A16z partner says that the theory that we’ll vibe code everything is wrong
aol.com
aol.com
(1) https://philippdubach.com/posts/the-saaspocalypse-paradox/
(2) https://philippdubach.com/posts/the-impossible-backhand/
Acharya’s framing is different from mine (he’s talking book on software stocks) but the conclusion is the same: the “innovation bazooka” pointed at rebuilding payroll is a bad allocation of resources. Benedict Evans called me out on LinkedIn for this (https://philippdubach.com/posts/is-ai-really-eating-the-worl...) take, which I take as a sign the argument is landing..
Excellent. And correct lol.
I don't understand this point. Can't it be possible that the ultimate effect is to devalue, hugely, software? As in it can totally both be true that AI capex has weak returns and at the same time most SaaS companies go bankrupt. To take an analogy: if ever we manage to successfully mine asteroids, and find some vast quantity of platinum, it could both be true that every existing platinum miner loses their shirt, and also that the value of platinum sinks so far that the asteroid mining company cannot cover its costs.
Sometimes I just wonder at how HN has no idea what enterprise software involves.
You can always also deploy open source since forever. What happens when it randomly drops logs or changes the text? If you get an alert and it is noise it starts becoming pointless.
And yet these type of stocks were at 50-100x earnings etc.
Or at least that is the theory. It is certainly true from observations of those around me. It also scales well. Even someone a bit innovative gets a multiplier by using AI intelligently. Those that just focus on the grunt work are the ones in trouble.
The next one a16z should walk back on is "AGI" given that they have just admitted that "vibe code everything" was just a sign of them being consumed by the hype.
AI being a next-token predictor will produce cheap and average products, we will likely see some (most?) software become a commodity, that goes through the same product development and "manufacturing" as a breakfast cereal. Made in a "dark factory", 24/7, with little supervision.
However I think down the line we will see many industries popping up that are like "organic food", "mechanical watchmaking" that provide above the usual slop that large businesses produce.
A lot of the AI and LLM argument on whether it is really eating the world misses one point, and I think Evans implied but not pointed out explicitly.
Had it not been AI investment, we wouldn't have the current hardware improvement and innovation rate.
Most people have heard about the limit of Moore's law, but every single time it appeared in headline is an economic model limit rather than limit of physics. We were predicting a stop to growth in 90s because we couldn't see a 400M PC market shipment in 2010. Turns out Smartphone carried that forward, and it is what funded growth of TSMC when most on HN even knew much or heard of TSMC. The same goes with LPDDR RAM, Pure Play IP, Wireless, Network, etc. All the hardware improvements that came with Smartphone is now continued to be developed at rapid pace due to AI and Hyperscaler.
What Evan were suggesting is much simpler, could AI automate things that previously were not possible for 99% of business outside of Tech and Software. The answer is a simple yes. And worth pointing out ChartGPT is closing in on a billion weekly active user.
A lot of HN discussion about AI often centered around software development. And whether it is good enough of it. Most of the world outside are happy enjoying AI for many things. What used to require a mildly technical person to do on excel can not be done without one. It is opening up software to even more people. It is creating more value than people imagine, and users are willing to paid for it.
Using vibe coding to build something to replace an enterprise SaaS offering for a medium to large company is not something to be taken lightly. The tool and the code is not everything. The operating environment, security guarantees, SLAs, support, and a bag of features you don't need today but might tomorrow is what the SaaS offerings bring to the table.
Imagine that I run a really good software house. I can literally build anything you want, feature wise, better than most. I do it quickly. You come to me and say you want to replace Slack for your team of 200, because Slack got too expensive. I say I can do it. Because I am feeling generous and you're my good friend, I will do it for free. However, I will just give you the code, a CI/CD script, and a README.md file. I will disappear and will not maintain or support your software, nor will I give you any guarantees on how well it will work, other than a "trust me."
I wouldn't take the offer.
Right now, you may not be faced with a valid choice to vibe code your own hubspot, but maybe some contractor firm will do it for you and sell an ultra low cost version.
Will opus 6 do the same thing without the contractors while managing the deployment and maintenance as a claw?
I hear Zulip could actually be a real alternative, though I have no experience with it.
We finally made the computer able to speak “our” language - but we still see computers as just automation. There’s a lot of untapped potential in the other direction, in encoding and compressing knowledge IMO.
AI is hugely beneficial in understanding a problem, or at least getting a good overview, so you can then go off and solve/do it yourself, but focusing on "just have the AI generate a solution" is going to hugely harm AI perception/adoption.
To have AI recreate something that was already in it's training set.
> in encoding and compressing knowledge IMO.
I'd rather have the knowledge encoded in a way that doesn't generate hallucinations.
The thing is incredibly good at searching through large spaces of information.
Did it work? Yes. Was it worth my time to maintain and scale the “platform” with the company rather than outsource all that to a CRM company? Not at all.
Time is finite. Spend your time doing what you do best, pay others to do what they do best.
The same thing happens with code; you may get great results from your prompt, but trying to customize it will drive you nuts and you may never get what you want.
Maintenance is another hurdle. How do you maintain code you might not have the skills to maintain?
Vibe-coding may reduce software creation time, but it's not taking over software engineering. The SaaS business is going nowhere. Most people, by far, will continue to rely on someone else for their software needs. But be very aware that the software business will change. We are seeing that already.
The key is that it makes new companies entering the market to compete with Salesforce immensely easier. More competition will just force lower overall margins in SAAS.
This is such an important key insight that will take the vibe coding folks another few years to really internalize.
Given that many engineers have never internalized this, you’re more confident than I am.
But give me the source code for something competitive with Solidworks, Jasper Gold, FL Studio, After Effects, etc. and I'm sure as hell making a business out of it!
