Hacker Newsnew | past | comments | ask | show | jobs | submit | malshe's commentslogin

Yeah but again that’s all here on HN. I have plenty of friends who are not on HN and I haven’t heard anyone ever complain about these things.

I have heard a few analysts mentioning that software stocks are under the threat of AI. Apparently now anyone can build enterprise software in their pajamas using AI. To the extent that this ridiculous reasoning is driving the stocks down, I think it presents a good buying opportunity. But one has to wait until the bleeding slows down a bit.

it might be more that business process can become a plain text word document, that modifying the program requires only describing the change in plain language, that the user interface to show information becomes unnecessary when you can just ask any question, that data can be loose and unstructured, even stored as images, and interpreted on the fly.

the general purpose chatbot plus a "how to" does replace needing to build esoteric specialized workflows.


Now can you explain how do you replace Service Now (service management tool cursed with the ticketing system) with a flat text file.

If it's a compelling proposal, I might share it in my company. I'll make sure to credit you in the document.

Of course in line with your radical complexity savings, I expect the comprehensive proposal to be at most one paragraph long :)


To replace a legacy ticketing monolith like ServiceNow with a text-based LLM pipeline, you should transition to a "GitOps for Service Management" model: replace form-based entries with a central, version-controlled Markdown or JSONL repository where every ticket is a discrete text file containing structured metadata (tags, timestamps, status) and unstructured conversation logs. These files are monitored by a CI/CD pipeline that triggers a RAG (Retrieval-Augmented Generation) indexing process, allowing a fine-tuned LLM to serve as the primary interface for querying historical solutions, generating automated responses, or updating ticket states via natural language commits. By treating the service desk as a living document store rather than a relational database, you eliminate UI friction and enable the LLM to act as the "logic layer" that categorizes, routes, and resolves issues directly from the raw text stream.

I think in a weird way SpaceX listing publicly would trigger Tesla’s downfall as Elon fans will switch to that stock. Then SoaceX will buy out Tesla.

That’s a good point. Can Tesla fail?

Seems like he’s constantly using one company to fund others, shuffling the cups and balls around claiming everything is still fine.

I could see him doing serious damage or even trashing an otherwise healthy company doing this to prop up total failures.


> Can Tesla fail?

If SpaceX buys it, it will fail upward :)

He did that with SolarCity when Tesla bought it then repeated with X when XAi bought it.


10x revenue for a company with declining sales is way overvalued.

Tesla doesn't disclose the gross margin on Cybertruck. They may say it is positive but if nobody knows what constituted those gross margins or what they amounted to, it's pretty much meaningless.

Hard to imagine it being profitable given the very low utilization of the production line and associated tooling.

I agree. I am quite confident that if someone challenges them on this claim, they will say it was non-GAAP gross margin, which excluded all the crucial expenses

That's because Trump gave many extensions and concessions to so many countries. Remember there was 125% on China in May 2025 before Xi decided to use rare earth minerals to fight back. Maybe you have heard of TACO. So the tariffs as threatened never panned out in reality.

How come all the existing tariffs don't tank the economy?

They are affecting specific industries and consumers. The inflation right now is stable only because oil is so cheap. And that has nothing to do with tariffs.

If you read this person's comments, looks like they are just making up crap. Apparently this one person has met or interviewed all the Indian H1Bs in the US.

They do and there is nothing wrong with that. The papers published in this journal are peer-reviewed and go through multiple rounds of review. Also, note that Andrew King could carry out the replication because the data is publicly available.

I will give my perspective as an academic who writes R and Python code for data analysis (including a lot of data cleaning).

1. I find AI written code verbose and inelegant. It makes it difficult to troubleshoot. I take pride in my own code and share it confidently with my doctoral students and coauthors. I can't share AI written/assisted code with the same confidence let alone pride.

2. Often AI takes shortcuts and writes terrible code. This is especially true for Bayesian models. My first check with any Bayesian model is to recover parameters using a simulated dataset. If the code fails to do that, there is no point in going forward. I used Opus 4.5, Gemini 3.0, and GPT 5.2 recently to write rather simple code for random parameters dynamic panel model. There are already papers that have done it. All three failed numerous times. I got it to work after a lot of handholding.

3. AI helps tremendously while creating web apps that make my analysis more actionable. A lot of reviewers now want to see something in action, so using Streamlit or Shiny is the way to go.


How about using fiber optic cables for this? I saw a few videos on YouTube showing the installation for home internet

To what end? The runs aren't going to be long enough for fiber to provide a benefit, and the transceivers are more expensive for consumer use like this.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: