Pinned post

It's time!

- ๐Ÿ‡ฎ๐Ÿ‡น geek in his mid-thirties, based in ๐Ÿ‡ณ๐Ÿ‡ฑ

- ๐ŸŽ“ M.Sc in computer engineering.

- My current job is about fixing and automating global supply chains, one line of code at the time, but I have worked in a wide range of industries over the past (nearly) two decades.

- My hobbies often involve automating everything around me.

- :linux: user since 2001. My experience as a Linux admin started back in a time when I used run my IRC and Apache servers on a repurposed Pentium 1 under my bed, and it still took about 10 ๐Ÿ’พ to install a full Slackware system.

- :arch: Linux and rolling release enthusiast.

- ๐Ÿ›  Creator and main developer of (, an open-source (mainly :python: and :vue:), general-purpose platform/framework to automate everything - from smart devices, to cloud services, to robots, to DevOps operations, to everything in between. With hundreds of available integrations, you can think of it as IFTTT+Tasker+SmartThings on steroids, scriptable, and runnable on almost any device. Or maybe like HomeAssistant's lighter brother.

- Admin of, a Mastodon instance where I may talk a lot about Platypush, automation, programming, electronics and maths. I tend to write a lot, so if you're looking for an instance with a 10,000 characters per toot limit...

- Looking for relays with instances dedicated to similar topics. My dream would be to build an experience, when it comes to , that is akin to curated lists, where admins can create curated federated experiences for the users on their platforms, rather than the open-to-everything overwhelming stream of toots on the federated timeline that most of the relays provide nowadays.

- ๐Ÿค– Machine-learning enthusiast. I have published a book on it, with simple computer vision exercises that can be run on a , and I did some academic research back in time where neural networks were still a green field, and I never stop learning new stuff.

- ๐Ÿงช๏ธ Physics, chemistry, biology, maths and astronomy enthusiast.

- ๐ŸŽต Music addict, ๐ŸŽธand ๐ŸŽน player, and occasional composer/producer You can find some of my music here and here

- I may often write about random politics/economics/philosophy. I may sometimes be very passionate on topics such as open-source, open data, open protocols, tolerance and social inequalities. I mostly belong to the progressive/social-democratic field. You are welcome to try and change my mind, as long as you do it in a civilized and data-driven way.

- ๐Ÿ„ and ๐Ÿ›น๏ธ rider. And, as a good Dutch resident, ๐Ÿšฒ enthusiast.

- ๐Ÿ‘ช Full-time father.

Rephrasing the article: " has become such a poor search engine riddled with trackers, sponsored content and SEO black magic, that people have resorted to search for stuff directly on Reddit and Quora.

Therefore, Google has decided to pivot its mission to that of a meta-search engine that displays results fetched directly from these platforms and other forums (plus trackers and ads), and stop pretending that it cares about being a search engine".

I've finally made the jump from to a self-hosted instance.

What held me back so far: mainly the UI, Instapaper's interface is extremely well-polished and optimized for readability (both the website and the app), and after using it for several years I couldn't settle for any of the alternatives. However, I also realized that a non-optimal font-family, an intrusive footer or a header nav with an inexplicable bright cyan background are nothing that a Greasemonkey script can't fix.

What kept me on Wallabag: feeds, feeds, MOAR feeds!! An Atom feed for my unread items and for each of my tags is exactly what I wanted the most from a framework to save my links. I can now read my unread items (already conveniently distilled) on any of my feed readers, and easily set up automation routines on my saved articles by simply polling a feed. And the API is surprisingly quite mature, so replacing my Platypush Instapaper hooks with Wallabag API calls has been a quite smooth process.

is now employing a new strategy for its propaganda: registering domains that look exactly like major Western media outlets, but whose purpose is to spew out the usual about .

An example: the domains and were registered just a couple of days ago, they looked exactly the same as the legitimate Guardian and ANSA websites (even the names and photos of journalists were the same), but their DNS records pointed to Russian IP addresses, and they published the usual FancyBear-powered fake news on Ukraine.

The sophistication of these campaigns is much higher than old ones - which mainly relied on setting up bots and fake accounts on existing platforms such as Twitter and Facebook, not on setting up their own infrastructure that mimicked that of major news outlets.

Buying their own domains, and hiring designers and web developers besides the usual army of fake news specialists, costs money, and quite a bit of it.

The fact that Russia is happy to spend hundreds of thousands of dollars on these campaigns to shape public opinion abroad shows just how desperate they are - and why we need to keep fighting their regime until it crumbles. As I said earlier, Putin is digging his own grave in Ukraine.

I also wish that IT specialists in Russia could have actual IT jobs in actually competitive IT companies, instead of maximizing their salaries only if they decide to work for the Kremlin's disinformation network, benefiting nobody but Putin and his lies.