Furthermore while good software may not guarantee business success, it is pretty much a requirement. I have seen many projects fail because the software turned out to be the hard part.
We're not going to see the end of software; we're going to see the end of margins.
I don't know for sure, but I suspect that other industries have experienced this. Would love to know which. Photography comes to mind, but I'm sure there are more meaningful examples.
You seem to be implying Salesforce’s business is successful because they have their own query optimizer. But the causality is reversed. Salesforce has their own query optimizer because they’ve built a successful business.
This is largely how things work now; AI may lower the cost and increase margins, but the economics of build vs buy seem the same.
Yes, the vendors want subscriptions and cloud and not owning anything, but customers also don't want to hire people to operate the infrastructure required to run this stuff themselves. That's the whole point of SaaS, and why some companies just run entirely on that model and basically have no in-house IT staff
That AI means you can write and run your own payroll system doesn't mean all of a sudden a world of people with zero technical skills can start doing it on their own
The person with zero technical skills could start learning primitive technology: youtube.com/@primitivetechnology9550 - living off the land really is the job for the unemployed. We've done it before! As long as there is a safe place to actually do it and a planet that can support it. Those'll be the problems.
Basically I expect way smaller companies popping up competing with the big ones and their offering will be priced way lower because their payroll is way smaller.
While there is no competitor, internal tools will pop up now.
Another great example is open source. I think PostgreSQL being free and usable by everyone is a more economic outcome than every Fortune 500 company building their own database engine. Payroll, ERP, and CRM fall into the same category of being commodity software in a lot of cases.
So now consider AI, perhaps the cost of building has decreased from 100k to 10k. What stops a Slack competitor from also building the product for 10k and reselling it at 10% of the cost of Slack? My point is that I don’t see how AI has changed the value prop.
I do not actually believe that you can trivially vibe code a viable slack replacement. But even if one day we could it wouldn’t mean that Slack as a company would just disappear overnight.
They would hang around serving companies who haven’t got the memo yet, or who are locked in a contract, or where the internal political situation is against such a move. The innertia of a bunch of humans behaving like a bunch of humans would provide a sort of “coyote time” effect where the fundamentals could fall out from under Slack yet the company would keep “floating” for a while.
It is funny how much of your question sounds like the old joke where an economist can’t believe their eyes that a $20 bill is laying on the pavement, because surely if it were so someone would have already picked it up. In a steady state the logic might hold up, but we are not in a steady state.
And that is separate from why do I think it is not realistic to just replace slack with vibe coded alternative: just in my company some people use the web interface, some the ios app and some the android app. To be a viable replacement you would need all 3 platforms supported with all features. That sounds in itself a nightmare. Then figuring out what features my company members really use is an other nightmare. There are some who craft custom emojis all the time, some who integrate all kind of weird apps. We various CI and data pipeline processes integrated with slack reporting. And then comes huddle. Video and voice chat and screen sharing. You can even draw on someone else’s screen with it! IT has their needs to archive things (maybe?) or snoop on certain things. Then comes of course interfacing with single-sign-on. I wouldn’t even volunteer to enumerate all the different features people just at my company depend on, let alone offer to replace it.
Also Is there not a self hosted open source solution that companies can host ? That’s easier than ai?
Slack exists in part because ten years ago it was a lot harder for big orgs to make good/modern software.
At the very least, the return is not worth the time and effort.
I think a more likely scenario here is that something good and free escapes containment at some point and Slack’s core product just kind of deflates. Not something better than Slack, but something good enough that people don’t care about Slack any more.
I don’t see it as a question of whether you build it or buy it, but a question of the time horizon for selling messaging software as a business strategy. Most business strategies have a finite time horizon. How long can you continue to sell messaging software before there are too many competing solutions available and you stop making money from it?
Web browsers used to be exclusively paid software, if you were serious. So did operating systems, SQL databases, C++ compilers, and video codecs.
You're too heavily discounting the effort to maintain, deploy, and scale a service like Slack.
I don’t think “categorically different” has legs. I work on the operations & hosting side of applications like this. There’s an operations burden to maintain, deploy, and scale web servers. There’s a burden to rolling out new Linux kernel versions to servers. But we still do it! There would be a burden to running your own Slack, just like there is a burden to running your own email, and people will choose to pay for hosted versions or host it themselves.
And then there are companies and organizations who have strong incentives to self-host to make compliance easier.
Of course it sucked and no one used it except executives and VPs. Everyone else did just enough to meet the minimum quarterly engagement metrics right before performance reviews.
1. They did, and still sell it. You can buy it.
2. They did, and then exited the market. Employees gradually migrated off the internal platform.
3. They weren’t in the business of selling software, and didn’t sell their internal messaging platform (which is idiosyncratic and closely integrated with other internal system).
I can't wait for orgs to try to vibe roll their own dozen clients, security models, and then try to talk to handle external integrations of some kind.
I imagine it's also infinitely better than anything an in house team could vibe code.
You don't need AI for a cheap slack alternative.
That's why I don't buy any of this.
Companies are not bothering with the free/open alternatives.
Unless the real power of LLMs is making it easy for greg in HR to self host these existing alternatives. But, that a trillion dollar market does not make.
This is called transaction cost economics, if anyone’s interested.
Salesforce is getting LLM superpowers at the same time the Enterprise is, so customizing and maintaining and extending Salesforce are all getting cheaper and better and easier for customers, consultants, and Salesforce in parallel.
Unless the LLMs are managing the entire process there’s still a value proposition around liability, focus, feature updates, integrations, etc. Over time that tech should make Salesforce get way cheaper, or, start helping them upsell bigger and badder Sales things that are harder to recreate.
