cx42net

The ones that loves the technicity, the challenge of building something, putting complex things together, the puzzle part of creating something.

The ones that wants to have stuff build to use them, were coding was just a mean to an end. (DHH mentioned in his podcast with Lex Fridman, that he is more in this category)

For the later, AI is the perfect tool that could come around, removing that barrier that was stopping them to achieve their needs (which you appear to be part of, I'm saying this without no jugement)

The writer of the blog post is in the former category.

A simple API that converts HTML documents to PDF or images. Used by more people that I would have hoped for.

I tried to follow Linux mentality ; do one thing, but do it well. I can only hope that I'm reaching it with PDFShift :)

It's built using Python (Sanic Framework) with Pyppeteer (Python port of Puppeteer)

Instead of arguing on the bad this brings, we should push our authorities to enforce mechanism that still respect the anonymity of a person.

France's "France Identité" started to implement an approach that respect anonymity by using what they call a "double anonymity: a principle that ensures that neither the adult site nor the age-appropriate provider can know who is accessing what. Using advanced cryptographic technologies—digital signatures and zero-knowledge proofs—the system generates proof of age without revealing identity or browsing history." (source: https://entrevue.fr/en/police-justice/verification-dage-sur-...)

French's state should already have these informations from us (as a French citizen) so asking them to provide a certificate that guarantee my age seems like an obvious choice, and giving that digital certificate to any site without having to tell the issuer where I'm going is, I think, the best solution for the worst project.

(There a lot of news about this, but mostly in French. If you are curious:

- https://www.franceinfo.fr/societe/pornographie/l-arcom-impos...

- https://www.frandroid.com/culture-tech/web/2962099_france-id... )

At first, I thought I had wrongly explained myself, so I checked what I wrote. Maybe it wasn't clear but I'm not looking for consciousness in the LLM era, but more about the philosophical research on what defines consciousness and the parallels that can be done with the LLMs era (parallels not in the sense of similitude, but in the sense of things that would appear similar but are not).

In a more generic and simple example, LLMs are good at producing language that resonates with us, and because of that, could be believed they have thoughts (they can write things like "I think you should .."), even thought they are not capable of thoughts, and are only producing the next logical word one after the other.

In the same sense, it would be interesting to read about philosophers that tried to define what consciousness is (or is not) and their age isn't relevant to today's LLM era. The connection can be done by the reader on what consciousness is, and what LLM is not, in the same vein as what thought and language is between our species and LLMs.

I'm not sure this more clear ^

A real-life scenario would be to refuse your snail mail because someone added an extra line in your address.

The only plausible reason would be to avoid polluting the data (like faking the ref= for marketing purpose) but that would assume you are already tracking the ref value for other services, so in a way, you either need it or not, but you can't have both.

If you don't actively track query strings, having them or not has no impact.

The "France Identité" use the identity card of the holder and generates a cryptographic proof of the age. The website wanting to check the age of the holder request the cryptographic proof, validates it against France's public key without querying government websites (that way, the government won't know you are trying to access pornhub).

This is, I believe, I concrete and working solution to age verification without having to provide any sensitive information to any website that will surely loose it in an breach.

Source (in French): https://www.frandroid.com/culture-tech/web/2962099_france-id...

And just with that hypothetical, you scratch the whole effort Bluesky is currently doing to offer an open protocol ? Which is the opposite of almost (Mastodon) all other social networks ?

Come on..

It honestly feel like this was written by someone being paid by a competitor, just to discredit the service.

Of course they will provide an infrastructure to ease users on to creating and using their service, they can't just provide an empty shell just for the pure purpose of decoupling the protocol and the service. You can't blame them for that (well, you can apparently..)

This.

(Thank you!)

I had the need for such a service back when I was working on ImprovMX.com. The checking system I had implemented back then wasn't efficient and had a few false positive.

At the time, I thought the market for such a need was very low so I didn't bother, but now that I'm working on https://getfernand.com, I realize I also need it. And with the advent of more and more products thanks to AI and the enthusiasm around tech, the market is growing, so it might be worth the effort to release it as a side project.

I took it as a challenge, on two simultaneous front: Technical, and AI.

On the technical side, the aim is to have a highly efficient script able to process millions of DNS record at the lowest time possible. I had to rewrite the script a few times and see how I could optimize it, but my tests lands me at around 500 QPS for now.

I'm testing against Cloudflare's DNS, and I know they have a limit at 1500 QPS, so I can't go much higher (per server), but once I reach that limit, I'll grow horizontally.

For the AI side, my approach was as follow: Use a standard web framework (Sanic in my case), build the API with a proper database structure, and once I'm satisfied with it, ask Claude code to write the OpenAPISpec file.

Once I had it, I used Mintlify to write the doc for this, and the help of Claude to have nicely written guides (rate limit, authentication, etc). Then, I asked Claude to generate the landing page, and the dashboard system.

So, except for the API and the core of the service, the other parts was mostly written in AI (I fixed some issues, improved the code Claude gave me, but that's all).

Less than three weeks after, the service is almost ready to go live. I still need to clean the code in the Dashboard part, and finalize the touches on the API server, but it can be used already (and it's used by us at Fernand).

Feel free to share your comments :)

I agree that the best thing would be to buy all the various keyboard I hesitate, use them for a while, and finally decide which one is best ... but it's costly :p

One thing that makes me hesitant to buy one of these, is that I'm hoping that with one such as the Moonlander or the Glove80, I can replace the used part easily, which should cost less in the long run (but maybe I'm wrong here).

Repairing mine is not possible, it's the wrist support that is worn out and it's glued to the keyboard...

I think I've narrowed down to the following:

- Glove80

- Microsoft Ergonomic Desktop

- Perixx : PERIDUO-505

- Mobility Lab : Clavier Sans Fil Ergo K30W

- Kinesis MWave (It's QWERTY and I need an AZERTY but I believe I can move the keys around).

Thanks for the advice!