timothygold

I made a quick prototype to demonstrate what I think A.I code assistance should be..

https://github.com/hibernatus-hacker/ai-hedgehog

This is a simple code assistant that doesn't get in your way and makes sure you are coding (not losing your ability to program).

You configure a replicate API token from replicate... install the tool and point it at your code base.

When you save a file it asks the LLM for advise and feedback on the file as a "senior developer".

Run this along side your favorite editor to get feedback from an LLM as your working on (open source code nothing you don't want third parties to see).

You are still programming and using your brain but you have some feedback when you save files.

The feedback is less computationally expensive or fraught with difficulty than actually getting code from LLM's so it should work with much less powerful models.

It would be nice if there was a search built in so it could search for useful documentation for you.

This is a simple code assistant that doesn't get in your way and makes sure you are coding (not losing your ability to program).

You configure a replicate API token from replicate... install the tool and point it at your code base.

When you save a file it asks the LLM for advise and feedback on the file as a "senior developer".

Run this along side your favorite editor to get feedback from an LLM as your working on (open source code nothing you don't want third parties to see).

You are still programming and using your brain but you have some feedback when you save files.

The feedback is less computationally expensive or fraught with difficulty than actually getting code from LLM's so it should work with much less powerful models.

It would be nice if there was a search built in so it could search for useful documentation for you.

Hedgehogs always recommend programming yourself to stay sharp.. so this is A.I assistance in a much simpler offering letting you do all of the work as the developer (how it should be).

Maybe this "backup" solution.. developed into commodity hardware as an affordable open source solution that keeps the model and code locally and private at all times is the actual solution we need.

Lets say a cluster of raspberry pi's / low powered devices producing results as good as claude 3.7-sonnet. Would it be completely infeasible to create a custom model that is trained on your own code base and might not be a fully fledged LLM but provides similar features to cursor?

Have we all gone bonkers sending our code to third parties? The code is the thing you want to keep secret unless your working on an open source project.

I think your mistaking programmer productivity with A.I that generates all the code for you allowing you to switch off your brain completely. prompt engineering code is not the same skill as programming and being good at it does not mean you actually understand how code or software works.

I tried cursor for a day or two and then asked for a refund... here's why:

* It has terrible support for Elixir (my fav language) because the models are only really trained on python.

* Terrible clunky interface... it would be nice if you didn't have to click around, do modifier ctrl + Y stuff ALL the time.

* The code generated is still riddled with errors or naff (apart from boiler plate)... so I am still * prompt engineering * the crap out of it.. which I'm good at but I can prompt engineer using phind.com...

* The fact that the code is largely broken first time and they still haven't really fixed the context window problem means you have to copy paste error codes back into it.. defeating the purpose of an in integrated IDE imo.

* The free demo mode stops working after generating one function... if I had been given more time to evaluate it fully I would never have signed up. I signed up to see if it was any good.. which it isn't.

AWS have a ridiculously unreasonable account verification process. Having sent everything they ask for I just keep getting the same big list of criteria without any more feedback on exactly what their issue is.

I hope your not relying on big tech / cloud providers who will suspend your account without warning and have a self hosted alternative.

Why will a financial institution accept a drivers licence, letter addressed to me and bank statements but AWS wont?

I'm so annoyed that I don't care if they don't let me in. It shows you exactly what AWS is really like and that they only serve huge international corporations.

> Forzani said the government needed to acknowledge that there are a small number of "very dominant suppliers" in the cloud market.

That's called a monopoly.

Would it be so hard for some of the smartest people in the world to run their own infrastructure and be more self reliant?

At least they aren't wasting tax payers money on all this total crap.. oh wait.. they are.

I'm surprised see Microsoft buying a competitor to only then shut them down at a later date. It's not like them at all to act like a Monopoly.

Disclaimer this is a rant, and I'm bitter about not becoming am A.I researcher or data scientist for job security reasons.

> Also please don't perpetuate the statistical parrot interpretation of LLMs, that's not how they really work.

I'm pretty sure that's exactly how they work.

Depending on the quality of the LLM and the complexity of the thing your asking about good luck fact checking it's output. It is about the same effort as finding direct sources and verified documentation or resources written by humans.