And, big picture, the LLMs are well trained on Salesforce API code. Homegrown “free” versus industry-standard with clear billing, whatever we know versus man-decades of learning at a vendor, months of effort and all the risk & liability versus turnkey with built-in escape goats… at some point you’re paying money not to own, not to learn, not to be distracted, and to have jerks to sue if something goes bad.
With that said, the entire business world does not understand that software is more than just code. Even if you could write code instantly, making enterprise software would still take time, because there are simply so many high-stakes decisions to make, and so much fractal detail.
But why? It was always dramatically cheaper for enterprises to build rather than buy. They stopped doing that becuase they did that in the 90s and ended up with legacy codebases that they didn't know how to maintain. I can't see AI helping with that.
I definitely agree with this.
I can't believe I'm saying this, but I guess you don't even really need to maintain software if it's just a tool you hacked together in a week. You can build v2 in another week. You'll probably want to change it anyways as your users and your org evolve. It's a big question for me how you maintain quality in this model but again, if your quality standard is "good enough", we're already there.
I think that’s gonna happen when you don’t need software and AI just does it all.
If it's not local, I saw this comment: https://news.ycombinator.com/item?id=47085906
"This entire stack could give you computing power equivalent to a 25k euro/month AWS bill for the cost of electricity (same electricity cost as running a few fridges 24/7) plus about 50k euros one-time to set it up (about 4 Mac Studios). And yes, it's redundant, scalable, and even faster (in terms of per-request latency) than standard AWS/GCP cloud bloat. Not only is it cheaper and you own everything, but your app will work faster because all services are local (DB, Redis cache, SSD, etc.) without any VM overhead, shared cores, or noisy neighbours."
Makes me think there will be these prompts like "convert this app to suit a new stack for my hardware for locally-optimized runtime."
How are people building the best local stacks? Will save people a ton of money if done well.
We have! It has different names, shapes and forms. For example WiFi. Last I checked, it’s not designed for human in the loop and you need multiple levels of decoding for a human to make sense of it.
Better and cheaper hardware too. Maybe it'll be DeAI? (decentralized)
Will combine with Crypto 2.0 - whatever that may be.
And the thing about many of these products isn't their feature set, it's their stability. It's their uptime. It's how they handle scaling invisibly and with no effort on your part. These are things you can't just write down from whole cloth, they are properties that emerge over time by adapting the the reality of scale. Coding isn't the whole deal, and your 20x clanker which can do nothing but re-arrange text in interesting patterns is going to have some trouble with the realities of taking that PoC to production. You'll still need experienced, capable people for that. And lots of time.
A lot of this "ermahgerd everything will change" drivel is based on some magical fundamentally new technology emerging in the near future that can do things that LLMs cannot do. But as far as anyone knows, that future may be never.
So even given a large improvement in agentic coding I'm not convinced it really changes the build vs buy equation much.
So I think big SaaS products are under attack from three angles now:
1) People replacing certain systems with 'vibe coded' ones, for either cost/feature/unhappiness with vendor reasons. I actually think this is a bigger threat than people think - there are so many BAD SaaS products out there which cost businesses a fortune in poor features/bugs/performance/uptime, and if the models/agents keep improving the way they have in the last couple of years it's going to be very interesting if some sort of '1000x' engineer in an agent can do crazy impressive stuff.
2) Agents 'replacing' the software. As people have pointed out, just have the agent use APIs to do whatever workflow you want - ping a database and output a report.
3) "Cheap" clones of existing products. A tiny team can now clone a "big" SaaS product very quickly. These guys can provide support/infra/migration assistance and make money at a much lower price point. Even if there is lock in, it makes it harder for SaaS companies to keep price pressure up.
Also, agents are not deterministic. If you use it to analyze data, it will get it right most of the time but, once in a blue moon, it will make shit up, except you can’t tell which time it was. You could make it deterministic by having AI write a tool instead… except you now have the first problem of maintaining a tool.
That isn’t to say that there isn’t small low hanging fruit that AI will replace, but it’s a bit different when you need a real product with support.
At the end of the day, you hire a plumber or use a SaaS not because you can’t do it yourself, but because you don’t want to do it and rather want someone else who is committed to it to handle it.
And regarding agents being non deterministic, if they write a bunch of SQL queries to a file for you, they are deterministic. They can just write "disposable" tools and scripts - not always doing it thru their context.
I think it will be a little different: black box the thing, testable inputs and outputs, and then go to town for a week or two until it is reasonable. Then open source it. Too big/complex for an agent? Break down the black box into reasonable ideas that could comprise it and try again. You can replace many legacy products and just open source the thing. If the customer can leave behind some predatory-priced garbage for a solution where they get the code I think they would be a lot more likely to pay for help managing/setting it up.
It would be interesting if, with all the anxiety about vibe coding becoming the new normal, its only lasting effect is the emergence of smaller B2B companies that quickly razzle dazzle together a bespoke replacement for Concur, SAP, Workday, the crappy company sharepoint - whatever. Reminds me of what people say Palantir is doing, but now supercharged by the AI-driven workflows to stand up the “forward deployed” “solution” even faster.
Or an industry specific Workday, with all of workdays features but aimed at a niche vertical.
I wrote about this (including an approach on how to clone apps with HAR files and agents) if you are interested. https://martinalderson.com/posts/attack-of-the-clones/
Apparently the meal is over and now we're just rearranging the plates.
I do think that I have worked on somewhat complex code bases. The reason why they are complicated is often "political" (e.g. at some point it was decided that this is the way to go, and from then on the specific abstraction was used. It turned out these wishes were not a good idea, but the code was never re-developed with a "more proper" architecture (also for the reason that removing some insanely convoluted feature would anger some users)).