In the meantime, ALWAYS remember to check the source before sharing an article: differences in the URLs may be subtle to notice, but they are really the best way to ensure its authenticity. And, as usual, ALWAYS distrust posts that only share screenshots without links to real articles.

almost got it right, but he forgot to expand the axiom.

in a mature industry (i.e. one where the product is mature, entry barriers are high and a few conventions and market players who have consolidated their position don't incentivize further innovations or disruptions) deters .

Capitalism without competition and without regulation favors the consolidation of capital and power in monopolies and oligopolies.

Monopolies and oligopolies are, by definition, the opposite of level playing field and open competition.

If capitalism without competition is exploitation, then capitalism without regulation inevitably leads to exploitation.

I'm on the side of the telecom operators on this one (and it doesn't happen that often).

Scaling up networks and band at the pace we do today (especially if we start adding smart cars and devices to the picture, not to mention the metaverse and whatever other network-intensive thing big tech decides to come up with) isn't cheap. Providing faster and faster networks to more and more people consuming more and more network-intensive loads isn't cheap, nor it is environmentally sustainable.

Big Tech so far could get away with treating the network supply basically as an unlimited resource, not even accounting for it in their "CO2 budgets". You want to add another 100 KB of trackers and scripts to a web page rendered by millions every day? You want to render ads as videos on every page view? You want to encourage people to share images or videos where they could have shared simple text? You want to run 100 requests to render a single-page webapp, even if not all of them are needed? You want to turn smartphones into devices that, even when idle, share a few MBs of data back with Google or Apple? You want to force everyone into virtual reality just because you invested a lot into it, even if the underlying networks aren't ready to handle the load? Sure, what's the problem?

The problem of treating any resource as unlimited is that, at some point, that resource becomes scarce, and your economic model isn't ready to cope with it.

I hope that gets indeed forced to share some of the network costs of its network-hungry applications with the network operators. It may finally encourage them to treat network and computing resources as finite resources, while building more efficient applications that can run on more devices in the process.

The only good news is that the far-right nightmare won't last long.

Meloni has spent most of her political career yelling from stages. She has no political depth whatsoever, and her party's stances on foreign policy, economics and energy management are just afterthoughts quickly scrambled together during the campaign itself. She has been catapulted into the current position just because she led the only opposition party to Draghi's, not for any merits of hers.

She is supported by two parties led by two unreliable and opportunistic leaders who are notorious for being addicted to electoral campaigns but not so much for governing, and for seeking the first useful occasion to unsaddle the current leader.

Moreover, those three leaders agree almost on nothing. From the war in Ukraine, to how to fund the much needed reforms in Italy, to how to reform the welfare plan previously put together by the M5S, to how to interact with the rest of the EU, to international alliances, the upcoming coalition is set to run divided on almost everything.

Just wait for the first big budget-related decision, or for the upcoming cold winter, and this coalition is doomed to fall apart as a straw house under the wind.

Then Italy will just be ruled by the next technical government appointed by someone else, in order to fix all the mess left behind.

And that's what Italians deserve. My people don't know how to vote. They keep voting out of protest rather than out of actual ideas, and they don't feel responsible for their decisions. They can't even pick a government that manages to stay in charge for more than a year or two. So they just deserve somebody more mature than them to take decisions for them whenever their elected hollow pipers fall.

UK politics condensed:

- Vote for a party that turns London into the biggest laundry machine of money in the world, deregulates everything that has to do with finance, removes the caps on bonuses for managers, lowers the taxes for corporations and for the highest earners, all while following the largely debunked "trickle down" economic theory.

- When nothing "trickles down" and people start to get pissed, blame the EU, blame the migrants, blame the left - blame everyone, but never, ever put the damaging Thatcherian-Reaganian doctrine in discussion.

- Rinse and repeat until nothing but the shadow of a country is left.

I have only played with for a few minutes, and I can't tell how much I love this tool.

As somebody who plays with tons of JSON datasets, and still receives tons of CSV files at work, this is exactly what I needed to make my life easier.

So far, performing some complex queries on JSON or CSV datasets required me to set up some temporary relational database and dump the data into it just so that I could query it. Or spinning off a Jupyter notebook or a Python script that could parse the data and run the queries I wanted via code.

All these workarounds are probably a thing of the past now that e.g. getting the countries with a highest number of cities from the dataset is something as simple as this command line:

columnq sql --table ./cities.json \
'select country_name, count(*) as cnt from cities group by country_name order by cnt desc'

"The whole Earth is going in ruin and the young generations should make more noise about it. A small hand of greenwashing isn't sufficient, we need to put the whole model of economic development in discussion".

