There is a difference between trying something and performing a test whose results are meant to be representative of most setups. A lot of people (in tech surprisingly) who see themselves as practitioners of applied science, apply flawed methodologies and try to generalise the results.
Just because testing is fast, it does not mean that it is free or cheap. Plus time is money and verifying every LLM setup every koolaid dev is proclaiming on bluesky could easily be a full time job.
Digital systems don't necessarily deteriorate immediately after the causal factors. Like technical debt, issues grow unnoticed and become visible gradually.
The personalised feeds is the whole point of their reason of being. No personalisation no profits. There just isn't a world where these huge companies can exist without personalised feeds. Meta platforms as well as YT would die and you would have instead something like Nebula which you pay for like Netflix but offers content creators content. Most people wouldn't pay for Nebula and so, you are back to the original point. People would put pressure to restore personalised feeds. Everyone supports these bans until they are the target of them.
This isn't a technological problem but a human one. The fundamental problem is that we haven't developed generalisable, scalable and profitable business models on the internet that aren't toxic.
Not true. YouTube, Facebook and Twitter were all successful products before they added personalised algorithmic feeds. Facebook just showed you everything by your friends. Twitter showed your follows (like bsky) and YouTube showed your channels.
The personalised algorithmic junk came later. It was never required for the websites to be wildly popular.
The tech isn’t that expensive or complex. Email didn’t have to add all this junk to be sustainable.
Something like the original Facebook where it’s just posts from your direct friends and no public content would be sustainable. It just wouldn’t make a trillion dollar company.
What happens when you have a codebase made with claude using this setup and claude is down for let's say 8 hours? Are you able to efficiently, smoothly and productively take over the codebase?
You could say the same thing about any always online software suite and it would be equally fair as we move into more agentic development workflows.
EX. Sure, you could go back to the old ways of using a drafting table for your engineering work if CAD went down but it would be exponentially slower…
Personally with my workflow I spend 30-60 minutes per Claude feature spec doc when I’m pair planning. If Claude goes down I would just prepare spec docs on my own until it came back online and then rapidly review them before calling the coding workflow.
> You could say the same thing about any always online software suite
But this is the reason "serious shops" do not use always online software and tools in critical parts of the SDLC. There is a difference between influencers/people on socials promoting things vs. reality where the expectation is that things don't just stop working because there is an internet outage or some 3rd party disruption
I would argue that it's really only toy projects that can continue in an Internet outage. "Serious shops" will be using cloud based version control, cloud based testing workflows, and most likely cloud based distribution of the software. isn't it only the little side projects you can get away with not needing the Internet for? Software long ago stopped being something one person on a computer did, today the professional SDLC includes many tools that are hosted.
Do farmers still plough fields with a Horse just in case their tractor runs out of diesel? Of course not, as technology moves on we all have to accept the inherent risks in exchange for the huge benefits, otherwise the work you do will be too slow and your job taken by someone willing to leverage the tools available today.
Which nobody is doing, especially not people who vibe code products. Saying "just prepare for it" as an answer to "what do you do if", is not really enough when that "prepare for it" is very expensive (time, tokens, effort etc.).
For someone to do this, they would have to think for themselves, which I've also not seen much of in the vibe-coding space.
Wait, what is "claude the model"? Anthropic's models are named versions of Opus, Sonnet, and Haiku. Claude, Claude Cowork, and Claude Code are their products which leverage those models. Right?
I assume it will be similar to when a person is out sick or on vacation. Another person on the team likely could take over the work for a day, but realistically it just sits until they're is back.
So work stops until Claude is back? What if Claude comes back and costs 10x the amount? The answer is obviously that you'll "bend over" and pay, because the AI vendor who convinced you that Claude is so great owns you, your codebase, and by extension your company now.