I see no reason why some (hypothetical) AI couldn't come up with a much better architecture (also good programmers are capable of this). The problem is rather "getting this architecture through politically"; for some reason "AI suggested/created it" is much more socially accepted by managers than "programmer X considers this change to be necessary" (I cannot understand why).
side note: indie games are not complex software.
and most "overvalued" and "impossible" and "walking it back now" comments are true in as many cases as they are not true and I really do not understand these commenters. smart people should not fall into the same category as people who think that "nobody cares" because they never met devoted lawyers, investigative journalists and law enforcers passionate about justice AND law. it's all so weird, man ...
> "You have this innovation bazooka with these models. Why would you point it at rebuilding payroll or ERP or CRM," Acharya said
> Instead, companies are better off using AI to develop their core businesses or optimize the remaining 90% of their costs
We work in Software ENGINEERING. Engineering is all about what tools makes sense to solve a specific problem. In some cases, AI tools do show immediate business value (eg. TTS for SDR) and in other cases this is less obvious.
This is all the more reason why learning about AI/ML fundamentals is critical in the same way understanding computer architecture, systems programming, algorithms, and design principles are critical to being a SWE, because then you can make a data-driven judgment on whether an approach works or not.
Given the number of throwaway accounts that commented, it clearly struck a nerve.
For example, being able to vibecode a UI wireframe instead of being blocked for 2 sprints by your UI/UX team or templating an alpha to gauge customer interest in 1 week instead of 1 quarter is a massive operational improvement.
Of course these aren't completed products, but customers in most cases can accept such performance in the short-to-medium term or if it is part of an alpha.
This is why I keep repeating ad nauseum that most decisionmakers don't expect AI to replace jobs. The reality is, professional software engineering is about translating business requirements into tangible products.
It's not the codebase that matters in most cases - it's the requirements and outcomes that do. Like you can refactor and prettify your codebase all you want, but if it isn't directly driving customer revenue or value, then that time could be better spent elsewhere. It's the usecase that your product enables which is why they are purchasing your product.
> The reality is, professional software engineering is about translating business requirements into tangible products.
ai/vibe coding may make that cycle faster but idk it might actually make things worse long-term because now the race course has rubber walls and there is less penalty just bouncing left and right instead of smoothly speeding down the course to the next destination...
> ai/vibe coding may make that cycle faster but idk it might actually make things worse long-term
By making the cycle faster it reduces the impact while also highlighting issues within the process - there are too many incompetent PMs and SWEs.
Additionally, in a lot of cases a PM won't tell you that you might actually be working on checkbox work that someone needs to do but doesn't justify an entire group of 2-3 SWEs because then you obviously won't do the work. This kind of work is ripe for being automated away via vibecoding or agents.
A good reference for this is how close is the feature you are working on directly aligned with revenue generation - if your feature cannot be directly monetized as it's own SKU or as a part of a bundle, you are working on a cost center, and cost centers are what we want to reduce either by automating them away, offshoring them, or doing a mix of both.
The reality is that perfection is the enemy of good, and this requires both Engineers and PMs working together to negotiate on requirements.
If this does not happen at your workplace, you are either working on a cost center feature that doesn't matter, you are viewed as a less relevant employee, or you are working at a bad employer. Either way it is best for you career to leave.
In my experience, if you've actually chatted with executive leadership teams in most F500s, when they are thinking about "AI Safety" they are actually thinking about standard cybersecurity guardrails like zero-trust, identity, authn/z, and API security with an added layer of SLAs around deterministic output.
But by being able to constantly interate and experiment, companies can release features and products faster with better margins - getting a V1 out the door in 1 sprint and spending the rest of the quarter adding guardrails is significantly cheaper than spending 1 quarter building V2 and then spending 1 more quarter building the same guardrails anyhow.
Basically, we're returning to the same norms in the software industry that we had pre-COVID around building for pragmatism instead of for perfection. I saw a severe degradation in the quality of SWEs during and after COVID (too many code monkeys, not enough engineers/architects).
> reality is that perfection is the enemy of good
We don’t like to always admit it but most jobs are fairly straightforward, as in the actual day to day tasks. Yes being smart is great and useful etc. but after a certain point it’s diminishing returns on the actual tasks you have to do. Dealing with other humans and their egos and eccentricities and the multitude ways each person sees the world is always what makes all jobs tricky. I suspect this whole ai wave/hype/reality is going to open many people’s eyes to this. We will laugh that we use to call them “soft” skills.
> AI is already much better at reviewing and even testing
The way NASA does it so that they can trust deliverables from the lowest bidder.
That is, have developers translate the wants/needs into detailed contracts of work.
What this means is that very simple apps will become easy to create quickly. So a todo manager is probably not going to be a very successful business. You’ll be competing with many many people and it will be commoditised.
But ultimately what happens here is the “complexity threshold” of a sufficiently complex product needed to make money will be raised. Existing products will become more sophisticated or, if there is not more “sophistication ladder to climb” then they will be commoditised.
There’s just no way people are going to vibe all their software, that’s a very self absorbed nerd take. But on the supply side we’ll see commoditisation, price drops, and increasingly good value for the user as features are shipped faster.
I also think that software quality is really going to tank, because using validation to test the output of Claude is not a good way to ensure quality or correctness. It’ll get you some of the way but you need powerful reasoning. The most obvious evidence for this is security flaws in AI code. We’ll see a new era of enshitification caused by AI code. Like outsourced manufacturing though, people will buy worse stuff at a cheaper price. That makes me sad, because I thought we were on a path to better software, not buggier software.
All state changes are made with MCP so it saved me from having to spend time on any forms and most interactions other than filtering searching sorting etc.
Means we will be ditching Linear soon.
I know I’m an outlier but this sort of thing will get more common.
Its not just adding features that Linear already provides but adding features and integrations that mets 100% your needs.
