found it: https://www.youtube.com/watch?v=Nc0FeUMQIA4
SNL skits have these great ideas every now and then, only if the execution were a bit better, they'd have timeless pieces (like the Monty Python's dirty fork). This one was very funny anyway.
One show is produced with edits and retakes if someone flubs a line or breaks, the other is live. Also consider, SNL puts out more content in a single season with 22 episodes at 90 minutes, then all of Monty Python's full run of 45 episodes at 30 minutes. You're comparing apples to oranges.
Sometimes you look at an orange, and think how convenient it'd be if you could bite it like an apple. Okay, many don't, but few who do came up with all the modern fruits :)
You would think that 20 episodes a season over 50 years, then on just pure statistical probability they would have had a funny skit. An infinite number of monkeys on an infinite number of type writers will eventually write a funny joke. But somehow, against all odds, despite hiring comedians, comedic actors, improvers, and writers. They have never, once, gotten a funny skit. It defies belief actually.
Great skit, thanks for sharing.
WhenI started watching, I was thinking I’ll share this on Teams to joke about the outage…then the third act came, and that it’s definitely nsfw.
I ASSURE you that SNL skits have wider appeal than Monty Python, which really just appeals to a specific subset.
This can't possibly be true.
Maybe to Americans? Not for the ROW
As a non american SNL is 99.9% asinine crap and pales in comparison to Monty Python even after all this time. May be that's why it has a wider appeal but the formulaic lack of humor and fear of actually punching up comes down to the bitter old man in charge.
And instead of using your time wisely to finish it manually, you're posting about it on HN?
Your intern rage quit.
We are all bosses now. Until it gets tired of working for us.
[flagged]
Of course it matters. If you used Claude just to accelerate production, let's say it makes you twice as productive. That means if you budget 30 minutes to do something, if Claude stops working, it means it'll take you 60. Or you just use the 30 and its half as good, not as polished, any number of things. Particularly if you know claude is really good at repetitive tasks, such as making sure all the font are the same size so you leave that until the end so you can focus on content.
The example are contrived, but still very real.
> such as making sure all the font are the same size so you leave that until the end so you can focus on content.
`ctrl-a`, select desired font in style dropdown, done.
Do you perhaps have a better example?
The criticism people are throwing out here is that while users often claim that they're only using claude to speed up the exact thing they would have done anyway, they are in fact ceding control and critical thinking to claude and just accepting what it does as if it were what they would have done.
Clearly, this mess would have been avoided by using two Claudes to cut prep time to 15 minutes.
People with more budget than me give the task of polishing up the powerpoint to an assistant. They don't hire 20 assistants to have the work done 20 times as fast, but that does not negate that having one assistant do it does free up some of their time
Responsible use of Claude behaves much the same, just within my budget
Huang needed to have a baby in 3 months, so,
why stop there? i hear Boris Cherny has thousands of instances running in parallel, he can do it in milliseconds
He makes Gary Tan seem thrifty.
I mean, if you were cutting deadlines this fine, with or without AI, then it was only a matter of time before your luck ran out.
> Particularly if you know claude is really good at repetitive tasks, such as making sure all the font are the same size so you leave that until the end so you can focus on content.
Surely you can come up with a better example than something that's solved by even the most basic tools we've had for decades at this point?
My thinking took a lifetime, my slides an hour or two. It's not the content, its the format.
Yes it matters regardless of how much thinking you offloaded. Having the tools you’re using vanish would stop anyone. Imagine if your favorite editor or other tools vaporized and you had 30 minutes left.
If the OP had hours they’d probably be fine.
My different smug answer is: I bought a laptop with 64G of RAM before the RAM bubble and would try throwing qwen3 or gemma at it.
If my favorite editor disappeared I would literally be fine.
Because there are many other editors. Perhaps this is a lesson to have backup LLMs ready to plug in.
Or always use tools that you have complete access to, i.e. that run entirely on equipment you own.