Or you point your Claude code at a different LLM provider. It's not complicated and there are lots of vendors (and in the open-weights space multiple vendors serving the same models competing on price). Sure DeepSeek 4 isn't quite Opus at the moment. But it's plenty good to do the work. We've got different competing front-end tools and different competing back-end providers. No one 'owns' your company. Maybe that will change as the market evolves and one of the frontier tools become so much better than one vendor will own the market. But that's not where we are now.
I didn't realize you could swap out the underlying model used by Claude Code. Aren't all of the Claude tools tied directly to Anthropics models, their authentication and billing, etc?
I have seen many many times in microcontroller forums posts from first timers in the liking of "hello sirs i have problem please show how to do this", followed by their own reply a few hours later asking again because they were holding up, where "this" was usually something really trivial, you just needed to read the docs and the rightful answer was "did you really not try anything in 6 hours?"
What happens when your engineer realizes they can make 10x more at another company? They leave and work stops. You then hire someone else or raise your pay to get better, more reliable engineers. The analogies keep going because AI is a tool, not a replacement. If it's a tool used by a non-technical person, so be it, but it's still just a tool.
Not really, realistically speaking it's now possible to use an agent to read code and make sensible summaries of a codebase faster than ever before, and it's exactly the thing you'd use to onboard yourself or someone else on the team.
The OP was asking what happens to productivity when your LLM is offline, I'd assume it isn't available yo onboard anyone at that time either.
More importantly I think, if devs become dependent enough on LLMs that they just put it aside when the model isn't available, they wouldn't be able to onboard quickly or at all.
It takes experience and a pretty deep understanding of programming in your language of choice to pick up a new code base and quickly understand how it works, the architecture(s) and pattern(s) being used, etc. Those skills would likely have been lost long before a dev simply can't work without the LLM.
AI should enhance your skills. If it's down and your first though is to buy another sub from a different vendor this might be a skill issue. (I'm afraid every day that this will happen to me btw.)
Please resist in future :) The guidelines ask us to avoid internet tropes, precisely because they're repetitive, and more apt to make us groan than smile.
The point is that, with a sufficiently complex setup (with skills, MCPs, prompts, etc.) the difference in AI models will impact the quality of work. You might not care now, but you might care when you have 2 million lines of code and zero idea whats going on.
The point is vendor lock-in. The vibe coding community has reinvented vendor lock-in and is bound to repeat every mistake associated with it.
Pretty much every single detailed prompt made after trial, error, and refinement is tailored to a specific LLM. They will all perform worse used with other LLMs than a similar prompt tailored for the second LLM would perform, and at times quite poorly.
How well would it work to ask the working LLM to rewrite the prompt to get the best results? Do the models understand enough about themselves to do that?
Claude has a /product-self-knowledge skill, and I am sure the others have something similar. So yes, it is possible if you work with care, as necessary with all things LLM related. There are hundreds if not thousands of skills on github that were created just this way.
It's not like you aim to do it, you are just in a feedback loop improving results for the tool you are using. It is inherent in any prompt developed through iteration.
Yes, but that's also a specific luxury I can choose for myself. Definitely a fun and interesting question. At some level of reliance, people would answer "no", but there's the large middle ground (assuming similarly-frontier models are down): having a weaker(?) AI model help you get up to speed ASAP by summarizing code pedagogically, and linearizing the code read order. Basically like an AI-assisted (but manual) code review to reorient yourself.
Just use a fallback, like Codex CLI. Takes a little effort upfront to ensure your configuration is wired correctly for both harnesses, but it is pretty easy to get them 90% identical (there will almost always be some experimental / edge case features that differ across harnesses, but in my experience those are negligible in practice).
I more meant feature-level differences. For instance, Claude Code has agent teams, and Codex CLI does not. Or for a while, Codex had "/goal" and Claude Code did not (though now Claude Code has it too). To your point, it is usually possible to polyfill these gaps either with custom code/skills/hooks or with third party plugins.