The full decision making equation is (cost of implementing it yourself + cost of maintenance + 10% additional benefit for a solution that fully meets your needs) versus (cost of preexisting solution that meets 90% of your needs). Cost of implementing it and cost of maintenance has just gone down. Surely that will mean on a whole more people as a whole will choose to make inhouse rather than outsource.
Thus demand for premade solutions will go down, Saas providers won't be able to increase their prices as this will make even more people choose to implement it themselves. The cost of producing software will continue to drop due to agentic coding and maintenance cost will drop as well due to maintenance coding agents. More people will choose their own custom solutions and so on. Its very possible we are in the beginning of the end for Saas companies.
> server operations, storage, scalability, backups, security, compliance, etc
We have products we're paying $100k a year for and using 3% of the functionality. And they suck. This is the target.
make me my own Stripe
make me my own Salesforce
make me my own Shopify
It will be more like:
Look at how Lago, an open-source Stripe layer, works and make it work with Authorized.net directly
Look at Twenty, an open-source CRM, and make it work in our tech stack for our sales needs
Look at how Medusa, an open-source e-commerce platform, works and what features we would need and bring into our website
When doing the latter, getting a good enough alternative will reduce the need for commercial SaaS. On top of that, these commercial SaaS are bloated with features in their attempt to work with as many use cases as possible and configuring them is “coding” by another name. Throw in Enshittification and the above seems to the next logical move by companies looking to move off these apps.
Velocity or one-shot capability isn't the move. It's making stuff that used to be traumatic just...normal now.
Google fucking vibe-coded their x86 -> ARM ISA changeover. It never would have been done without agents. Not like "google did it X% faster." Google would have let that sit forever because the labor economics of the problem were backwards.
That doesn't MATTER anymore. If you have some scratch, some halfway decent engineers, and a clear idea, you can build stuff that was just infeasible or impossible. all it takes is time and care.
Some people have figured this out and are moving now.
LLMs don't have attention to detail.
This project had extremely comprehensive, easily verifiable, tests.
So the LLM could be as sloppy as they usually arez they just had to keep redoing their work until the code actually worked.
I linked the paper! I read the paper. Yeah. they wrote the tests, which is how this worked! how the heck do you think it was supposed to work?
the fact that they needed to write the tests was just the means to implementation. It didn't change the non-LLM labor economics of the problem.
You've unfortunately committed the big sin of anthropomorphizing the LLM and calling it diligent.
An LLM cannot be diligant, it's stochastic so it's literally impossible for it to be diligant.
Writing all those tests was diligant.
I’m not worried about the personal character of diligence. I’m interested in what the technology unlocked and how things made with it are materially different in terms of labor configurations.
Fadda forgive me!
C’mon pal.
"Did you check?" is the most expensive question, and one of the most feared in my experience in tech circles. Spent quite a few years as a dedicated tester once I developed the knack for it. Everybody gangsta til it's time to prove the damn thing works.
Here it’s a climb-down. Writing tests to validate translation is orders of magnitude less work (and less likely to fail or be too dull to do properly) than the alternative available prior to 2025.
The fact that they wrote tests and clearly established operational control is not some kinda gotcha! It’s how they managed to get this piece of technology to allow them to do the IMPOSSIBLE.
I am just really struggling to understand someone who reads that paper and thinks “yup, everything is still the same and we don’t need to re-evaluate any ideas” Like, if you want to say that they still need engineering discipline in order to do ISA transitions at scale then…ok? That’s true? But Gemini meant that this formerly impossible thing was now not only within reach but done.
This has been how all previous innovations that made software easier to make turned out.
People found more and more uses for software and that does seem to be playing out again.
https://arxiv.org/abs/2510.14928
Just read some of that. It's not long. This IS NOT the past telescoping into the future. Some new shit is afoot.
First, a vendor will have the best context on the inner workings and best practices of extending the current state of their software. The pressure on vendors to make this accessible and digestable to agents/ LLMs will increase, though.
Secondly, if you have coded with LLM assistance (not vibe coding), you will have experienced the limited ability of one shot stochastic approaches to build out well architected solutions that go beyond immediate functionality encapsulated in a prompt.
Thirdly, as the article mentions, opportunity cost will never make this a favorable term - unless the SaaS vendor was extorting prices before. The direct cost of mental overhead and time of an internal team member to hand-hold an agent/ write specs/ debug/ firefight some LLM assisted/ vibe coded solution will not outweigh the upside potential of expanding your core business unless you're a stagnant enterprise product on life support.
Most of us* are working for places whose analytics software transitively asks the user for permission to be tracked by more "trusted" partners than the number of people in a typical high school, which transitively includes more bytes of code than the total size of DOOM including assets, with a performance hit so bad that it would be an improvement for everyone if the visitor remote desktop-ed into a VM running Win95 on the server.
And people were complaining about how wasteful software was when Win95 was new.
* Possibly an exaggeration, I don't know what business software is like; but websites and, in my experience at least, mobile apps do this.
They invested in ERP/CRM? I built one (fairly complete to the German/Italy/EU tax system) and it saves a ton of money vs commercial offerings. So yeah, of course we will.
To illustrate, I'll share what I'm working on now. My companies ops guy vibe coded a bunch of scripts to manage deployments. On the surface, they appear to do the correct thing. Except they don't. The tag for the Docker image used is hardcoded in a yaml file and doesn't get updated anywhere unless you do it manually. The docs don't even mention half of the necessary scripts/commands or implicit setup necessary for any of it to work in the first place, much less the tags or how any of it actually works. There are two completely different deployment strategies (direct to VM with docker + GCP and a GKE-based K8s deploy). Neither fully work, and only one has any documentation at all (and that documentation is completely vibed, so has very low information density). The only reason I'm able to use this pile of garbage at all is because I already know how all of the independent pieces function and can piece it together, but that's after wasting several hours of "why the fuck aren't my changes having an effect." There are very, very few lines of code that don't matter in well architected systems, but many that don't in vibed systems. We already have huge problems with overcomplicated crap made exclusively by humans, that's been hard enough to manage.
