digitcatphd

Respectfully, starting one's argument with a factually invalid statement is not a good way to argue against a bubble. If by 'every' you are referring to the foundation model providers, this is not 100% of the 'AI service' market and even then, I would argue that a lot of this demand will need to be answered to by companies measuring ROI after the FOMO or unreasonable expectations get settled in and right now, my primary argument is this ROI is driven by speculation rather than empirical measurement.

I have spoken with many companies and nearly all of them, when speaking about AI, have gotten to the point they don't even make any sense. A common theme is 'we need' AI, but nobody can articulate 'why' and in-fact they get defensive when questioned. It is almost perfectly parallel to the 'we need blockchain' argument or 'we need a mobile app'. That isn't to say those are not useful technologies, but the rapid rise, steep decline, then gradual rise is a theme in tech.

I have started to notice this as well over the past several months, in-fact, I would say it is orders of magnitude larger than the Crypto bubble and when it bursts will be significantly more impactful. Right now, everything is propped up on the premise the hyperscalers will grow EPS proportionately to their investment and that ROA is being priced in as a best case scenario (hope) in their share price. Maybe its not ROA at all, maybe its simply FOMO, we keep citing this 'AI Race' as if there is some end objective to 'win' drawing parallels to the nuclear arms race that only resulted in massive wasted CAPEX in decaying nukes sitting in unused bunkers since the cold war. (Not to mention it isn't even clear this arms race played a direct factor in the war beside depleting resources) Everyone is happy right now. Execs get stock bonuses, investors get returns, vendors get contracts. Once this is questioned, and history demonstrates it undoubtedly will be, it will cause a cascading collapse. History is a mathematical truth.

I ran some tests on a similar concept using LangGraph. Unfortunately, I think while the results are meaningfully different from a foundation model, they don't supplement actual real-world data yet and don't provide sufficient diversity of thought and opinion between cases. For instance, asking 1,000 people you get a different response every time because each person is slightly unique, but with the LLM it is probabilistically different, not because of the slightly unique differences and for Meta reasons.

Agreed. I think most of the arguments are premised on AI getting infinitely better for some reason, without anyone outlining clear arguments addressing the inherent architectural limitations of today's LLMs. I have been in this space since 2021 and honestly, beside maybe voice and Gemini Deep Research, things aren't much better than GPT-4.

I’m sure most of the problems cited in this article will be easily solved within the next five years or so, waiting for perfection and doing nothing won’t pay dividends

If I’m not mistaken they are using high valuations of top companies to conclude AI is overhyped?

Sorry, but weren’t these valuations escalated because of low interest rates and quantitative easing? Perhaps combined with increased concentration in Top 10 by investors navigating uncertainty?

Typical BS coming from a mega fund only supported by management fees. not saying AI isn’t hyped but this is laughable.

this is smart, but I think NVIDIA's paper on fine tuning small language models presents a sightly more efficient approach

I built a GitHub repo that allows you to quickly spin up LangGraph agents using Claude Code with built-in tools and pre tested configurations.

https://www.braid.ink

Makes sense, yes, I think a spectrum of tools will be useful depending on the build. N8N, LangGraph, and working with primitives all have their use case and then tools built on each will be useful much like the landscape of RAG builds have evolved.

All these arguments remind me of early LLMs. In fact, most of the arguments are nearly identical. Heck, even Bitcoin back 10 years ago was demonized and now governments are pushing for national reserves.

If you have used GPT o3, you have used Agentic models. If you use Claude Code or Cursor, you have used Agentic models. If you have used Claude + MCP, you have used agentic models.

This isn't theoretical - its already being applied. In fact, Langchain just hosted an entire conference to showcase how Fortune 100 companies were using LangGraph and I can assure you if they could have used a more simple architecture they would have.

If you wait around until everyone considers it to be acceptable in the mainstream all the good opportunities to create disruptive tech will be gone. So please - keep publishing this luddite trash and throw shade on the sidelines. I will keep doing agent research and building with it. We will see in 10 years who was wrong I guess.

https://www.qodo.ai/blog/building-agentic-flows-with-langgra...

Thanks!!!

As of now yes. But we are still in day 0.1 of GenAI. Do you think this will be the case when o3 models are 10x better and 100x cheaper? There will be a turning point but it’s not happened yet.