I think that ship has sailed. The entire world runs on SaaS tools which cannot / do not run locally. For mission critical ops, it can make sense to pay the enormous premium for local tools, but when one runs the risk/impact calculation, it rarely makes sense. Especially when there are four 9s attached to most of these online services.
> The entire world runs on SaaS tools which cannot / do not run locally.
Some of us have made different choices and therefore have different exposure to these kinds of failures. Yes, sometimes those choices mean higher costs or fewer features, but a lower probability of getting kneecapped by others' problems is a pretty good feature.
Life is tradeoffs.
That ship is still very much available for boarding, but if your not already on it, it will take a bit of effort to get onboard.
Many of us refused to go along with the subscription model, and so are not affected. I've used primarily LibreOffice for ~two decades, and even more emphatically when Excel became primarily Subscription. Stuck with Gimp even after buying Photoshop for the Pantone support, but killed PS when it went online model. Was very interested in Fusion 360, but stuck with RhinoCAD when Fusion was online-only. I refuse to use online accounting/payroll "solutions".
I use Claude $20/mth a good amount, and am hesitant to use Chinese technology, but the last day makes it look a lot more sensible to setup local AIs, even at the cost of higher initial cost, lower capability, more system management effort, and potential compromise.
It's just one small shop, but many others could make similar decisions, instead of going with what's "just easier today".
If you use open router you can choose whatever you want whenever you want. Before I switched to mostly local I discovered like a lot of people how 99% of the time using the us labs is a complete waste of money.
I don't need to imagine it, this actually happened to me. OK, it was a bit more than 30 min, but I used Arch Linux back in the day, around '07. Coursework was due the next day and had loads to do. Was looking for a distraction so decided to `pacman -Syu` in the middle of the day. It completely borked my system. Didn't have enough time to fix so chucked in an Ubuntu CD and within an hour I was back with all my favourite tools.
It's very different when you own your tools rather than renting them.
I had the same experience with Debian unstable and tetex, cca. 2003. Don't upgrade your production system, a good life lesson.
> Having the tools you’re using vanish would stop anyone.
No. Not really. Depending on the presentation, it wouldn't stop professionals. More importantly, this is why you have backups. It's why you don't wait until the last minute to prepare your presentation, or, if you have to, you are knowledgeable enough to forego the presentation.
Simply put, in this case, absolutely not.
Edit: And to be clear, before you put too much weight on the "depending on the presentation" part, data heavy slides are not what I'm talking about.
time for a chalk-talk.
If you completely stop being able to work with without Claude, then I say a deep analysis of the root cause is needed and it may have some uncomfortable findings.
But for many, they use it as a productivity multiplier. They can still get the job done, just slower. The problem is that the rest of the supply chain likely doesn't care, management deadlines likely don't have 'do without AI' timelines any longer, so if we define the "job" as needing to meet the deadlines and not just complete the task eventually, people will find themselves unable to do their "job" without AI.
These both can be simplified as "not being able to do your job without AI", but I propose we have little to gain and a decent bit to lose by doing anything that treats these two situations as the same.
Why stop at Claude? Forget computers, bring the good ol' slider projection machine thingie and do super old school.
Even better, if cave painting was good enough thousands of years ago it is good enough today.
No need to be so dramatic. There are a lot of things you can do with just speaking supported by a whiteboard with markers.
When I was still teaching, I’d do this sometimes. Just do a lecture with a whiteboard and draw data structures and algorithms. It’s often much more engaging than prepared slides.
Yes, I know it doesn’t work for every possible presentation, but for many it does.
Totally. But today, the room will not be well set up for it, markers may be dry, whiteboard covered in old, settled in diagrams, etc. And half the people dialing in. All solvable, but not in 30 minutes.
And for larger work presentations the expectation even 30 years ago was not whiteboards; that is very inefficient. It was transparencies. But try printing those in a pinch. And finding a projector.
That's not to say that a key tool outage should kill a presentation (and if it was important, do it early); but it is not a do-not-care either.
How does any tech company or team operate without whiteboards?
Microsoft Surface Tables!
It's much easier to anticipate dry markers (keep a small stock) than outage of Claude or Microsoft.