Vibe coding consistently gives the illusion of progress by fixing an immediate problem at the expense of piling on crap that obscures what's actually going on and often breaks exiting functionality. It's frankly not sustainable.
That being said, I've gotten some utility out of vibe coding tools, but it mostly just saves me some mental effort of writing boring shit that isn't interesting, innovative, or enjoyable, which is like 20% of mental effort and 5% of my actual work. I'm not even going to get started on the context switching costs. It makes my ADHD feel happy but I'm confident I'm less productive because of the secondary effects.
A few times a week I’m finding open source projects that either have a bunch of old issues and pull requests, or unfinished todos/roadmaps, and just blasting through all of that and leaving a PR for the maintainer while I use the fork. All tested, all clean best practice style code.
Don’t complain about the outputs of these tools, use the tools to produce good outputs.
I’ve had a similar experience to what you’re describing. We are slower with AI… for now. Lean into it. Exploit the fact that you can now iterate much faster. Solve smaller problems. Solve them completely. Move on.
> (...) ops (...) a bunch of scripts to manage deployments.
Devops is prime example of work to be minimized and ultimately eliminated entirely by automation. Yes, it's a complex domain rich in challenges and there's both art and skill to do it right, but at the same time, it's also not the thing we want, just the thing we have to do to get the thing we want, because we can't yet do better.
Your model doesn't improve. It can't.
Hilarious that you start with that as TAO requires
- Continuous adaptation makes it challenging to track performance changes and troubleshoot issues effectively.
- Advanced monitoring tools and sophisticated logging systems become essential to identify and address issues promptly.
- Adaptive models could inadvertently reinforce biases present in their initial training data or in ongoing feedback.
- Ethical oversight and regular audits are crucial to ensure fairness, transparency, and accountability.
Not much magic in there if it requires good old human oversight every step of the way, is there?
Of course it needs human supervision, see IBM 1979. Oversight however doesn’t mean the robots wait for approvals doing r&d and that’s where the magic is - the magic being robots overseeing their training and improvement of their harnesses.
IOW only the ethics and deployment decisions need to be gated by human decisions. The rest is just chugging along 1% a month, 1% a week, 1% a day…
You absolutely could print things like cups, soap holders, picture frames, the small shovel you use for gardening, and so on an so on.
99% of people still just buy this stuff.
- JIRA/trello/monday.com - benchling - obsidian
this is what i buy and have no intent to replace:
- carta - docusign - gusto/rippling - bank
this is what might be on the chopping block:
- gsuite
plus, being able to crossref internal data types is chef's kiss.
im paying for claude pro so it's use it or lose it. when i finish everything and have it battle tested i can end my claude code. and anyways when i have 10 employees, it's parity.
for gsuite: i want to own everything internally eventually ans having internal xrefs will be nice. the gsuite data is incidental, what is truly valuable about gsuite is spam detection and the oauth capability
If AI gets good enough to replace all human labor then actual physical moats to keep the hungry, rioting replaced humans away will be the most important moats.
If AI renders human labor a cheap commodity (say you can orchestrate a bunch of agents to develop + market a Windows competitor for $1000 of compute), what used to be "Satya + his army vs. you" now becomes mostly a 1:1 fair fight, which favors the startup.
Everyone who founds a company needs to remember that they're building a system of systems that all interact and influence each other, and you have to balance short term cash flow against long term strategy.
It's not uncommon that this maintainer actually wants to get away from maintaining this code, and would actually be quite willing to teach some successor how everything works.
The problem is that it is often hard to find someone who is similarly passionate about this system (often the system only keeps working because the original maintainer invests a lot of energy into keeping it alive), and is thus brutally willing to learn this system inside out. You can't force this mentality from above: either a suitable programmer has this mentality or he is typically not suitable.
I've certainly inherited and also caused these problems in my younger years.
People running businesses want to focus on their core business and are happy to pay for pain points to go away, for money to come in or less money to come out. It's that simple.
I think GenAI opens Pandora’s box and all of these decisions change.
Theres a whole lot between "I waste my time and do non-scalable things" and "I need an enterprise contract as a SMB".
> Shitty python script with DuckDB running locally on laptop, can get you long long way.
Yeah ... no. You do you, but if you want to get your business to a good size you need proper tools. Like a chef needs a good knive, your business needs good tools as well. That's not to say that a quick DIY integration is off the table, but the one thing you don't get back is time, not money. So when you waste time by doing shit that can go away with minimal investment, you're preventing your business from getting traction.
Sorry but that's a coder mindset. "I can do this myself". That's what's keeping you stuck. You need to think about every decision as an investment. Time vs money. What's the ROI for all alternatives? Then you decide on what moves the needle. It's that easy.
The calculation is simple: every minute you do stupid stuff like coding shit, you can't sell to someone paying you to make things. So you lose twice by doing something yourself: a) you can't sell your services or time in the mean time and b) you add complexity and maintenance to your very complex business life as well. For me, the biggest problem is too much to do and not enough time. So to scale my business I need to reduce complexity. If I can throw reasonable money for that to go away: good, I'll do it. Most business owners I know think like that. That's business, not sales.
There are so many good SaaS for cheap or even with a good free tier that get you started that it's just incredibly junior to just build it yourself.
AI systems are also prone to writing code which they can't effectively refactor themselves, implying that many of these code bases are fiscal time bombs where human experts are required to come fix them. If the service being replaced has transactional behaviour, does the AI produced solution? Does the person using it know what that means?
