chrismarlow9

I'm finding it challenging to believe they wouldn't just cannibalize anything dependent on them in that way or at minimum launch a directly competing product.

You can use a frontier model to create a plan that's specific enough for a local model of a very small size to execute on. The more specific you are and compartmentalize tasks the "dumber" the local model can be.

Edit: Obviously you'll be using more tokens but this is the trade off for running a smaller model and running locally. Similar to time memory trade off but in token economics. Sorry I need more coffee

That's cool but don't get heart broken when engineers leave after repeatedly being told the thing they worked on for months wont ship because someone at the top said so.

I've seen it before and over time it always plays out similarly.

- the cloud was invented and we were told CTOs would be able to just point and click and make infrastructure and apps! What did we actually get? Another layer of abstraction to debug through. Does it have is perks? Yes. Does it have it's own problems? Yes. Is it more expensive than setting up bare metal and having a solid team? Well that depends on what you're doing and the economics of it and the team.

- then document storages came along like and got wildly popular like mongo and people were calling it the end of SQL! And no more complexity or relational nonsense. Everything is JSON and life is great. What actually happened? These companies realized over time their data was getting trashed, adding things and fixing bugs became complex in pure docstore systems. So the initial v1 was easy and looked beautiful but only 4 years in you have a production db with orphan data that's twice as big as it should be. New features take forever to see a clear path in adding it to the model because it's no longer as intuitive to get performance for a feature.

Anyway. I see AI taking both of these roads at the same time. In 5 years I believe the code will be a giant pile of unfixable mess for most vibe coded things. I also don't see it getting rid of programmers but just adding another layer of abstraction to the mix that yes is helpful, but only if you already know what you're doing, much like what came of the clouds.

surely concats of user input, stdout of external dependencies, and non-deterministic output feeding back directly to an eval is safe. it's never been a problem before. not even trying to check the boxes when it comes to security anymore.

It's been my default search for years. Lately for quick one shot AI prompts I use duck.ai (they put some basic effort into anonymizing your chat: https://duckduckgo.com/duckduckgo-help-pages/duckai/ai-chat-... ).

For the search, some of the local results are wrong but I live in a very small area so it may be more reliable for highly populated areas. Lately I've been checking out Kagi for a few things just to see what the quality is like on competition. The anonymized chat (proxy) for AI is cool but very small context limit. Good for looking up random questions and they typically include references.

Yeah, ETF is a function of strategy, asset, financing methods and many other things. Varying leverage alone can multiply one ETF into 3 or 4. Other than showing that there's more risk in a collapse than just the single security I'm not sure what I'd do with this information.

We've been down this road when backlinks ran the game. It eventually ends with parasitic hosting. Find a domain with authority and spam whatever mis information or spam you'd like AI to run there. Or buy a domain that has trust already. Or for the darker hats just literally hack the site and use cloaking to send fake info to the AI bot. It's probably already being done.

Everything old is new again when you start a new market. If you think that AI is bad imagine what old tricks are new with polymarkets

Sounds about right. Security is a joke everywhere right now. First to market is all that matters anymore and security is the very first thing to be thrown out when it stands in the way.

100% agreed. use the non-deterministic thing that is right 90% of the time to generate a deterministic thing that is right 100% of the time. one of the key things I add to my prompts is:

- Please consult me when you encounter any ambiguous edge cases

Attaching the AI to production to directly do things with API calls is bad. For me the only use case where the app should do any AI stuff is with reading/categorizing/etc. Basically replacing the "R" in old CRUD apps. If you want to use that same new AI based "R" endpoint to auto fill forms for the "C", "U", and "D" based on a prompt that's cool, but it should never mutate anything for a customer before a human reviews it. Basically CRUD apps are still CRUD apps (and this will always be true), they just have the benefit of having a very intelligent "R" endpoint that can auto complete forms for customers (or your internal tooling/Jenkins pipelines/etc), or suggest (but never invoke) an action.