(Yes, I know that the slippery slope is in jest.)
This is reminding me of a time in school were we thought if we stole all the chalk we might get a day off of Calculus class.
The professor came in and looked around for two minutes before shrugging and pulling a fresh stick of chalk out of his back pocket and began writing on the board.
This is the same professor who we'd never seen use a fresh piece of chalk ever, he always just soldiered through the nubs and remains littering the board's chalk tray.
We had no idea he actually kept a pack of sticks in his back pocket at all times.
Last time I worked in an office I kept my own private stash of markers exactly for that reason.
Magnetic drawing boards are also a thing that last forever.
Do you have a way to resolve the issue where the surface becomes degraded and much harder to erase over time? I think it's partly due to wear and tear and partly because people try to clean off the boards with solvents that degrade the surface that is erasable. Is there a way to renew it in some way?
The ones I have usually seen do have some coated steel (presumably because they are cheaper). At home we have a glass whiteboard (tempered glass I think?), which is extremely strong and the surface very smooth.
Or whiteboards and blackboards.
Are people seriously so helpless they can't stand up and talk about what they've been getting paid to do anymore? If so y'all need ai detoxes, you arent even people anymore.
Standing up and talking is not the problem, it's getting the information into other people's heads effectively. The adage "a picture is worth a thousand words" does have a basis, especially when the subject is very technical.
Yea but, you just think about it. Or you ask a colleague to rubber duck. AI usually has really weird takes.
Only Claude understands their codebases... so they probably are not thinking anymore.
Asinine argument. Are you outsourcing your thinking to the compiler because it gives you warnings, instead of getting an error two hours into your punch card programs's runtime?
People have done this work by hand for years and years, and now when they've been using LLMs to do it 5x faster for a couple of years, suddenly they'd be unable to work the traditional way? Please
> My car broke down and now I'm late for work
> Well you shouldn't have outsourced your mobility to a machine
> People have done this work by hand for years and years, and now when they've been using LLMs to do it 5x faster for a couple of years, suddenly they'd be unable to work the traditional way?
Yes. Not completely unable all at once, but they certainly get worse at it fast.
https://www.thelancet.com/journals/langas/article/PIIS2468-1...
Losing skills due to over-reliance on technology is not a new phenomenon and not exclusive to LLM use, but they of course exacerbate it by so pervasively touching so many areas.
https://www.scientificamerican.com/article/how-gps-weakens-m...
You take a clockmaker who's used to working with hand-operated lathes, mills etc. and let him work for a few years using CNC versions. For sure he'll probably be worse at using the traditional machines when he comes back to them, but that doesn't mean he's any dumber or worse at making clocks. He's just out of practice, same as if he'd have spent a few years staffing a convenience store or anything else.
That seems to be precisely what your study shows as well, though I only read the summary. Say the AI lowered the endoscopists' workload by (for illustration) 20%, so they got 20% less practice per unit of time, which made them have a performance that was 20% worse in the trial. I would expect the same to happen if they got assigned to do paperwork for 20% of their time as well.
I'm not trying to be argumentative, perhaps we're in agreement here, I'm just clarifying my point.
There's a famous talk about the importance of knowing when to use which level of automation, and maintaining currency at all levels you expect to need to use: https://youtube.com/watch?v=5ESJH1NLMLs
That only works because the clockmaker had the manual experience to begin with and only needs to remember to get back into the groove. If however he had always only worked with CNC machines the story would be very different.
I'm not convinced that's true. If you have experience driving an automatic car with lane assist, you'll have an easier time driving a 1980s manual car than somebody with no driving experience at all.
Likewise, the machinist still knows the principles of construction even if he's only used CNC machines. The endoscopist still knows the principles of colonoscopy even if he's only done it with AI assistance.
Just anecdotally, back when I took courses in assembler and logic circuits, there seemed to be no great difference in success between people like myself that had previous experience with that, and people who'd at most done some high-level scripting.
If you have enough curiosity I doubt using AI will be a roadblock to true understanding, just like high-level programming languages weren't.