The reality is I suspect one will use different models for different things. Think of it like having different modes of transportation.

You might use your scooter, bike, car, jet - depending on the circumstances. A bike was invented 100 years ago? But it may be the best in the right use case. Would still be using DaVinci for some things because we haven't bothered swapping it and it works fine.

For me - the value of R1/o3 is visible logic that provides an analysis that can be critiqued by Sonnet 3.5

Exactly. Everything is massive kindle for a wildfire and it comes to show how tiny of a spark it takes. Yet - people keep adding more kindle.

This doesn't have anything to do with DS, despite what the media claims, it comes to a heightened sense of fear people have with playing with fire.

Nah, this just means training isn’t the advantage. There’s plenty to be had by focusing on inference. It’s like saying apple is dead because back in 1987 there was a cheaper and faster PC offshore. I sure hope so otherwise this is a pretty big moment to question life goals.

Don't make me tap the sign.

Economic cycles. As soon as the next cycle builds up momentum all the low risk entrepreneurs will rush in again.

All of the arguments presented in this were equally true in 2013 as they are in 2025 and there are many more dimensions than you are taking into consideration. For instance, the demographics of HN users could have changed, or the economic climate is cyclically more defensive and 'risk off', which is well known. I suspect this is more related to economic cycles than a cultural paradigm shift and would suggest the author to think more deeply than forming vague generalizations around some HN comments of anonymous users.

I disagree, these search indexes already exist, they just need to be navigated much how Cursor uses agentic search to navigate your codebase or you call Perplexity to get documentation. If the knowledge exists outside of your mind it can be searched agentically.

I wasn't referring to OAI, but rather:

1. Existing legacy players with massive data lock-ins like ERP providers and Google/Microsoft.

2. Massive consolidation within AI platforms rather than massive fragmentation if these legacy players do get disrupted or opportunities that do pop up.

In other words - the usual suspects will continue to win because they have the data and lock in. Any marginal value in having a specialized model, agent workflow, or special training data, ect. will not be significant enough to switch to a niche app.

It is indeed unfortunate and niches will definitely exist. What I am referring to is primarily in enterprise.

I agree with you at this time, but there are a couple things I think will change this:

1. Agentic search can allow the model to identify what context is needed and retrieve the needed information (internally or externally through APIs or search)

2. I received an offer from OpenAI to give me free credits if I shared my API data with it, in other words, it is paying for industry specific data as they are probably fine tuning niche models.

There could be some exceptions to UI/UX going down specific verticals but eventually these fine tuning sector specific instances value will erode over time but this will likely occupy a niche since enterprise wants maximum configuration and more out of box solutions are oriented around SMEs.

Not for filing. The hard part is understanding how to find exemptions, deductions, ect. It has already outperformed both my CPA and tax attorney and at this point I am using it to correct them before taking their advice.

It's very simple actually. For instance, let's assume Walmart automotive takes maybe 2-3 FT employees to operate, where this may normally be fragmented across 2-3 local mechanics employing 3-6 FT employees each. Then of course you have the 2-3 local services which catered to them (E.g. cleaning services, real estate agents, CPAs, ect). Then, multiply this for each Walmart department (Electronics, Groceries, Pharmacy, Auto, Household, ect.) and there's your town.

Using creative new methods, they find that the costs Walmart imposes in the form of not only lower earnings but also higher unemployment in the wider community outweigh the savings it provides for shoppers.

> This consolidation is the root cause

o3 fixes the fundamental limitation of the LLM paradigm – the inability to recombine knowledge at test time – and it does so via a form of LLM-guided natural language program search

> This is significant, but I am doubtful it will be as meaningful as people expect aside from potentially greater coding tasks. Without a 'world model' that has a contextual understanding of what it is doing, things will remain fundamentally throttled.

Good to see studies of incidents in comparison to human drivers. Historically what I have seen is 'self-driving car crashes into X, which seems to be concentrating exclusively on outliers to prove a point.

Their demo video was uploading a picture of a birdhouse and asking how to build it

The reality is AI is largely in the uncanny valley right now. This is great because it weeds out the noise and gives the focused long term builders breathing room.