Do you feel the product managers are the people pushing the hardest for LLMs and Vibe coding?
Posted by Friendly-Nobody8023@reddit | ExperiencedDevs | View on Reddit | 109 comments
I have read a few books on product management in the past. And a recurring theme in all those books is a strong emphasis on experimentation. Collecting data. Making data-driven decisions. Doing A/B testing and everything. All those books made product management seem like a scientific endeavor. After I read those books, I developed deep appreciation and respect for product management.
And literally data is a PM's bread and butter. And if they don't have data they can't make any decisions. If they are not making any decisions they are just chasing the engineers around. Chasing the marketing folks around. At that point they are just a glorified project managers.
And the problem with collecting the data is that you actually need to build something and put it in front of money customers most of the time to see what they think. You can do surveys and stuff. Call your customers up and ask questions. Talk to the folks in the marketing department and all. But at the end of the day nothing beats having a working prototype that you can show customers to get their feedback. And that is what gets these product managers excited really excited about. Really fast prototyping. A chance to cut off the middle man AKA software engineer and run the experiments themselves.
And I follow so many product managers on LinkedIn. Almost every single one of those people is posting about LLMs and vibecoding incessantly. And I have actually friends on LinkedIn who have a Ph.D. in ML/AI. Even they don't post about LLMs and AI agents as much as these product managers.
A couple of days ago, one of those really famous product managers was saying that Google is including a Vibe coding round in the Product manager interview rounds. And the funny thing is that a director of product management from Google commented on his post and said that this was not true. That whole episode made me feel that these product managers are so desperate for this LLM thing to take off.
They seem to have this fantasy that if LLMs really take off, they will get to do both product management and software engineering? And that all the software engineering jobs will go to them as well.
I feel most of these product managers don't seem to understand the difference between building prototypes and building production ready code. Some product manager on LinkedIn was saying that he built a Chess Engine using Claude. I told him "Great, why don't you host it online. Try to add features to it. Try to do A/B testing with it. Try to fix bugs. Try to incorporate user feedback into it, while ensuring nothing breaks. Tell me how long it took before you gave up and started manually coding again". I got so many upvotes on that comment. This dude never replied back.
Do you folks think that all of this enthusiasm about LLMs and vibecoding comes mostly from Product managers and not actual software engineers or researchers and scientists?
Eliarece@reddit
The goal of a product manager in most companies is to deliver as many features as possible with as little cost as possible. This makes LLMs perfect.
The issue being that they usually aren't the one responsible for fixing the mess.
Suspicious-Letter459@reddit
reminds me of the no-code movement hype
CodelinesNL@reddit
There's always a lot of "shit" in hypes. Low-code and blockchains were great examples that turned out to be mostly pointless.
But "cars", "planes", "electricity", "the internet" and "the cloud" were similar; many people promising every single of our problems would be solved. Of course they were wrong. But these are also "things" that greatly influenced the future.
lol_wut12@reddit
and they actually work, unlike AI
CodelinesNL@reddit
If you are still claiming "AI doesn't work" in 2026, you're going to be in for a rude awakening in 2027.
lol_wut12@reddit
yeah yeah you guys have been saying "in 6 months" for the last 5 years
asarathy@reddit
So much this. Support, maintenance etc are not their concerns. They only things that are are security issues that can cause reputational harm (but only after it happens), stablity and speed of delivery of change. They haven't seen how vibe coded stuff has affected the other 2 things yet so can pretend it doesn't exist or will just be solved by more vibe coding. Which who knows maybe it will sooner than later.
bonkosaurusluke@reddit
Product Manager with 6 years experience. This has been the reality in my entire product career. The role that's sold on Linkedin by the PM influencers is a myth.
My ability to make decisions on what to build based on data and analytics means nothing to a hippo that can use their authority to override priorities based on feels. Every product job i had started as "strategy" then moved into execution work like backlog management and running Agile teams.
I'm ok with this, but framing the experience accurately in any Product Manager interview is an automatic no, even though that's the actual problem they are trying to solve.
