mickael-kerjean

Having a path for a shared filesystem to be used by agents with strict access control is something I've been working on with my oss work: https://github.com/mickael-kerjean/fdrive, and https://github.com/mickael-kerjean/filestash

I have been working on exactly this with Filestash (https://github.com/mickael-kerjean/filestash), out of millions of users, 99% of them don't seem to care so much about the malleability, it gets very usefull when a company ask for something that 99.99% of people don't care without having to add yet another setting toggle. A recent example was when supporting gobd compliance deployment which required us to sign logs via a vendor who can deliver a qualified electronic signature for which nobody outside Europe would care about

Malleability is a great answer to avoid the death by a thousand feature where you bloat your software because a company asked for something 99% of your users don't care about. That's the path I went with for my open source work with plugins that can run from a wasm runtime to prevent those from doing crazy things [1]. Unsurprisingly plugins represent more than 10x the code than what the core provides.

[1] https://github.com/mickael-kerjean/filestash

> utterly unlike the sort of security in place already for power plants and weapons

As someone who has been working on scada used in those place from one of the biggest firm out there, I was shocked by how they released their software while working there, the thing literally went in nuclear plant but nobody published any checksum, + they were shipping hard drive al around the world full of the software ready to be deployed in those super sensitive environment to the marketing team to publish everything on their joomla website without 0 care to ensure the integrity ofanything. It was sloppy as hell, and while talking to the people who were doing those things, they either did not care, did not know or a mix of both saying there are other layer to deal with security in the chain. At some point I remember an accident in that same software that made a engine do things outside what it should have, that literally killed someone and the email we receive was along the line of: we know journalist are snooping around , you just shut the f up, I left soon after and that company valuation has gone nothing but more up

This got to be the extinguish phase from the EEE Microsoft playbook. Prior acquisition, Github was very liked and very focus, then Microsoft happened in their embrace phase, telling us EEE was something of the past. While trying to show good faith they extended the platform and this is now the very last phase

Filestash (https://github.com/mickael-kerjean/filestash) has a s3 gateway plugin that I made. It proxy the S3 traffic to any downstream storage: SFTP, FTP, another S3, SMB, NFS, IPFS, ...

There is this infamous xkcd (https://xkcd.com/435/) going like this: sociology is applied psychology -> physchology is applied biology -> biology is applied chemistry -> chemistry is applied physics -> physics is applied math -> math is way up there looking down on other fields

I would argue the main reason AI labs have been focusing on programming is to unlock industrial scale automation, next logical step is to solve math as it's the key to unlock everything else. Once you hold the key for math, everything downstream fields become a matter of compute

That's the path I've been on with my Dropbox alternative [1]. The 80% is the fast core that focus on file management, the remaining 20% is coming via plugins which implement one of core interfaces so you can handle your own storage, authorisation, authentication, custom apps to handle file types, ... Fun fact, there is 10x more code in the various plugins [2] catering to everything than the core code that was supposed to be the 80%, unsurprisingly everyone needed a different 20%. Plus I find it cool that the guy who needed a gobd compliant log system that have ton of properties to keep auditors happy enforcing hash chained signed with rfc3161 does not contribute to making the system shittier for everyone else

[1] https://github.com/mickael-kerjean/filestash https://github.com/mickael-kerjean/fdrive

[2] https://www.filestash.app/docs/plugin

Apart from learning, I cant see the point of spending that kind of money and energy to get such awfull token generation speed. Assuming memory bandwith is the bottleneck, is it just a matter of time until we start to see hbm4 based chip able to run qwen3.8 for normal people running at more than 500 token / seconds? Is memory speed the only technological bottlenecks that prevent us from having fast local model?

Kubo is absolutly amazing, is there a group of people interested in keeping kubo alive? If so I'd love to join that conversation

I am working on making the Dropbox launch first comment a reality: https://github.com/mickael-kerjean/filestash It gives you the Dropbox experience on any storage you want inclusive of FTP but also let you access your Dropbox storage via FTP, SFTP, S3, ...