As an atheist, I would have never expected the more leftwing thing of the Italian electoral campaign to come from a Pope. And I couldn't agree more with him.

has just released, a new open-source model for speech detection.

While after a couple of tries I'm impressed by its accuracy (you need to use the small model or a larger one if you want enough precision though), I'm also still unimpressed by its resource usage and performance.

The small model took ~30 seconds to process an audio file with 2 seconds of speech on my 6-year-old laptop with an i7 CPU, and in the meantime it used up more than 4 GB of RAM.

Mozilla's model was also heavy when I last used it ~1 year ago, but not THIS slow (although it was also slightly less accurate).

For now I definitely see the use-case for using OpenAI's new model for offline transcriptions, but they are still very far from being used for real-time applications such as voice assistants.

I'm still looking for a good open-source model that can be run on a RaspberryPi as a stable voice assistant. Ideally, it needs a small and simple model that can be used for hotword detection (I used to use Snowboy, but that project is now dead), and a more complex model to be used once the hotword is detected in order to transcribe the speech. And the audio transcription needs to be done within max 5 seconds in order to be compatible with the real-time expectations from a voice assistant.

Ideally, it needs to only include the model, not a lot of bloat around it that makes it harder to embed it - so is excluded.

So far, I haven't found any such model. My RPi still run the Google Assistant's push-to-talk script that I adapted into Platypush years ago, and a Snowboy hotword detection model that I managed to train before the project was shut down. If anybody knows of better solutions that could cut this last dependency on Google, I'd be happy to try them out.

p.s. there's something like this apparently:

The UFS apparently used to be open-source, and now the whole Github project consists of a README that invites people to pay $119 for a license and a Zoom call where the author shares the code and the instructions to install it.

I want to build something really FLOSS also to make sure that these borderline scammers who profit from a real-world need and exploit the open-source community have no bread left to eat.

Show thread

Today I feel the insatiable urge of building a tool to scrape the shit out of and export my timeline into an feed, or at least something that can be easily integrated into the .

I haven't posted anything on my Facebook profile for months nor accessed their website unless some friend shared a direct link to a photo or a video with me. Granted, I feel like the Fediverse is a much healthier place that doesn't make me feel as guilty as if I were chain-smoking and consuming junk food while driving a huge CO2-spewing SUV.

But, even if I met a lot of amazing people here, and I have even managed to bridge Twitter profiles and content from a vast trove of RSS feeds, and I have built bots that take a lot of interesting content directly to my door, and I have even managed to keep messaging my friends on Messenger/WhatsApp through Matrix and Bitlbee bridges, there's still an uncomfortable truth that doesn't make me sleep at night: Zuckerberg is still holding most of my family and friends as hostages, they will probably never move to the Fediverse, and I'm missing out on the lives of my loved ones (as well as on a lot of interesting events happening around me) because that content is behind a huge impenetrable wall.

I'm sick of hearing "Facebook should be compelled to federate, or at least open up their APIs for personal usage, but we don't know where to start". Or "the is great on paper, but it's hard to enforce". If regulators don't take the matter into their hands, then I will. And, if Facebook dares to sue me or lock my account, I'm ready to sue them back for violating the DMA. I'm ready to take this matter in front of courts and spend my money on lawyers, because I want Facebook and their highly immoral "high switching costs" strategy to die amid the worst conceivable pains in this universe - or at least I want them to be forced to open up the data of my loved ones.

I was looking around for some up-to-date Facebook scrapers, but all I could find was this project (which only scrapes public pages) and some commercial solutions that provide Facebook scraping for profiling and ads purposes (which makes me wanna spit and puke on the people behind those businesses for proudly showing off the worst that a human being can be capable of and making a profit out of it).

I made a Facebook scraper around 10 years ago, but back then their pages were relatively simple, and a bit of beautifulsoup scripting was enough to scrape the shit out of them. I've now taken a look at the developers console while browsing the website, and I've been horrified by how much effort they've put to prevent exactly what I was trying to do - the whole Facebook feed is basically a bunch of <script type="application/json"> tags that download some custom minified JavaScript for each post, that in turn is used to decrypt some other JSON requests.

So I'm appealing to all the hackers and tinkerers out there: are there FLOSS projects that already do what I'm trying to do (basically allow you to sign in to Facebook with your account, get an access token, and scrape posts and comments from your own timeline)? If not, are there any volunteers out there who would like to join forces with me in a new dog-and-cat war with Facebook - starting with reverse engineering whatever mechanism they've put in place to obfuscate the HTML on their timelines?

You don't need to stick to , or any other cloud just to get new music suggestions and new releases delivered to your doorstep.