CodelinesNL@reddit
In our case product management and their ability to make decisions is becoming the new bottleneck, now that "writing code" is much less of one. So we're pushing back on that, putting them back into their role of figuring out what does and doesn't work.
If vibe-coding prototypes helps them with that; great. We will just take what works and implement it ourselves.
igharios@reddit
and one of their challenges is the limitations of people.
capacity - not enough people
capability - bad engineers, building "inflexible" products - we all worked with at least 1
PM_ME_UR_PIKACHU@reddit
This is my engineering groups problem right now as we try to convert a myriad of shitty written ai prds that make no sense compared to our stack and then product both sets the due date and gets mad when we have to make changes to create actual engineering plans. If your engineering is lead by product run teams, run away as fast as you can.
pydry@reddit
The enthusiasm seems to be inversely correlated to programming ability.
A lot of not very good developers are very keen on it also.
Smallpaul@reddit
This is the in inventor of Rust:
https://graydon2.dreamwidth.org/322732.html
Someone of substantial programming ability.
pydry@reddit
Smallpaul@reddit
If you were honest you would quote the next sentence as well.
pydry@reddit
If you were honest you'd have acknowledged that sentence.
Smallpaul@reddit
What did I say to deny the existence of the sentence?
vibes000111@reddit
This only makes sense if you ignore all of the top engineers who are excited about it. A very public example is someone like John Carmack - clearly a brilliant engineer, very excited about what's happening with agentic coding. Not that you need to look at engineers who are that extraordinary or that public to see that your logic is just wrong.
Smallpaul@reddit
Why didn’t you link to evidence of your closet that John Carmack is excited?
pydry@reddit
I havent seen anything to indicate that Carmack has gone hard on vibe coding or that he warrants that it works fantastically well.
All ive seen is one tweet saying "dont be anti progress" to an engineer.
CodelinesNL@reddit
There is a big difference beween some product manager vibe coding and experienced engineers using these tools. They are extremely powerful, especially in the hands of experienced devs.
It's mostly junior devs, or developers who need fully specced stories that are completely fucked. But this applies to most office jobs anyway. HR? Legal? Ivory tower architects? Office management? Most of the people in these roles will have to adapt the coming 1-2 years, or be royally shafted.
pydry@reddit
There is typically a big skill difference between somebody who is unimpressed with agentic coding and somebody who thinks it's a game changer that everybody needs to adapt to.
CodelinesNL@reddit
If "I'm smarter because I am unimpressed by AI" is your position; by all means 😉
In my experience it's mostly stubborn/afraid developers who haven't bothered getting a 20-euros a month pro Claude Code license and actually started building stuff that have this stance.
Like I said; I'm more than happy to actually explain to people what does and doesn't work. Or keep pretending it doesn't; either way is fine.
And I do get that most of us want to learn only on the job. But if your job is shafting you with the wrong tooling, I'd urge you to consider spending some personal time to at least somewhat stay with the times.
upsidedownshaggy@reddit
I guess I wanna know what kinds of problems advocates like yourself are solving with AI coding tools/agents because I've found they excel in some areas and are extremely lackluster in others, I assume because of the data sets they were trained off of. E.g. everyone and their mom can spit out a professional looking React portfolio now because there's thousands and thousands of hours of tutorials for that kind of thing, but I couldn't get Copilot with Claude Sonnet to make me a functioning third person camera script in Godot a few months ago.
I'm trying to catch up on the tooling as my work is now finally adopting AI tooling for developers, so I genuinely wanna know.
CodelinesNL@reddit
The current models are reasoning models. The notion that they're good or bad at certain codebases/lanugages holds no water.
Copilot is absolute garbage. The harness is just as important as the model. I have created a simple data analysis agent that uses Sonnet underneath, but that can't do with CC does, simply because it misses the extensive harness that CC provides.
Primarily; use Claude Code together with OpenSpec, and get some experience using that workflow building stuff. Work on narrow slices using that workflow.
