the bubble has burst or, rather, currently is in the process of bursting.
My job involves working directly with AI, LLM’s, and companies that have leveraged their use. It didn’t work. And I’d say the majority of my clients are now scrambling to recover or to simply make it out of the other end alive. Soon there’s going to be nothing left to regulate.
GPT5 was a failure. Rumors I’ve been hearing is that Anthropics new model will be a failure much like GPT5. The house of cards is falling as we speak. This won’t be the complete Death of AI but this is just like the dot com bubble. It was bound to happen. The models have nothing left to eat and they’re getting desperate to find new sources. For a good while they’ve been quite literally eating each others feces. They’re now starting on Git Repos of all things to consume. Codeberg can tell you all about that from this past week. This is why I’m telling people to consider setting up private git instances and lock that crap down. if you’re on Github get your shit off there ASAP because Microsoft is beginning to feast on your repos.
But essentially the AI is starving. Companies have discovered that vibe coding and leveraging AI to build from end to end didn’t work. Nothing produced scales, its all full of exploits or in most cases has zero security measures what so ever. They all sunk money into something that has yet to pay out. Just go on linkedin and see all the tech bros desperately trying to save their own asses right now.
that’s interesting in all honesty and I don’t doubt you. all I know is my bank account has been getting bigger within the past few months due to new work from clients looking to fix their AI problems.
I think you’re onto something where a lot of this AI mess is going to have to be fixed by actual engineers. If folks blindly copied from stackoverflow without any understanding, they’re gonna have a bad time and that seems equivalent to what we’re seeing here.
I think the AI hate is overblown and I tend to treat it more like a search engine than something that actually does my work for me. With how bad Google has gotten, some of these models have been a blessing.
My hope is that the models remain useful, but the bubble of treating them like a competent engineer bursts.
Agreed. I’m with you it should be treated as a basic tool not something that is used to actually create things which, again in my current line of work, is what many places have done. It’s a fantastic rubber duck. I use it myself for that purpose or even for tasks that I can’t be bothered with like creating README markdowns or commit messages or even setting up flakes and nix shells and stuff like that, creating base project structures so YOU can do the actual work and don’t have to waste time setting things up.
The hate can be overblown but I can see where it’s coming from purely because many companies have not utilized it as a tool but instead thought of it as a replacement for an individual.
At the risk of sounding like a tangent, LLMs’ survival doesn’t solely depend on consumer/business confidence. In the US, we are living in a fascist dictatorship. Fascism and fascists are inherently irrational. Trump, a fascist, wants to bring back coal despite the market natural phasing coal out.
The fascists want LLMs because they hate art and all things creative. So the fascists may very well choose to have the federal government invest in LLM companies. Like how they bought 10% of Intel’s stock or how they want to build coal powered freedom cities.
So even if there are no business applications for LLM technology our fascist dictatorship may still try to impose LLM technology on all of us. Purely out of hate for us, art and life itself. edit: looks like I commented this under my comment the first time
the bubble has burst or, rather, currently is in the process of bursting.
My job involves working directly with AI, LLM’s, and companies that have leveraged their use. It didn’t work. And I’d say the majority of my clients are now scrambling to recover or to simply make it out of the other end alive. Soon there’s going to be nothing left to regulate.
GPT5 was a failure. Rumors I’ve been hearing is that Anthropics new model will be a failure much like GPT5. The house of cards is falling as we speak. This won’t be the complete Death of AI but this is just like the dot com bubble. It was bound to happen. The models have nothing left to eat and they’re getting desperate to find new sources. For a good while they’ve been quite literally eating each others feces. They’re now starting on Git Repos of all things to consume. Codeberg can tell you all about that from this past week. This is why I’m telling people to consider setting up private git instances and lock that crap down. if you’re on Github get your shit off there ASAP because Microsoft is beginning to feast on your repos.
But essentially the AI is starving. Companies have discovered that vibe coding and leveraging AI to build from end to end didn’t work. Nothing produced scales, its all full of exploits or in most cases has zero security measures what so ever. They all sunk money into something that has yet to pay out. Just go on linkedin and see all the tech bros desperately trying to save their own asses right now.
the bubble is bursting.
The folks I know at both OpenAI and Anthropic don’t share your belief.
Also, anecdotally, I’m only seeing more and more push for LLM use at work.
that’s interesting in all honesty and I don’t doubt you. all I know is my bank account has been getting bigger within the past few months due to new work from clients looking to fix their AI problems.
I think you’re onto something where a lot of this AI mess is going to have to be fixed by actual engineers. If folks blindly copied from stackoverflow without any understanding, they’re gonna have a bad time and that seems equivalent to what we’re seeing here.
I think the AI hate is overblown and I tend to treat it more like a search engine than something that actually does my work for me. With how bad Google has gotten, some of these models have been a blessing.
My hope is that the models remain useful, but the bubble of treating them like a competent engineer bursts.
Agreed. I’m with you it should be treated as a basic tool not something that is used to actually create things which, again in my current line of work, is what many places have done. It’s a fantastic rubber duck. I use it myself for that purpose or even for tasks that I can’t be bothered with like creating README markdowns or commit messages or even setting up flakes and nix shells and stuff like that, creating base project structures so YOU can do the actual work and don’t have to waste time setting things up.
The hate can be overblown but I can see where it’s coming from purely because many companies have not utilized it as a tool but instead thought of it as a replacement for an individual.
At the risk of sounding like a tangent, LLMs’ survival doesn’t solely depend on consumer/business confidence. In the US, we are living in a fascist dictatorship. Fascism and fascists are inherently irrational. Trump, a fascist, wants to bring back coal despite the market natural phasing coal out.
The fascists want LLMs because they hate art and all things creative. So the fascists may very well choose to have the federal government invest in LLM companies. Like how they bought 10% of Intel’s stock or how they want to build coal powered freedom cities.
So even if there are no business applications for LLM technology our fascist dictatorship may still try to impose LLM technology on all of us. Purely out of hate for us, art and life itself. edit: looks like I commented this under my comment the first time