Blackhat SEO spamming knew this 20 years ago

I am a Senior DevOps/Infrastructure/Full-stack engineer. First 10 years of my career was software development (LAMP & C# work). Last 10 years have been DevOps, SRE, and Infrastructure where I often still ended up in the code. I enjoy working in the security, finance, healthcare, and educational industries.

Location: Texas, US

Remote: Yes

Willing to relocate: Yes

Technologies: Terraform, Kubernetes, Python, Golang, Ansible, PostgreSQL, AWS, Ansible, Agentic Coding (OpenCode + vLLM + Local Models + Frontier/ VSCode + Copilot), Bash & Linux (all distros)

Résumé/CV: https://www.linkedin.com/in/chris-marlow-37661177/ (full resume on request)

Email: chrismarlow9[at]gmail.com

I use local models on a Mac mini for most things and fall back to the hosted ones when they can't get the job done. Of course you have to break the work into smaller pieces yourself that a local model can understand. One good side effect of this is that you end up actually learning the code and how it's structured.

A better rule is "complicate only if necessary"

You're not alone. I was hoping for the same.

We propped the entire economy up on it. Just look at the s&p top 10. Actually even top 50 holdings.

If it doesn't deliver on the promise we have bigger problems than "oh no the code is insecure". We went from "I think this will work" to "this has to work because if it doesn't we have one of those 'you owe the bank a billion dollars' situations"

Do you feel the same about AWS?

I don't have time to put together a submission but I'm willing to bet you can use this:

https://github.com/kjdev/php-ext-jq

And replicate this command:

jq -R ' [inputs | split(",") | {url: .[0], date: .[1] | split("T")[0]}] | group_by(.url) | map({ (.[0].url): ( map(.date) | group_by(.) | map({(.[0]): length}) | add ) }) | add ' < test-data.csv

And it will be faster than anything you can do in native php

Edit: I'm assuming none of the urls have a comma with this but it's more about offloading it through an extension, even if you custom built it

The slow part as a senior engineer has never been actually writing the code. It has been:

- reviews for code

- asking stakeholders opinions

- SDLC latency (things taking forever to test)

- tickets

- documentations/diagrams

- presentations

Many of these require review. The review hell doesn't magically stop at Open source projects. These things happen internally too.

Not nearly enough context to be a front page post but here we are. Does everyone just give up on all fundamental expectations including determinism when they see AI in the title? Is the AI final boss to take over the front page of hacker news? /rant

To answer the question I would highlight the wrong regions in neon green manually via code. Now feed the code (zipped if necessary) to the AI along with a screenshot. Now give it relatable references for the code and say "xxxx css class/gtk code/whatever is highlighted in the screenshot in neon. I expect it to be larger but it's not, why?"

I call this difference being a developer who is on call vs being a developer who is not on call

I can't wait to see how these AI models maintain backward compatibility

Yeah the under 31 surprised me. That alone would bring a shortage.

To have easy compatibility with other devices. Why wouldn't they say yes?

That's some fancy words friend.

This smells like when Google released Gemini to have a product in the space.

Because most places never needed kubernetes but used it to put their technical debt on a credit line. So what do you do when they try to collect? Well you just take out another loan to pay off the first one.

You can use similar tricks to sniff auto fill data with arrow keys, a fake pacman game, and hidden form fields using focus.

Oh there's more luck required than that. You have to get lucky many times to win at a 10x stock.

- You have to be lucky enough to find it when it's cheap.

- You have to be lucky enough to hold on to it even if it loses money

- You have to be lucky enough to not sell it when it's at only 5x and hold off for the top

- you have to be lucky enough to have bought enough initially that the return is meaningful to you

These are the thoughts that made me clean up how I invest and stop thinking I'll get lucky at some point just rolling the dice. It's way more luck required than just buying in early.

What's to stop me from installing custom certs and MITM your login session proxying the info. Or an extension to harvest the data after you login. I'm pretty sure if I have root it's game over one way or another. The surface is massive.