We have 3 big competitors in the space: Anthropic, Google and Microsoft. I think they can all use the same base configuration. So it's not that we are out of options here.
if there is 8 hours of downtime (even before AI) I take that opportunity to do other codebase maintenance, debugging, file organization, renaming all the things I said I'd rename or take a break.
pre AI if my IDE was down for whatever reason I wouldn't switch IDE's, I would do something else.
Some agent-written tools and modules are easily the best codebases I've worked with. Documented correctly to the T with various charts and explanations for everything, "start here" guides, concepts defined clearly, and very good Git commit messages.
Naturally you can also have a LLM one-shot a 14000 line PHP monstrosity - it's up to you still, LLM or not.
The main problem is that it'll probably be a waste of time to code anything yourself if Claude is back online in 8 hrs. It's like walking to the next bus stop when you missed your bus - it won't make you get home any sooner.
8 hrs will probably be better spent reading specs or checking things with stakeholders so the next features you let Claude implement are the ones the business actually wants.
In my experience the answer is "no". If I am reviewing some slop and I ask Claude's human babysitter why this class has these constructors, they don't have any idea. Without Claude they don't understand the output at any level.
What happens when you have a codebase made with gcc for let's say 8 hours? Are you able to efficiently, smoothly and productively take over the assembly code?
You can use a local model, which will go down exactly as often as gcc will. We may still have hopeful notions of being able to understand the codebase, but the reality seems to be that the codebases we don't understand will be the ones that will win out in the market, because they'll be cheaper while still only having about as many bugs as they had when people wrote them.
Because you're better able to take over the codebase a local model wrote than one Claude wrote? The original question was about taking over an LLM-written codebase, it doesn't sound to me like the argument was about a codebase that Claude, specifically, wrote.
What does it matter what the codebase is made with? If Claude is down, use Codex, or Gemini, or Deepseek. That version of the argument is just way too easy to counter.
Exactly this. Anti-AI Devs/Techies have their heads in the sand or/and resorting to binary thinking when it comes to AI.
No one is going to vibe code a Photoshop replacement just like no average smartphone user is going to take prize winning photographs with their phone or directly compete with professional photographs.
What is going to happen is what happened to videographers and photographers and what is happening to record musicians: the medium is going to become more accessible by reducing the cost and skill required to make lower quality items.
Just like random selfies don't need you to be a photographer, neither will the one off random app that only your household uses require you to be a programmer.
Making a music video of a trip doesn't require you to know technical knowledge of video recording nor basic music theory. You click buttons and it is done. It won't win prizes but it will be satisfying for the use case it occupies: a one off low scope purpose.
Making tiny one off apps is definitely going to become a thing among people beyond tech and tech adjacent fields. It won't be code clean, it won't be code reviewed or even code versioned but it will be useful and that's what matters ultimately.
> neither will the one off random app that only your household uses
This reminds me a bit of the 2010s idea that every house would have a 3D printer to make one off repairs. Years later, this still seems far out of reach. If anything, it seems to have been settled that most non-technicals don't want a 3D printer.
Vibe coded apps are great, but unless they're hitting an already open API, they're effectively hermetic. There aren't many useful, high quality APIs out there without a companion app these days.
I encourage you to ask members of your household what apps they use which don't connect with any other apps, sites, or companies. I think we'll find the number is pretty low.
In your mind, what are some apps which don't currently exist which would be solving a bespoke household issue that non-techies will be reaching for vibe coding to solve?
I'd be happy to be proven wrong, but I'm just not convinced the puddle is very deep. It's really hard to compare taking a photo with vibe-coding an app.
The problem with that is, like many people found out the hard way, that printing is the easy part and 3D CAD design is much harder.
Many people now have 3D printers to print all kinds of useful tools, though, and there are businesses dedicated to one-off prints for the very occasional repair.
>printing is the easy part and 3D CAD design is much harder.
I'm curious if AI will make that part more accessible. You can ask Gemini to make you a parametric openSCAD model and it can do a pretty good job for most designs I've tried. Then just plug in your measurements, export the stl to your slicer, and print.