What I see it struggle most is codebases with a lot of technical debt, and conflicting information. But it's actually better at at least identifying that, than a typical dev is.
Unfair-Sleep-3022@reddit
Can you share where he showed such excitement?
Wonderful-Habit-139@reddit
Yes Mr vibes, we know you like vibecoding.
vibes000111@reddit
Wasn't even on my mind when I chose the account name.
Wonderful-Habit-139@reddit
And the biggest issue is that they're the most arrogant about it. They're so quick to accuse you of skill issues and not knowing how to prompt.
shokolokobangoshey@reddit
One of my closest friends is a former Meta PM who is mainlining the kool-aide. I mentioned picking up a new language for a thing I’m experimenting with for work, and he frowned “Why are you learning to write new code in 2026”. Ok bro lol
dagamer34@reddit
How long until people realize that if you can’t do it yourself by hand, the quality of your AI-generated code will suffer?
Unfair-Sleep-3022@reddit
It's like they're indoctrinated, isn't it? I notice very clear patterns in how they talk
sosumi17@reddit
I call vibe coding “the revenge of the mediocre developer”. It gave voice to a lot of developers who were mediocre in their whole careers and now found a way to cover their mediocrity.
ninetofivedev@reddit
What’s your definition of a mediocre developer and what makes you better than them?
sosumi17@reddit
To me it’s people who don’t care about the code or the quality of it. They were not willing to learn or to upskill, they joined software engineering because of the hype and how easy it was during certain times. Those people would be out of the market quite soon because after a point they would be considered redundant. However vibe coding gave them a “free pass”. Now they don’t need to upskill, they can be the same mediocre and hide their mediocrity behind Claude or any other agentic tool.
Regal_Kiwi@reddit
Not only a voice, but it was already hard trying to sustain a minimum of code quality in a product, now people can summon 3x the amount of code in the same time.
Secret_Jackfruit256@reddit
I've seen deep write ups about exactly this on hacker news, and in some of they, people were even bragging about being able to do that.
ninetofivedev@reddit
There is also a lot of not very good developers who think they’re above it.
Honey, the LLM writes better quality code than you do.
pydry@reddit
I was waiting for somebody who fitted the criteria to announce themselves this way.
ninetofivedev@reddit
I wasn’t talking about you, to be clear.
CodelinesNL@reddit
I agree, to some extent. I am very enthusiastic about Claude Code specifically, but it removes a lot of the "bottleneck" in trying out different approaches to see what works. Proving an approach with a proof of concept is MUCH cheaper now.
But at the same time product managers suffering from Dunning Kruger think they can completely replace devs, and lazy / bad devs use it as a crutch to seem productive.
Without sound technical leadership that understands the dangers but can also set up a good development process using AI, I can imagine a lot of companies are going to see their codebases turn to shit.
chickadee-guy@reddit
There is no learning curve to use AI. Prompt and read the slop. No, using Skills and MCP in claude code is not a skill. No, "prompt engineering" is not a thing
CodelinesNL@reddit
I didn't say there's a learning curve? the "behind the curve" is not a reference to a learning curve, but an experience curve. Knowing what the most recent models can and cannot do, and how to best use them. Your "read the slop" is an indication you're behind that curve too.
Yes Claude gets stuff wrong sometimes. But using it the right way it absolutely does not "produce slop".
How to use it well; I taught our devs in an hour. Knowing what does and doesn't work takes 1-2 weeks tops.
Prompt engineering is dead, MCPs are inefficient (use CLIs) and creating skills is something Claude is really good at. "Hey we keep doing X, please create a skill for it". That's it.
chickadee-guy@reddit
The latest models are a full on regression from prior models. There is no curve. Youre a scammer hawking vaporware.
CodelinesNL@reddit
It takes a very specific type of ego to respond in such a fashion to someone who has a different view on something than you do.
vanit@reddit
I've noticed it also correlates with how hands off you've become. I've been so disappointed to see some staff engineers and EMs particularly taken with it.
pydry@reddit
I find the people in high level technical positions who dont code actually tend to be the least competent technically.