The other side is that AI as an industry still needs to recoup trillions in investment, and enterprise users are potential whales for that. Good prices in AI systems today are not guaranteed to last because even with hardware improvements these systems need to make money back that has been invested in them.
it is not needed to automate everything. some joys should not be automated away, people wont let them be either way.
the world could be much more optimal in .any place, but its boring, so the optimisations go elsewhere.
I don't even know what this means, but my take: we should stop listening to VCs (especially those like A16Z) who have an obvious vested interest that doesn't match the rest of society. Granting these people an audience is totally unwarranted; nobody but other tech bros said "we will vibe code everything" in the first place. Best case scenario: they all go to the same exclusive conference, get the branded conference technical vest and that's were the asteroid hits.
it's not just "replace snowflake", there are a lot of times i wish i could build a very focused thing to accelerate some of our internal workflows and the nocode solutions either were too simplistic that you ended up spending just as much time trying to wrangle some generic solution to your own use case. OR it was not worth throwing significant engineering resources behind internal ops stuff. now that barrier is dropping fast and it's feasible for us.
whoever can create the framework/tooling for people build their own systems will win this, but i don't think it's something that can be "productized" like a saas.
You don’t understand what’s happening if you dismiss the leverage provided by AI as “vibe coding”.
I really hate the expression "the new normal", because it sort of smuggles in the assumption that there exists such thing as "normal". It always felt like one of those truisms that people say to exploit emotions like "in these trying times" or "no one wants to work anymore".
But I really do think that vibe coding is the "new normal". These tools are already extremely useful, to a point where I don't really think we'll be able to go back. These tools are getting good enough that it's getting to a point where you have to use them. This might sound like I'm supportive of this, and I guess am to some extent, but I find it to be exceedingly disappointing because writing software isn't fun anymore.
One of my most upvoted comments on HN talks about how I don't enjoy programming, but instead I enjoy problem solving. This was written before I was aware of vibe coding stuff, and I think I was wrong. I guess I actually did enjoy the process of writing the code, instead of just delegating my work to a virtual intern while I just watch the AI do the fun stuff.
A very small part of me is kind of hoping that once AI has to be priced at "not losing money on every call" levels that I'll be forced to actually think about this stuff again.
I think I would still kind of ask the same questions, though maybe a bit more conceptual. Like, for example, I might see if I could get someone to explain how to build something, and then ask them about data structures that might be useful (e.g. removing a lock by making an append-only structure). I find that Codex will generally generate something that "works" but without an understanding data structures and algorithms, its implementation will still be somewhat sub-optimal, meaning that understanding the fundamentals has value, at least for now.
| The virtue of formal texts is that their manipulations, in order to be legitimate, need to satisfy only a few simple rules; they are, when you come to think of it, an amazingly effective tool for ruling out all sorts of nonsense that, when we use our native tongues, are almost impossible to avoid.
- Dijkstra
- Had a boss give you confusing instructions?
- Argued with someone only to find you agree?
- Talked with someone and one of you doesn't actually understand the other?
- Talked with someone and the other person seems batshit insane but they also seem to have avoided a mental asylum?
- Use different words to describe the same thing?
- When standing next to someone and looking at the same thing?
- Adapted your message so you "talk to your audience"?
- Ever read/wrote something on the internet? (where "everyone" is the audience)
Vibe Coding has its uses and I'm sure that'll expand, but the idea of it replacing domain experts is outright laughable. You can't get it to resolve ambiguity if you aren't aware of the ambiguity. If you've ever argued with the LLM take a step back and ask yourself, is there ambiguity? It'll help you resolve the problem and make you recognize the limits. I mean just look at the legal system, that is probably one of the most serious efforts to create formalization in natural language and we still need lawyers and judges to sit around and argue all day about all the ambiguity that remains.
I seriously can't comprehend how on a site who's primary users are programmers this is an argument. If we somehow missed this in our education (formal or self) then how do we not intuit it from our everyday interactions?
[0] https://www.cs.utexas.edu/~EWD/transcriptions/EWD06xx/EWD667...
[1] https://en.wikipedia.org/wiki/Principia_Mathematica
[2] Most programming languages are some hybrid variant. e.g. Python uses duck typing: if it looks like a float, operates like a float, and works as a float, then it is probably a float. Or another example even is C, what used to be called a "high level programming language" (so is Python a celestial language?). Give up some precision/lack of ambiguity for ease.
> we developed the languages of Math and Programming for a reason
Sometimes I wonder why companies are so resistant to making profits. It can be really strange. To be so profit focused yet throw away so much just because it is a bit more effort or a bit slower. But I guess most people are penny wise and pound foolish.
> It's extra sad because they would be more profitable if they recognized this.
I don't think that's the argument. The argument I'm seeing most is that most of us SWEs will become obsolete once the agentic tools become good enough to allow domain experts to fully iterate on solutions on their own.
That’s been the argument since the 5PL movement in the 80s. What we discover is that domain expertise an articulation of domain expertise into systems are two orthogonal skills that occasionally develop in the same person but, in general, requires distinct specialization.
A Lotus 1-2-3 vibecoded by a Product Manager in 1979 would probably had a hotkey for a calculator.
> The argument I'm seeing most is that most of us SWEs will become obsolete
Though my big concern right now is that we'll get rid of juniors and maybe even mid levels. There's definitely a push for that and incentives from an economic point of view. But it will be disastrous for the tech industry if this happens. It kills the pipeline. There can be no wizards without noobs. So we have a real life tragedy of the commons situation staring us in the face. I'm pretty sure we know what choices will be made, but I hope we can recognize that there's going to need to be cooperation to solve this least we all suffer.