3D cad is avoided by having a critical mass of other people putting their cad online. I used https://makerworld.com/models/77668 to replace my one that broke. Easy-peasy.
The people empowered by AI don't have to be nontechies, just everyone who has the will for an app to exist but didn't have the means (like interest) to yak-shave into a software engineer just to build one.
It doesn't mean people who still don't have the interest are suddenly going to build apps.
Also, the idea that there is no more room for apps just because apps already exist is wrong. Incumbent apps would love for you to believe that.
I just vibe-coded my own pedometer app after the most popular steps app on iOS started charging for Duolingo-like "Streak Phrases". The main input was my own interest/energy/attention which is the filter for whether someone will build an app. It uses the iPhone's steps API.
Just because most people don't have the interest/energy/attention to build an app doesn't mean AI hasn't made app-building trivial.
As long as you have to do something, like open a new conversation tab in an AI app, there will always be a filter for the segment of society that will do something.
The puddle for doing some pushups at home isn't very deep yet involves a little bit of time and discomfort. Almost nobody does it despite the upsides. The conclusion you can draw from that is less about the process and more about human disinterest.
Practically speaking it’s rarely cost efficient to vibe code your own all compared to buying. It might cost $100-200 in tokens alone and then there’s the time cost, maintenance, etc. Paying $10/month is cheaper.
I’ll only make an app if there’s nothing else comparable out there (surprisingly common)z
I think there's a use case for LLM's being able to treat websites as API's (I mean, it is an API, really, running on port 80/443) but this is why attestation seems designed to ensure only large companies can do this and not end users.
> Years later, this still seems far out of reach. If anything, it seems to have been settled that most non-technicals don't want a 3D printer.
Considering the 3D printing barrier for entry is buying, setting up, learning, and maintaining a whole new piece of equipment vs going to a website on a phone or computer everyone already has and telling it what you want in plain english, I think the comparison is lacking.
It's really funny to be declaring the idea of 3d printers being ubiquitous as dead because they weren't a smartphone scale change in how the world works, while pretending that 3d printers aren't more common and accessible than ever.
Yeah, a better comparison than the smartphone would be something like a table saw. You don't assume everyone has one, but it's not very unusual to have one in the garage.
> Years later, this still seems far out of reach. If anything, it seems to have been settled that most non-technicals don't want a 3D printer.
They would if you could print things out of durable materials that had weight and structure. I haven't seen any 3D printers that do anything except for that light resin-plastic that feels like you could snap it easily. But if I could print a PVC section for my sink that would totally change the calculus.
You can, in fact, print perfectly well in any thermoplastic, including PVC (although it's unpopular due to toxic fumes). Nor is strength neccessarily an issue. In fact you can 3D print polycarbonate parts strong enough to scratch-build a drone - props and all.
No - the reason you wouldn't want to print parts for your kitchen sink isn't because you can't, it's because you rarely need such parts, and when you do you can simply buy off the shelf parts for next to nothing. A printer simply does not justify its overhead for most people. It's like having a lathe: useful if you're seriously into manufacturing or crafting, but not worth it if you want something pre-designed. There's just not much that it wouldn't be easier to just buy.
How much is a 3D printer that can print durable thermoplastics that I can use for replacement of trivial household items? I thought that would require an industrial setup to do. If you're telling me that I can just start replacing plastic crap in my house, including critical parts like plumbing, with a 3D printer that can sit in its own corner, I probably WILL buy a 3D printer.
A Bambu Lab P1S is $600, and then you gotta buy PET-CF filament, and there's shipping consider as well, so let's say ballpark $800 to start. They make more expensive printers too if you have the budget for it, but the underlying technology is mostly the same. Filament comes in 1kg rolls, and small stuff is like 50 grams, but a large piece of pipe including support structure could be as much as 500g, so if you end up printing a large amount of large items, eg https://makerworld.com/models/1441653 instead of https://makerworld.com/models/77668, which is 2g, and cheaper PLA, filament costs are going to add up.