They lose the feedback loop that keeps them competent.
Unfair-Sleep-3022@reddit
I'm so tired of managers and VPs which clearly haven't written a line in decades glazing this bullshit
ExperiencedDevs-ModTeam@reddit
Rule 9: No Low Effort Posts, Excessive Venting, or Bragging.
Using this subreddit to crowd source answers to something that isn't really contributing to the spirit of this subreddit is forbidden at moderator's discretion. This includes posts that are mostly focused around venting or bragging; both of these types of posts are difficult to moderate and don't contribute much to the subreddit.
Smallpaul@reddit
No.
There are many extremely talented engineers who are also excited about LLMs and vibe coding in its appropriate place.
https://agent-wars.com/news/2026-03-15-graydon-hoare-llm-shift-software-development-2025-2026
https://www.remio.ai/post/linus-torvalds-vibe-coding-how-the-linux-creator-uses-ai-for-hobby-projects
https://www.teachingpython.fm/150
Yes product managers love them for the reason you describe. So do many managers.
Oakw00dy@reddit
Vibe coding is a good way to try out and express ideas. I'd much rather take a vibe coded app as the specification than a 200 page Word document. The disconnect is when PMs and stakeholders equate the effort for coming up with a vibe coded UI with a full fledged solution that can actually be shipped and maintained.
LuckyEmphasis2480@reddit
how do you define "vibe coding" in your experience?
dagamer34@reddit
Even with Opus 4.6, the code it writes, while technically correct, is organizationally sloppy. It’s pretty easy to see if no human has ever attempted to maintain the code, it’ll repeat itself, it doesn’t structure things well, it’ll happily generate multi-hundred line files. The codebase is not engineered in any way.
The nature of whatever you are producing will be the forcing function, the risk tolerance for bugs. If you are making a web app with a user base of 10 that only reads data from an authoritative source vs building an operating system which needs to get installed to be run, the latter “should” have no bugs. The turn around time for fixing a bug and getting it to customers is weeks to months. The desire for data loss is zero. Engineers absolutely have to be acountable for whatever gets made.
disposepriority@reddit
Famous product manager on LinkedIn definitely sounds like something you should be following!
Eligriv@reddit
I'm an former dev now pm. You know why i still lurk in this sub ? Because there are no equivalents in product management spaces : it's all posturing, people selling themselves, nothing genuine.. It's linkedin EVERYWHERE with these guys. So obviously they'll push whatever garbage they think'll make their carreer advance. The AI lot were all in the blockchain a few years prior.
Ok_Confusion_4746@reddit
Are you saying that my bored ape is not worth 2 million dollars ?
What's next ?! My metaverse villa is a glorified Sims' villa ?
Please...
Lyelinn@reddit
Nothing like reading daily dose of hard cringe with a cup of coffee every morning lol
ionforge@reddit
Is there anything else but hard cringe in LinkedIn?
CodelinesNL@reddit
Recruiters whining about developers being mean to them? 😉
Lyelinn@reddit
occasionally I get "happy birthday" messages there!
CodelinesNL@reddit
I think it's actually quite dangerous to have these kind of black and white views on what's happening. It's a pattern I see a lot in this sub, and on Reddit as a whole. We use Claude Code in our team and it works very well for a lot of things, especially if you give it proper structure. There are also some situations it's really not good at.
A lot of developers are reasoning from a lack of experience with the right tools and the right processes. I've been there; C-level mandated use of AI, and then giving us a crap Copilot license. F right off. But in our company (small fintech) it's immensely useful, for product, devs, QA, you name it.
I'm happy to explain the process, but I've tried before and people reacted as if I was paid by LLM vendors directly, so unless there's a clear interest, I won't.
So what it's really good at: writing useful stuff when it has clear guidance and a clear picture of what is needed. So proof of concepts, greenfield stuff, etc. What it struggles with is large old codebases that have a LOT of technical debt. In our company we have both. But what I also notice is that the more we clean up that tech debt, and the more context we give it, the better it gets at that too.