Spotify's discover weekly and release radar playlists are definitely a useful feature for those who want to discover new music, but replicating such features isn't rocket science.

In this article I show how to generate your own discover weekly and release radar playlists (as well as keep a local record of your listening history) using , feeds, the APIs, some scripting and a pinch of SQL.

As an Italian (even if living abroad), I'm very worried for the winter that is awaiting my country.

Italy is the EU country that relies the most on Russian gas, after Germany and some eastern European countries.

And France just announced that it may cut its energy exports (which constitute about 4-5% of Italy's mix) amid concerns about the capacity of its aging and strained nuclear plants.

Algeria promised to make up for the losses by exporting more gas, but Eni just confirmed its serious doubts about Algeria's capacity of supplying so much gas.

Overall, Italy really faces the risk of getting 15-20% of its energy supplies cut over this winter.

But I also see a silver lining.

This is the occasion for people to soberly acknowledge that there's no such thing as a painless green revolution, and whoever promised such a thing was nothing but a hollow piper.

There's no such thing as migrating to a sustainable energy mix while we keep burning gas, while people keep moving around in cars that burn fossil fuels, and while Italy's energy mix is completely based on the import of fossil fuels. And if your house still relies on gas for cooking and heating, and you've never consider buying an induction cooker or a heat pump, then you're part of the problem.

My country of NIMBYs may also finally realize that there's no point in opposing nuclear plants while a lot of the energy we import comes from nuclear plants just across the border. They may also finally realize that nobody should give a fuck about their opposition to wind turbines because of concerns about noise or landscape views, when the alternative is thousands of homes lit up by candles.

There's a lesson to be learned here, and it's a lesson that I've been trying to convey for years: changing the way we produce and consume energy will require change and sacrifices at all levels. Opposing the construction of plants that help a country be less reliant on energy imports, or have a greener energy mix, is just going to harm a whole country for the narrow interests of a small minority of selfish NIMBYs. And expecting somebody else to "migrate" while we don't care about making our houses more energy efficient/self-sufficient, or making sure that our way of moving around is more sustainable, is pure idiocy.

I wish that these lessons were conveyed without the pain that my country will soon face. But sometimes pain is a required step towards acceptance.

We are taking incremental steps to build and strengthen the support for free and open source. facilitates the open development of software projects from the Commission as well as the other European Union institutions.
@EC_DIGIT_director_general at

Hey Fediverse!

The @owncast project could really use some help! ๐Ÿ™

At the moment the lead dev @gabek is having to handle an awful lot at once. The platform would be much more sustainable if there were more people on the team.

Everyone who wants to help is welcome, but of course there are some specific areas where help is especially welcome:

Please get in touch with @owncast or @gabek if you want to help out!

For those who don't know, OwnCast is a Twitch-style livestreaming and chat service for the Fediverse. You can follow OwnCast accounts from Mastodon etc and you will see a post in your timeline when the account goes live. There's a site about it at and a directory of featured streams at

Boosts welcome!

#Fediverse #OwnCast #HelpNeeded

One of the arguments in favour of surveillance capitalism is the great usefulness of cloud-based ML predictions.

After all, who can deny the usefulness of photo apps that automatically recognize faces, detect your speech or help you making sense of the deluge of information in a social feed?

The argument usually goes like this: these features require large neural networks, which in turn require a lot of computational power to train the models, and a lot of memory and disk storage in order to load and save those models.

You can't do such things on small devices that run on batteries. Therefore your phone *HAS* to send your data to servers if you want to get those features. Otherwise, you just won't get those features.

Except that... What if this whole argument is bollocks?

(Private Optimal Energy Training) proves that you can run both the training and the predictions locally, without compromising neither on precision, nor on performance.

After all, the really expensive part of training is back-propagation. POET breaks down the back-propagation performance issue by quantizing the layers (so real-number large tensor multiplications get reduced to smaller multiplications of integer tensors, without sacrificing precision too much), and a clever way of caching the layers that are most likely to be needed, so we don't have to recalculate them, without caching everything though (which would be prohibitive in terms of storage).

The arguments in the paper sound very convincing to me. The code is publicly available on Github. I haven't yet had time to test it myself, but I will quite soon - and try to finally build an alternative voice assistant that can completely run on my phone.

As a crypto-skeptic, I'm following the merge with interest.

On one hand, shifting from PoW to PoS really feels like "so far lighting up a cigarette required burning a whole tree, now we've figured out that these things called lighters are a bit more efficient".

On the other hand, the high initial stake required in order to make cheating financially inconvenient will just make the network more and more centralized - and only those who already have a lot of money can afford to be part of it. So much for the dream of decentralization.

Show older

A platform about automation, open-source, software development, data science, science and tech.