> They would if you could print things out of durable materials that had weight and structure
You and I totally would, but we're nerds!
Think of how much coaxing it takes to get the average North American homeowner to replace a leaky shower head or a spark plug. A lot of normal happy folks will spend their lives not really learning to fix things much, and that's quite alright, IMO. We don't all need to be good at everything.
I have my doubts, yes there will be tinkerers who build their own apps, but this will be roughly the same crowd who today tinker with home automation, soldering or model trains as a hobby (or as Douglas Adams said: "I am rarely happier than when spending an entire day programming my computer to perform automatically a task that would otherwise take me a good ten seconds to do by hand" - just replace "programming" with "vibecoding").
I don't see 'grandma' building here own calendar app via Claude Code that reminds her of the family birthdays.
Grandma likely isn't able to use most existing web apps beyond facebook, her default email client and little else either.
Uncle Bob on the other hand will stop nagging you to make him those apps you never have the time to make him and will do it himself. He is a handyman, literate and numerate and able to use a computer like most middle age folks outside of tech can. Uncle Bob's mates at the local bar will see the software he wrote and will get into it themselves.
The Gen X+ non-techie population is made up of more than just grandmas.
> I don't see 'grandma' building here own calendar app via Claude Code that reminds her of the family birthdays.
If you think of apps in the traditional sense I think I agree with you, but I have a feeling things are about to become a lot more messy.
Grandma might not even know she's building her own calendar app.
I don't think we are that far from being able to ask a general purpose AI to "help me not forget my family's birthdays" and it creating and maintaining code for that purpose. Not quite an app, but more than a one off script, I think AIs are going to unlock this weird situation where they're running a bunch of barely organized code almost as an extension of thinking.
> Anti-AI Devs/Techies have their heads in the sand or/and resorting to binary thinking when it comes to AI.
The goal posts are being moved, yet again, as the reality of generative AI's usefulness starts to narrow. I think most "anti-AI" devs wanted the technology to be supplemental in the first place, in the hands of responsible engineers. The hype riders are the ones who are saying our job is over.
> reducing the cost
The evidence is the contrary. The tools are become more expensive by the month it seems.
As a more emotions based response to your post: I find it pretty gross that we are ready to accept that this tech should be used in art whatsoever. I think saying this is a barrier-to-entry-lowering tech is a misnomer, because even those who use computers still need to understand the program, mechanics must understand the function and implications of a torque wrench; there is no effort or skill involved with generating slop, you always get a result. Additionally, the first part of your post was to argue that we should be using these tools to do narrow scoped tooling and one-off script, and then you moved to generating videos and music, which shows that you aren't even aware of the "scope" involved in those efforts.
> The evidence is the contrary. The tools are become more expensive by the month it seems.
Maybe, but compare the monthly cost of a ChatGPT subscription to the cost of a face to face CS education, the cost of a dev machine, the cost of spending your time building the software that you want. The subscription easily wins. And yes, yes, overtime the sub is more expensive but the point is that your average layman is not going to front five figure amounts in costs in something like this just like most people don't buy a gym's worth of fitness equipment, they instead go to their local gym and pay monthly. Now think of all the things LLMs cover even if unreliable and low quality and you can see why the fast-food/fast-fashion era of software is upon us. People won't be experiencing the software equivalent of Michelin rate restaurant food or wearing Gucci but they will certainly be having their needs met in a way that they didn't use to before.
Hasn't Anthropic being experiencing issues due to extremely high usage? Being their investor, you would think Amazon wouldn't do Anthropic dirty by weakening their ability to handle user traffic
Your approach is good for catching stuff that human reviewers might miss not as a first line default-only unit. The whole reason this is happening is because humans are not doing their job. Your solution (humans not doing their job) is just increasing the scope of the problem.
Most people don't have cars
reply