LLMs generate human like answers by using statistics and other techniques on a huge corpus. They do hallucinate but what is less obvious is that a "correct" LLM output is still a hallucination. It just happens to be a slightly useful hallucination that isn't full of BS.

As the LLM takes in inconsistent input and always outputs inconsistent output you * will * have to fact check everything it says. Making it useless for automated reasoning or explanations and a shiny turd in most respects.

The useful things LLMs are reported to do where an emergent effect found by accident by natural language engineers trying to build chat bots. LLM's are not sentient and have no idea if the output is good or bad.

Hosting a website with an RSS feed is federated. It doesn't solve the problem of finding interesting RSS feeds unless maybe RSS aggregator feeds exist (which they probably do).

The fediverse is a marketing buzzword designed to sell fake p2p social media platforms.

Bluesky -> not actually properly disintermediated or federated. mastadon -> equally misleading.

Bluesky host and manage centralized infrastructure that would break the system if turned off.

If an owner of a mastadon server deletes their server all the connected users profiles get deleted too.

react... over my dead body.

typescript... just no.

Project owners don't really know what they are asking for half the time.. when they move the goal posts after the thing has been built and tested and ready for release * sigh

I'm not bitter obviously LOL

Spending three months arguing over tech stack, frameworks, whether or not to lint but never really worrying about test driven development because that is too difficult for the junior developers.

BTW just use Elixir and self host if in doubt : )

Cloud company products are over engineered piles of crap.

Vendor lock-in..

Oh yeah AWS have an authentication system.. let's just use that and we can always switch to a different one!

eeerm yeah right.

Agile teams... which is a euphemism for let's just start building and figure out the nitty gritty details later.

At least we all sat round a whiteboard doing waterfall and discussed the architecture, requirements and the best way to build things holistically.

Choice blindness.

Also certificate authorities are largely opaque corporations and there is no community verification of actual trust between a community of people sharing content with each other.

It is not unheard of for a CA to get hacked or issue fake certificates to aid hackers.

No beadyw because SSL does not provide protection from A.I bot scrapers or DDOS attacks.. or manage federated digital identity.

I've updated the blog again with some tweaks.

A malicious reputation server could also potentially ok and allow bad requests if that aspect of the system isn't resilient. so reputation server's would also need a reputation system of some kind... and down the rabbit hole we go again.

Another issue that I thought might be pointed out with this idea is...

If you need to interact with a reputation service of some kind for every new client connecting. That would require network calls and traffic correlating to the number of new inbound requests.

So you could DDOS a website by providing requests with junk public key's causing the server to consume resources checking whether those requests are genuine.

As a defense against scrapers and bots I think the idea has some merit. It may be replacing one form of DDOS for another if there is a targeted attack against the authentication bit.

Good point mrkeen.

I'm trying to solve many interconnected problems at once and generating new ones... standard.

Let's say you have a public digital identity... this would indeed make it possible to tie activity back to that identity granted.

Let's say you add a system of trust whereby you get a random public key as the server. The reputation system confirms it's linked to some real identity but doesn't tell the server which one.

Something like that.. it's possible to enhance the privacy to browsing clients by building on top of this idea.

Websites have no way to differentiate between bots or real human beings. The website has to load content to bot's and implement ever increasingly complex resource hungry or expensive defences that are ineffective. Bandwidth is being used up blocking legitimate clients and legitimate clients user experience is deteriorating.

[snippet from article]

This is my blog post on how we can fix the internet by breaking it again. Without needing to wait for so called "solutions" from the same people that broke the internet in the first place..

namely:

Advertisers, Big tech, A.I companies, Social media.

etc.

Articles without references and fact checking seem to be obvious targets for getting flagged. Unfortunately innocent articles that provoke toxic comments seem to get flagged too. I really enjoy hacker news because of the quality of articles that make it in the list. However I do feel that the moderators and maybe some commenters have their own motives. This is a larger free speech issue and not a hackernews issue. There are increasingly less and less decent forums for meaningful discourse.

Yes I'm also in the 1% and I wrote the article... I guess there is the 0.001% that need's addressing first. I don't know if that makes me an entitled narc for writing the article. Genuinely if Wanted attention I could do it in a million other different ways and I just want humanity to evolve.