> If you have enough curiosity
That’s a gigantic “if”. Anything can become a positive in the right hands, but we have to consider the societal impact of broadly available tools.
Analogies with physical tools don’t work well for LLMs because the former enable you to do things you couldn’t before, regardless how much you train and learn. A microscope and a needle allow us to see and work with precision beyond our human physical limitations. Furthermore they don’t make decisions, you are in control the whole time.
The talk looks interesting, thank you. I’ve added it to my list to watch later.
My compiler does not require billions of dollars of opearational infrastructure to work, except for the electrical grid.
And even in the case of outage, I can keep my computer running on a battery or a backup generator.
One claude instance also doesn't need billions of dollars of operational infrastructure to work.
Its great that you can keep your computer running, 1/3 of my information is ithe cloud (which is not a problem at all, even today) and 100% of my infrastructure is in the cloud.
Its not a problem at all, that a service has an outage. My laptop also had an outage when the logic board broke. Took 2 days to fix so what
> One claude instance also doesn't need billions of dollars of operational infrastructure to work.
Of course it does. It's the marginal cost that is low.
> 100% of my infrastructure is in the cloud.
https://en.wikipedia.org/wiki/Turkey_illusion
> My laptop also had an outage when the logic board broke.
And if you had a local backup, you'd be able to resume your work immediately.
> My laptop also had an outage when the logic board broke.
> And if you had a local backup, you'd be able to resume your work immediately.
Immediately after buying a new laptop. Or perhaps you mean you keep a spare laptop around? In any case, it's also an object that requires a huge environmentally destructive supply chain to create. Quite like a LLM...
> Immediately after buying a new laptop
No. I could go by with an older one, or in a pinch I could repurpose my current media server. Or I could put my data in an SSD with a bootable partition and borrow my wife's laptop...
> huge environmentally destructive supply chain to create.
Sunk cost fallacy? If the major electronics factories explode today, the existing computers won't stop working. If any of the Anthropic's datacenters go under, the compute capacity would have to be rebuilt.
If you're worried about the environment, you're better of stopping "outsourcing" your food gathering to factory farms or your walking to cars.
That's not what I'm talking about though, the question is whether the compiler does the thinking for you or not.
Or do you mean that the question of whether you're outsourcing your thinking to a LLM or not depends on whether you're running it locally? Would you claim you're outsourcing your thinking every time you run a script on a remote server?
My argument is specific to Claude, and it has nothing to do with "the environment".
It seems you didn't finish reading my comment before replying. Here's the relevant part in case you missed it:
> Or do you mean that the question of whether you're outsourcing your thinking to a LLM or not depends on whether you're running it locally? Would you claim you're outsourcing your thinking every time you run a script on a remote server?
In any case, your "argument" is more of an observation that
> Claude requires billions of dollars of operational infrastructure to work
Okay, so does a bunch of other things as well. So what?
Me and other commenters have tried quite hard to divine what you feel is relevant about this fact, without success. Perhaps it's time for you to try to articulate what significance you see in it.
My point is that your "outsourcing to a compiler" analogy does not make sense. And that running a hosted, proprietary LLM is not the same as "running a script in a remote server".
The whole "AI is a tool" argument only works if the "tool" is a commodity that is universally available and still exists even if the tool manufacturer disappears. Open-weights, locally hosted LLMs fit this criteria. Anything that depends on a cloud provider does not.
> The whole "AI is a tool" argument only works if the "tool" is a commodity that is universally available and still exists even if the tool manufacturer disappears.
You have quite strange criteria for what constitutes a "tool", but sure, at least I see where you're coming from then.
Linus Torvalds might be able to describe some obscure part of Linux Kernel memory management that LLMs might not be able to come up with, but he won't be able to type it out faster than he can have it generated off his explanation
are your audiences going to be exactly as pleased with just the prompt in the center of the sole slide just as they would with the complete generated slides? probably not
will they find AI-generated slides sloppy and worthless anyway? I might, but that is not exactly my business. also I know gcc or clang won't accept a prompt as source file(s) yet
This is where putting a provider-agnostic code agent between you and claude can save your bacon.