Yes, there's a lot of dunning kruger and devs who are lazy and use it to pretend they're useful. Yes those large enterprises that shove Copilot through everyone's throat are going to suffer. But what the top of the line tools can do right now, is incredibly impressive. Just as an example; we need to make a decision on some pretty fundamental architectural decisions. Instead of spending a month doing an analysis, I just created a proof of concept, learned from it, and settled on a design in a single day.
ChallengeDiaper@reddit
You are spot on and that’s been my experience as well.
Folks who can’t see that either haven’t learned how to use the tools effectively or are letting their emotions blind them.
I’m glad I’m at the tail end of my career. Within five years, this profession is going to look VERY different and a lot of people are scared of that.
CodelinesNL@reddit
For me personally; I like my job a lot more. I'm currently in a role where my main purpose is enabling the other engineers. We have a lot of backlog in improvements in our CI/CD and infra stuff for example. I generally hate to do that work because it's A) boring and I'm B) constantly looking up CLI commands, trying to figure out what's going wrong, etc.
Claude Code is rediculously good at this. The quality of the stuff it produces with the right process (openspec) and right rules (TDD approach, modular design, etc.) is frankly better than most devs I've worked with.
Oh, and the data analyst role won't exist more in a few years. It's extremely good at writing queries, finding correlations in data, and building dashboards for them.
So by all means "hate on" the AI slop that some of your colleagues are producing with crappy tools like Copilot. But do yourself a favour; learn what the right tools can actually do. Because our work will fundamentally change, this change has already started, and it's hard to imagine where will be in a year from now looking at what it could do a year ago.
chickadee-guy@reddit
That isnt impressive? Like that was the point of the whole thing? To tell us that you would have spent a month reading documentation, now you can offload that to an LLM?
Professional_Ebb5640@reddit
that's rough. seems so counterproductive
Bricktop72@reddit
It's the senior developers and system architects leading the push where I am. Developers and architects get access to the latest models on copilot. The versions of AI available to PMs are hobbled by security.
GreenOrg@reddit
In my company, all managers look like additcted for using LLM. Every news in AI (theoretical or a statement about the future of AI from some top guy in AI) is shared with developers; they are very obsessive.
Most of our discussion about developing moved from "let's create a plan, docs, estimates, test" to "what AI will you use? Why not Claude? Cursor is bad. etc ..."
They think AI is a golden pillow that will help our company become a 1BL unicorn.
I really dont like it and start to hate it.
rcls0053@reddit
No. The enthusiasm is a byproduct from AI companies that are pushing their product down the throats of companies and spreading fear, saying you will never be able to compete unless you use our product to gain an edge, and those companies don't currently have a problem that needs solving with AI.
chickadee-guy@reddit
The losses are at the scale of hundreds of billions with a B at this point. The technology and its rollout is an unmitigated disaster
vibes000111@reddit
No, it’s not mainly product people.
I think some people are looking for all kinds of excuses to explain away the hype around agentic coding. It’s just hard to accept that it’s genuinely useful and a massive leap over where we were a few years ago. It’s hard to change how you work, some people will find the new way of working unappealing, some people will find the fact of change itself uncomfortable, some people will feel like they’re falling behind and resent that feeling.
And the fact that there’s actual hype and exaggeration and uncertainty and imperfect solutions and genuine bullshit - all of these things make it very easy to find excuses and form negative opinions instead of accepting that you need to change something like your opinion or your ways of working.
chickadee-guy@reddit
It isnt hard to accept it, it just isnt true and there is 0 evidence showing the above statement is true. Opus 4.6 was a full on regression from its prior model, per AMD.
another_dudeman@reddit
The thought of a developer feeling left behind using these tools is stupid lmfao
vibes000111@reddit
What do you mean? I genuinely can't tell which side of the argument you're on.
another_dudeman@reddit
Devs learn shit everyday way more complicated than using LLM agents.
vibes000111@reddit
It's not about how complicated it is, it's about the fact the it's a different way to do work that many people aren't used to. Getting used to working with LLM agents takes time, and some people aren't even trying (because they're telling themselves that it's bullshit propagated by people selling LLMs / product managers / bots).