Same as op, I rewrote away from react my Dropbox alternative (https://github.com/mickael-kerjean/filestash) in vanilla js a couple years back around the idea that components are plain exported functions like this:

  export default function component(render) {
      const $dom = xxxx
      render($dom);
  }

example: https://github.com/mickael-kerjean/filestash/blob/master/pub... the only library in there is rxjs to manipulate events in a functional fashion with no imperative code and everything is composed from those simple component functions. The wasm is used mostly as an interface for plugins to both run apps to display various file types like RAW, PSD, TIFF, CDR, .... as those tend to have better support in C than other languages, and on the server side to have a safe environment to execute plugin code.

> From Windows or Mac, this FTP account could be accessed through built-in software.

This is exactly what I've done: https://github.com/mickael-kerjean/fdrive https://github.com/mickael-kerjean/filestash

The premise of the article that "Dropbox was .. a feature, not a Product" is complete nonsense. Remember the infamous FTP guy's top comment during their launch? I have spent almost 10 years working on making that vision a reality [1], and there are entire industries built around some variation of it: digital asset management, managed file transfer, digital preservation software, electronic document management systems, ...

[1] https://github.com/mickael-kerjean/fdrive | https://github.com/mickael-kerjean/filestash

Any plan to release on ollama?

A Drobpox like app that works with FTP and every other storage protocol along gateways to access your Dropbox over FTP, SFTP, S3:

- the desktop client: https://github.com/mickael-kerjean/fdrive

- the web client: https://github.com/mickael-kerjean/filestash

> dropbox clone get off the ground

Thats exactly what Im working on:

- web client: https://github.com/mickael-kerjean/filestash

- desktop sync: https://github.com/mickael-kerjean/fdrive

In line with the infamous top comment during their launch, it does of course support FTP as a storage, and the other way around exposing your Dropbox as a FTP server, while supporting theme so you have a similar look and feel as well: https://www.filestash.app/img/screenshots/theme_dropbox.png + tons of stuff Dropbox never did

Genuine question: how reproducible / usable / verifiable are these architectures from the published documentation? Are they similar to PDF/DWG/PSD specifications, where the format look like an open spec at first sight until you attempt to implement it and realize the crucial implementation details are undocumented?

The infamous attention is all you need paper has this: "The Transformer ... reach a new state of the art in translation quality after being trained for as little as twelve hours on eight P100 GPUs.", the p100 is now quite a bit under 100$ and that's from a time where nvidia was valued 100x less than it is today. Lets wait and see

Building a Dropbox like client that work with every protocol there is: S3, SFTP, SMB, NFS, Azure Blob, IPFS, ...

https://github.com/mickael-kerjean/fdrive

If you are interested in small language model to fine tune, gemma3:270m is quite interesting for its size

> If you want a job that you should care about then work in public service

I did that in the health sector of my local gov, the whole place was full of consultant who either got contracted directly from Oracle, used to work at Oracle before but moved there or took the Oracle pill early on and never got the idea to see how things get done elsewhere. It was impossible to ship anything that's not made of Oracle technologies and that was not an accident but a deliberate construction.

Would love to know the technical details of how you made such an awesome game

I've had the opposite experience. I used to run k8s on bare metal, troubleshooting something at least once a month (DNS going down was a recurring favorite). The breaking point came with the churn in the ecosystem, got bitten by the deprecation of the community darling weave net cni plugin, the killing of the nginx ingress was the nail in the coffin knowing I had way too many annotation tight to the ingress that it would take longer to migrate those than go the ansible way with k8s imposing tight upgrade schedules. While I agree ansible feels a lot more dirty than k8s, I spend much less time on infrastructure, sleeping betterat night and handling things like databases is much simpler

I applied for my oss project Filestash which satisfy all their criterias but never heard back despite the project having millions of users, more than 14000 stars on github and representing more work than a single person can cope with

I've got one set aside to play with v4l2m2m. Once that's working, it should improve video encoding support in my OSS project (https://github.com/mickael-kerjean/filestash)

It's refreshing to see those kind of projects coming back on HN. Pretty awesome to see your example (https://evmar.github.io/theseus/) fit in about 1MB when a lot of those wasm projects can be super heavy (like libreoffice which is 50MB after brotli compression)

I did work on a plugin system for Filestash leveraging wasm. Plugins made with tinygo were 10x slower than the same code in rust or c compiled to wasm

[1] https://github.com/mickael-kerjean/filestash/blob/master/ser...

Tell that to CVEs