I can't speak for how other agents work, but in juggler the conversations are stored in a cross-model format, so if claude runs out of tokens or goes titsup during something urgent, I can just switch models to GPT and carry on without needing to restart the task.
From what I've seen the pricing is intentionally hostile towards third party, and therefore provider-agnostic, code agents. I have never used "Claude Code" or "Google Antimatter" or whatever it's called, I only use OpenCode, but I think the intent behind them and the reason only they, and not API access, is included in their static pricing is very transparent. If you want API access you are forced to pay-as-you-go, and many "normal people" do not want that. My way of solving it is simply using OpenRouter, setting up blacklists for models and pretty strict spending limits.
Yup.
exe.dev works the same way.
Also convenient to continue working when security rules cause the agent to refuse to help you. Today I got Claude to finish something GPT refused to do.
My strategy: diversification. I run deepseek harness (which I made a vs code integration for and extended it a lot with plugins) and integrated it with my codex and claude subscriptions as well as an openrouter account. This way I can switch models and providers during my sessions and does not get blocked on such things.
Useful for - outages - running out of quota - failing agents e.g. because of "cyber" flaging
Good diversification. I like to have a strong local model like qwen3.8-flash-next available because I live in the mountains and my Internet is not 100% reliable. I don’t get it that companies allow themselves to get locked in on one vendor. Two or more, with local makes more sense.
You're on HN, and "vendor lock-in" surprises you? ;)
How long has this been working for you? (Or others).
Since Anthropic doesn't seem to like subscriptions being used in third party harness, do they actually enforce that nowadays?
https://github.com/micw/claude-wrapper-advanced I wrap their CLI ^^
Edit: CLI in headless mode is an official feature
Okay, this is slightly horrifying to me but also a bit genius...
But I guess Anthropic models are trained for Claude Code's tool call shape (i.e. the edit tool). Does it deal well enough with the MCP tools that Deepseek harness gives the model instead?
I don't think that "Anthropic models are trained for Claude Code's tool call shape". It's all about system prompt and model description (which is part of that prompt).
I observe that most models can deal well with the tools you provide.
Form time to time I analyze my sessions to see which tools could improve my (or my agent's) work. E.g. a background timer/watcher is on my todo list, so that I can chat with the agent session while its waiting for some shell response.
> Okay, this is slightly horrifying to me but also a bit genius...
If you're old enough to remember CGI web services, originally every request to a web server for non-static content launched a binary. Most of the web ran on that approach for a long time, later optimized with things like FCGI.
dsh looks intriguing. Do they have any plugins that help effectively extend the context memory of each of the LLM's? I am starting to watch some videos on it but it will take me a while to get through them.
I think there are plenty ways to "extend the context memory". Compaction is the most obvious: reduce old context to it's core and free tokens. Another tool is session search. The models can look back in parts of the session that got "compacted away" and also in other sessions.
But honestly with 500k or 1m token sessions, I rarely run out of context memory and barely notice any loss of precision after compaction.
Can OpenCode work in the same way?
For coding, I wouldn't use deepseek as a replacement. At least not for anything hard.
Except that DeepSeek Harness is very well done. It is not just the model.
somewhere down the line, a hook that triggers through an outlook email, didn't trigger today
Don't worry, it'll trigger twice tomorrow. Or the day after. And everything will be fine, right?
It was a load bearing hook.
I have a bad bad feeling about the outages happening today
I have a great feeling, it's cruisy as hell without emails! I'm able to get a morning of focus work done without interruption, and then I think it'll be a POETS Day.
I've not seen anyone use the term poet's day for many years. My dad used to be a tax accountant for British Rail and said that it was old corporate slang. (Piss off early, tomorrow's Saturday.)
So you are just waiting for the HN outage now?
Nah, I was kickin' round HN with a beer in hand as I listened to a mate DJ on Twitch. Not a bad afternoon!
(I did log in and do 30 mins work later on once systems were up but we all have our moments of weakness...)
its not that deep
Could be singularity or just Friday 13th
Remember all those recent stories about "AI is really good at hacking now"?