Unfair-Sleep-3022@reddit
No, it really really doesn't take any time at all.
Let's stop pretending it's a "skill". The whole point of the tool getting better is that you don't need complicated prompts to make it work.
vibes000111@reddit
It's a skill and it's much more than just writing good prompts especially in the last six months (multiple agents, skills, spec driven development). And it's important to understand that it's a skill, otherwise you get the "I tried it and it's not that good, people online are overselling it" - yeah, some people are overselling it but a big part of that experience might be that you don't understand how to get the maximum from it.
Writing good prompts is a topic from 3 years ago.
Unfair-Sleep-3022@reddit
- Multiple agents is something that gets fixed by the harness like conductor. Humans are proved to be bad at multi tasking, so you have a biological ceiling to how many complex tasks you can do concurrently without losing up to 40% of productivity.
- Stupid things like skills, progressive context disclosure etc are just different flavors of prompting. The truth is that the whole thing gets bundled and passed to the LLM so just putting a bow on it doesn't make it any different from "better prompting"
- Calling out the years like the thing has fundamentally changed from the first day is peak vibe coder lack of understanding. The models got better, the harnesses got better and the technology works largely exactly the same way.
Why do you think using tabs for a bunch of agents while you progressively can't keep up and understand what you're doing is a skill? Your results will be trash.
No, at best what any biological human can do is focus on one complex task and have other agents on silly side quests that wouldn't get done otherwise but that don't "move the needle" to use the same parasitic language.
This is also why the productivity gains are fringe and there's no provable / measurable gain from using the technology.
vibes000111@reddit
I don’t necessarily disagree with a lot of this, not sure I got the “people like you” comment, I think you might have gotten the wrong impression about my position (I’m not defending the specific tools that are popular now).
My point is - you have understanding that many people don’t have and it affects how you work - that’s what I’m calling a skill. You’re not the same as someone saying “I don’t need agentic coding now and there’s nothing about this tool that I need to learn”.
Unfair-Sleep-3022@reddit
Sorry for my bluntness then
It's just been a bit exhausting over here lol
"I'm tired boss"
shephrrd@reddit
You present like developers either love LLMs and have hopped on the wagon or they hate it and are resistant to its use.
I hate it, and I use it. I recognize its usefulness. I hate it because it has sucked the joy and soul out of writing code. I’m now a glorified babysitter who does code reviews and tweaks. There are loads of developers like me, I’m sure.
vibes000111@reddit
Didn't mean to present it that way, I had a specific segment of the anti-LLM crowd in mind. I'm sure there's many people like you, it's a very understandable reaction.
another_dudeman@reddit
Again, not hard for a dev that's had to learn things continuously their whole career.
hondacivic1996@reddit
8 month old hidden account called «vibes000111» posting an obviously LLM generated text about how real the hype is. Can it get any more obvious?
vibes000111@reddit
I'm a real person who's been using reddit since 2013 - I used to make a new account every 8-12 months because how easy it is to correlate personal information but last month I changed to just hiding my history instead.
The "hiding your history = bot" logic is just paranoia.
You can add that to the list of stories you can tell yourself if you don't like where things are going with agentic coding - everyone who disagrees with you is a bot... or a product manager like OP does, depending on which one of the two you think less of.
shimroot@reddit
I work in product and I’m really excited about what I can do with LLMs.
The way I see LLM product work is that instead of spending stupid amounts of time writing docs, making diagrams, building clickable prototypes, and so on - I can just chat with an LLM to get a quick prototype that I can then show to the dev team so we have a better starting point for discussions.
zinguirj@reddit
My experience is quite the opposite, PMs using LLMs as argument to reduce deadlines, I heard "what would that take so long, can you just ask Claude to do it? Here I just did and got me almost everything done" also got one where the PM asked claude code to implement a feature when we got to the refinement instead of a especification we got a PR done by claude and the PM asking how can we get live "It shouldn't take more than a day or two since the heavy work is already done"
shimroot@reddit
Yeah, I can see how that can be the majority of cases. Until now dev estimates were taken at face value, but LLM tooling gives people a reason to push back without knowing what they’re talking about.