I believe that business analysis and computer science are essential domains, of (respectively) properly defining the logical/mathematical constraints of problems, and of understanding what it means to solve such problems. But software engineering is "accidental", in the sense that no one other than us in the industry actually cares about what we do, but rather only about their use-cases implemented; what we do is just implementation details.
I don't have a magic ball, but I wouldn't be surprised if in a century, (human) software engineering goes the way of the switchboard operator, elevator operator or knocker-upper.
He's not wrong about the problems of natural language YET HERE ARE. That would, I think, cause a sensible engineer to start poking at the predicate instead of announcing that the foregone conclusion is near.
We should take seriously the possibility that this isn't going to be in a retrenchment which bestows a nice little atta boy sticker on all the folks who said I told you so.
> Dijkstra also said no one should be debugging
Interestingly I think his "gospel" is only more meaningful today.
| Apparently, many programmers derive the major part of their intellectual satisfaction and professional excitement from not quite understanding what they are doing. In this streamlined age, one of our most under-nourished psychological needs is the craving for Black Magic, and apparently the automatic computer can satisfy this need for the professional software engineers, who are secretly enthralled by the gigantic risks they take in their daring irresponsibility. They revel in the puzzles posed by the task of debugging. They defend —by appealing to all sorts of supposed Laws of Nature— the right of existence of their program bugs, because they are so attached to them: without the bugs, they feel, programming would no longer be what is used to be! (In the latter feeling I think —if I may say so— that they are quite correct.)
| A program can be regarded as an (abstract) mechanism embodying as such the design of all computations that can possibly be evoked by it. How do we convince ourselves that this design is correct, i.e. that all these computations will display the desired properties? A naive answer to this question is "Well, try them all.", but this answer is too naive, because even for a simple program on the fastest machine such an experiment is apt to take millions of years. So, exhaustive testing is absolutely out of the question.
| But as long as we regard the mechanism as a black box, testing is the only thing we can do. The unescapable conclusion is that we cannot afford to regard the mechanism as a black box
https://www.cs.utexas.edu/~EWD/transcriptions/EWD02xx/EWD288...
He literally said those exact words out loud from the audience during a job talk.
And yeah, the total aim and the reason why he might just blurt that out is because a lot of the frustration and esprit de corps of programming is held up in writing software that's more a guess about behavior than something provably correct. Perhaps we all ought to be writing provably correct software and never debugging as a result. We don't. But perhaps we ought to. We don't.
Is control via natural language a doomed effort? Perhaps, but I'd be cautious rather than confident about predicting that.
> He literally said those exact words out loud from the audience during a job talk.
Unfortunately despite being able to provide a summary I'm unable to actually read it for you. You'll actually need to read the whole thing and interpret it. You have a big leg up with my summary but being literate or not is up to you. As for me, I'm not going to argue with someone who chooses not to read
Which is what I was referring to. I read what you wrote, pal. Did you read what I wrote?
> I sincerely doubt you produced the source
> I read what you wrote, pal.
We're basically in agreement, but you want to act like you're teaching me something. It's irritating.
One can fully understand that the goal is to write provable programs and yet we do not, we write programs that need debugging. So therefore, I don't think it's hard to imagine that if we get along with that, we may get along with natural language in the control channel, despite that being also proscribed in that vaunted essay you linked to me.
Vcs are choke full of companies that can be cloned over night, SaaS companies that will face ridiculously fast substitution, and a whoooole lotta capital deployed on lousy RAGs and OpenAI Wrappers.
This is your regular reminder that
1) a16z is one the largest backers of LLMs
2) They named one of the two authors of the Fascist Manifesto their patron saint
3) AI systems are built to function in ways that degrade and are likely to destroy our crucial civic institutions. (Quoted from Professor Woodrow Hartzog "How AI Destroys Institutions"). Or to put it another way, being plausible but slightly wrong and un-auditable—at scale—is the killer feature of LLMs and this combination of properties makes it an essentially fascist technology meaning it is well suited to centralizing authority, eliminating checks on that authority and advancing an anti-science agenda (quoted from the A plausible, scalable and slightly wrong black box: why large language models are a fascist technology that cannot be redeemed post).
Seems like he's focused on fintech and not involved in many of their LLM investments
> ...
> For this WE WANT:
> On the political problem:
> Universal suffrage by regional list voting, with proportional representation, voting and eligibility for women.
> ...
> On the social problem:
> WE WANT:
> The prompt enactment of a state law enshrining the legal eight-hour workday for all jobs.
> ...
> On the military issue:
> WE WANT:
> The establishment of a national militia with brief educational services and exclusively defensive duty.
> ...
> On the financial problem:
> WE WANT:
> A strong extraordinary tax on capital of a progressive nature, having the form of true PARTIAL EXPROPRIATION of all wealth.
> ...
0: https://it.wikipedia.org/wiki/Programma_di_San_Sepolcro#Test...
The message of universal suffrage vs. that of preventing an out group from “stealing” an election are not far apart semantically. Same with workers rights - in practice the worker protection laws that were passed in Italy at this time were so full of loopholes and qualifications that ultimately the workers do not gain power in that system.
It is this fair, in my view, to question the spirit of the manifesto in the first place.
Or maybe they own the debt.
Listen to some of the Marc Andreessen interviews promoting cryptocurrency in 2021.
Do that and you will never listen to him or his associates again.
Machine assisted rigorous software engineering is an even bigger wealth transfer from unscrupulous people to passionate computer scientists.
Most SaaS companies are just expensive wrappers on top of existing tools. For non-VC-funded companies, SaaS tools are a serious cost. If you can re-create them in-house with AI, why wouldn't you? The result is saving capital (which you can then employ to do the more innovative things), and being in control over your own data.