Meanwhile SEGA's arcade machines have been offline for days: https://www.techpowerup.com/354714/sega-network-hack-keeps-j... and fanfiction writing site AO3 is down for two weeks so far. Then there's the freezer hacks, one of which impacted the French blood service.
It's been a long time since I read COMP.RISKS (and indeed it's no longer published) but this feels like an unusually high level of incidents.
good friday gift for europeans who don't work at Microsoft that's for sure
Duplicate of https://news.ycombinator.com/item?id=50481187
Good thing is at can expect some compensation such as resets. And yeah - diversify your providers, is the same with electricity from utility perspective…
I only have one grid connected where do I get a second?
I fully agree on the snark, but if anyone seriously considers backup from grid...
Get an islanding-capable inverter separating your internal power setup from grid, connect it with battery systems - you can control battery charging and have "microgrid UPS" then. Combine it with solar power feeding as second feed into inverter. Scale battery and solar panels as needed. If you have third input, can add three-phase genset (I recommend looking for propane fueled one if you don't want to care about fuel going bad)
I have this pretty much and generating way too much power for our needs which is great. And that includes car charging for light but significant use. I export a lot during peak hours from the battery as there is a lot of surplus.
But I was snarking on "provider". There are 100s of providers where I live but one grid and you can't use two concurrently.
My backup in the past has been a local library for internet and laptop power and everything else can wait.
This is all getting quite silly but really large sites with critical power needs (hospitals, factories, data centres, etc) will sometimes have two independent grid supplies running in from separate substations, preferably going as far back down the chain as possible before hitting a single point of failure. They'll also have UPS and diesel generators able to kick in and keep things running if somehow they entirely lose grid power. Its all totally doable, but as you say for most people going to the local library or a coffee shop is a perfectly fine fallback plan.
That's what a UPS and a backup generator were for, even before PV and industrial-scale batteries became a thing.
Large consumers often have two independent grid feeds.
OK. I’m not a large consumer.
If you were to setup a GPU cluster in your basement and consume TWs of power, the grid going down would no longer be your problem. It would become their problem! /s
.. to the same grid. If you're lucky they're to different substations.
The number of people who had "independent" fiber connections which were routed through the same channel and destroyed by the same backhoe is so high it's a meme among sysadmins.
The problem with trying to have two grids is that if anyone ever connects them together, now they're one grid with a weak link in it.
See Japan (2) and the US (3).
DC links are now feasible for these situations, and don't have certain failure properties of AC grids (synchronization/frequency/phase).
Now they're one grid with a controllable weak link. They might solve the all-or-nothing interconnection problem of AC grids, but they can still cascade failures - and they will, if one goes down and everyone switches to the other.
The two Japanese systems have 130+ GW of generating capacity each and the links between them have a total capacity of about 2 GW (<1%).
Do you disambiguate between failure of the grid and failure of the usine? Like failure of the power generator is not the same as cut wires. And in terms of AI and internet providers - both can be easily diversified.
Besides, what is it that you guys expect from a large scale provider of essentially experimental technology being tuned live as it goes? Is amazing this all works so good considering the complexity underneath.
> A fix has been deployed for the issue affecting Claude services and we are monitoring for recovery. > Nov 13, 09:58 UTC
Claude was down yesterday too - started around the time of the m365 outage for me.
Does every large company have a big server with some h100s and a couple of open weight models for just such an occasion. Especially if you don’t have the luxury of being able to wait or postpone eg under cyber attack?
Only until NVIDIA starts cryptography verification and block anything not from US closed models, which should be anytime soon
I hope that doesn’t happen. There is a good alternative: if MOE models work for you, then Apple Silicon does a decent job at inference using whatever harnesses you like.
Just get an OpenRouter or Vercel account. Your option aren't pay Anthropic or buy hardware yourself, there are hundreds of providers out there.
Claude has us locked in right now while they subsidize cheap usage to get us subscribed and dependent on their monthly plans.