I’m building some side stuff with LLMs to get a better feel for what I can and can not do and a v1 is aleays done quickly. But throw some different use cases at it and then it’s a few days trying to figure things out and make it work properly.
Friendly-Nobody8023@reddit (OP)
That sounds like the perfect application of this LLM tech if you ask me!
nkondratyk93@reddit
nah - most of this is exec mandate, not PM initiative. PMs I know are asking 'what does success look like' not driving the hype. we're translating
More_Foundation21@reddit
this is just patently untrue for most experienced PMs, which most of the ppl you see on Linkedin are not, especially the influencers
Aggravating-Slip5857@reddit
The tools are real. The process around them isn't.
Tons of lines of code in zero time, but the rest of the pipeline is not there yet - releases, deployment, tracking what's deployed where.
You know, the engineering around it.
PMs probably don't look at this first.
campid0ctor@reddit
I hate this phenomenon so much. PMs suddenly think they can now give technical recommendations because they vibe coded a prototype and they think it's "performant". Maintainability be damned, just keep on excreting code slop so that sales can have something shiny to play with.
AfricanTurtles@reddit
This is the most frustrating for me personally. It's like just shut the f up you actually don't have any idea what you're talking aboooooooooout 😡
Tasty_Condition1212@reddit
i'm not sure it's just product managers pushing for this
Unfair-Sleep-3022@reddit
Nah it's mostly investors. The same people that invest in AI also invest in other publicly traded companies. They have a stake in the massive amount of money pumped into the AI economy, so they force other companies to use the products. This is also how you end up with very dangerous cycles like the one between NVIDIA, OpenAI and Oracle
Unfair-Sleep-3022@reddit
No, it's investors
lolninja@reddit
My feeling as a PM is that as LLMs become more capable the gap between “I know what I want” and “I know how to build this” collapses.
03263@reddit
I already considered a big part of my job to be figuring out what people actually want and helping them align their dreams with their budget
Most people do not know what they want, or what is even possible. They are better at describing a problem, but want to describe a solution and it takes some serious work to get the full idea of the problem and why they gravitated towards that solution.
03263@reddit
They overestimate how useful it is.
Especially noticeable with ones at smaller companies that think they will get the same results as those spending big bucks on tokens, but within the free tier of their MS365 subscription.
propostor@reddit
Honestly in my world the only people I see obsessing over LLMs is on Reddit, where I am becoming increasingly sure that those people are actually bots.
ForeverIntoTheLight@reddit
Don't go looking at what LinkedIn lunatics do. Just don't, if you value your sanity.
TerribleShoe2440@reddit
it feels like some pms might be oversimplifying software development challenges
Wide_Obligation4055@reddit
No. The head of engineering is pushing hardest. Mythos/Capybara worshipper. Question is will it be a mythical new land of milk and honey OR a giant hairy stinking rat 😁
But as a software developer it's 24/7 Claude code shells ... or sling your hook.
another_dudeman@reddit
The people pushing hardest for vibe coding are the people selling their LLM
Soleilarah@reddit
IMO, LLMs are perfect for product managers ; it allows you to write quickly (emails and such) and to make prototypes that can give a general idea of where the product needs to go.
This means being a better "bridge" between departments (i.e. "translating" the needs from marketing to the devs and avoiding thousands of meetings during development to adjust the project on the go)
Van_Quin@reddit
No, its my idiot engineering boss that thinks that he could vibe code through the whole planet
rxvf@reddit
i’m a tpm and believe me when i say that linkedin pms are delusional
Outside-Storage-1523@reddit
In my place the management is the most keen on it. Someone even vibed code an app by himself.