It depends on what "us" means. I am switched off of Claude almost entirely now except for background things where I want to abuse my subsidized account. Otherwise, OpenRouter + open weight models are my daily drivers. It takes a bunch of work but I'm doing it for roughly a max20 account cost, a bit more, but it's not like the difference is huge. They are less capable but I've modified my workflow to mitigate that as much as possible, and I can still do things when Anthropic goes down.
What models are you using? I was satisfied with deepseek v4.1 while ai was still using my brain, before being time-pressured into shipping something I don't understand and now I feel I need stronger models to steer it. Kimi K3 or the closed models through openrouter is too expensive compared to subsidized plans though so I'm using a $20 sub for light reviews/consultations.
As temporary backup, one can use claude models hosted on GCP - my understanding is that those don't depend on platform.claude.com and should be available even if main claude platform is dead
There is an absolute tonne of competition.
Maybe I can help mop the warehouse or something until it’s back (I’m half joking at this point…)
Working fine for me here in Europe.
it was down in the UK for approx 20 mins but only affected auth
Down in Sweden
Same
Received 3 emails in the last 2 hours. But widely communicated that "Teams is the way now!".
Luckily I can sometimes forget to check my email for an entire day, and it nobody even notices. At least _I_ am not that dependent on it.
Also some safe browsing thing is offline.. we just click "disable it and continue"... people are feeling very safe indeed.
Future is local, onpremise
Unacceptable. Just started my workday and four projects are now ruined due to modified source files.
‘Unacceptable’ is exactly what I would say to anyone relying on a cloud service that has exactly zero SLA.
Unless you’re on a negotiated 99.9% contract, you accepted this.
They don't even guarantee that the model stays the same without hidden nerfs. The most obvious thing one would demand if paying for their service.
Where is the version control?
"ruined due to modified source files"?
I think it's a bit. If it is, it's a pretty funny one.
It's one way to describe the way some people write code I suppose.
Taking immutability discipline to new levels!
Your next question on Claude should be "how do I setup a git server and how to use it"
Working fine for me.
Is it? Huh
"Software is solved" they said.
Just need GitHub to stop by and join the outage party.
And Cloudflare too, please God.
The two of you are going to jinx it.
Microsoft+Claude make it a perfect AI day.
I didn't think an over-reliance on AI and firing all the human coders would come back to bite them so quickly.
I'd guess what we see today is just a demo version of what's coming.
I'm more concerned about what's coming when AI works fine. We aren't in the enshittification phase yet.
I think AI started in the shitty phase, we may need a new term for when it gets worse than it already is.
> Great catch!
A trial run.
How is this biting anyone?
Its friday, people will just do certain things later.
I got logged out of Claude in the morning, she couldn’t get back in without email. I guess Anthropic can keep their twenty bucks / days in November, for today.
Where is your god now?
It really is Friday the 13th today with outlook and now Claude both being mostly down. I wonder if this is related though. Either way not going to be a great ad for their IPO.
> Either way not going to be a great ad for their IPO.
Utility customer satisfaction goes up after power outages.
Obviously there are caveats to that. They must be rare, have decent communication, etc. But the point is that people will take always-on services for granted and be reminded of how much their lives are improved when they suddenly don't have them.
Microsoft PM-s overloading Claude with outage solution prompts.
Will we be able to call in sick in future if Claude is down?
The prayers to Yellowstone are working. Give the magma chamber everything you’ve got, we are so close to ending it.
[clarkson_reaction.gif] Oh no! Anyway.
Time to do some downtime aktivities like in xkcd 303.
Oh no. Anyway...
Hey claude users: is there a way to access claude without being gated by a whatwg cartel web engine?
Alice what the fuck
oh no wha happand, how terrible, nooo
millions/billions around the world suddenly lose the ability to speelchek
"I have a drug that will make you twice as smart. Just pay me $20/month for the rest of your life."
"What happens if I stop paying, or Microsoft fucks up and cuts off your supply?"
"You get half as smart as you were before you ever started taking it. First month is free!"
"Sign me up!"
Imagine relying on AI for work and not having backup.