Stefano Maffulli
maffulli.net
Stefano Maffulli
@maffulli.net
Also an architect, GIS enthusiast, sailor.
Reposted by Stefano Maffulli
At the risk of starting the flame war to end all flame wars...

Modern LLMs (GPT-5.1, Claude 4.5, Gemini 3) produce excellent code and can be a significant productivity boost to software engineers who take the time to learn how to effectively apply them - especially if used with coding agent tools
November 27, 2025 at 7:55 PM
Reposted by Stefano Maffulli
DeepSeek-Math-V2 means we now have an open weights (Apache 2) model that can achieve gold medal performance on this year's International Mathematical Olympiad - previously proprietary models from OpenAI and Google DeepMind had achieved that

689GB from Hugging Face! simonwillison.net/2025/Nov/27/...
deepseek-ai/DeepSeek-Math-V2
New on Hugging Face, a specialist mathematical reasoning LLM from DeepSeek. This is their entry in the space previously dominated by proprietary models from OpenAI and Google DeepMind, both of …
simonwillison.net
November 27, 2025 at 4:03 PM
Reposted by Stefano Maffulli
some might know how @roost.tools is approaching open safety tool development, but many more may not! @annebdh.bsky.social lays it all out here, including our governance structure, how people can get involved, and how working in the open doesn't mean more abuse!

roost.tools/blog/open-by...
Open by Design: ROOST's Approach to Safety Tool Development
Robust Open Online Safety Tools or ROOST is a new non-profit entity designed to address the urgent need for accessible, high-quality safety tools in the rapidly evolving digital landscape.
roost.tools
November 26, 2025 at 5:23 PM
Reposted by Stefano Maffulli
Fifteen Years

xkcd.com/3172/
November 26, 2025 at 10:32 PM
Reposted by Stefano Maffulli
I write about some interesting developments in the ATmosphere (and the broader universe of decentralised and federated social media). Very intrigued by #Eurosky, especially the idea of shared moderation infrastructure. I will be following (and hopefully writing about!) this with a lot of interest.
🇪🇺 Eurosky is building EU-hosted social media infrastructure on ATProtocol. Priorities: Commons for Content Moderation (CoCoMo), decentralized architecture, and reducing Big Tech dependency.
Public funding + civil society collaboration = digital sovereignty. openfuture.eu/blog/eurosky...
Eurosky dawns: Building Infrastructure for Sovereign Social Media – Open Future
The Eurosky initiative is building European social media infrastructure on the ATProtocol, representing a new step in Europe's digital sovereignty agenda as policymakers seek alternatives to Big Tech ...
openfuture.eu
November 27, 2025 at 11:01 AM
Not sure how to take this: Mastodon already exists and it's a European project... and it's struggling to take off. I don't think that the problem is the ActivityPub protocol. What does ATProtocol have that will lead to wider adoption?
🇪🇺 Eurosky is building EU-hosted social media infrastructure on ATProtocol. Priorities: Commons for Content Moderation (CoCoMo), decentralized architecture, and reducing Big Tech dependency.
Public funding + civil society collaboration = digital sovereignty. openfuture.eu/blog/eurosky...
Eurosky dawns: Building Infrastructure for Sovereign Social Media – Open Future
The Eurosky initiative is building European social media infrastructure on the ATProtocol, representing a new step in Europe's digital sovereignty agenda as policymakers seek alternatives to Big Tech ...
openfuture.eu
November 27, 2025 at 9:37 AM
Reposted by Stefano Maffulli
Galloway thinking outside the box about GLP-1s (not sure why he pits it against AI...)
America’s Best Bet | No Mercy / No Malice
America is now one giant bet on AI. If not for the Magnificent 10, the markets would be flat for the year. There’s a different technology that offers more asymmetric upside. If we want to transform th...
www.profgalloway.com
October 26, 2025 at 5:20 PM
1/ The idea that licensing deals can scale to supply training data for cutting-edge AI systems is deeply flawed. As OpenAI, Google, Anthropic and others show: licensing deals so far cover only a tiny sliver of the content needed.
The False Hope of Content Licensing at Internet Scale - ProMarket
Is there a world where AI developers could get the training data they need through content licensing deals? Matthew Sag argues that content licensing deals between developers of artificial…
www.promarket.org
November 26, 2025 at 7:04 PM
"What have you done to secure the Open Source packages you rely on since Log4j?" asks the US and the EU. What do you think the big tech companies will answer?

about.gitlab.com/blog/gitlab-... #security #CyberResiliencyAct #OpenSource #sustainability
GitLab discovers widespread npm supply chain attack
Malware driving attack includes "dead man's switch" that can harm user data.
about.gitlab.com
November 26, 2025 at 5:30 PM
OpenAI picked the new ROOST Foundation rather than other hosting foundations. Given the competition out there, the Foundation’s leadership deserves praise. https://www.maffulli.net/2025/10/31/roost-foundation-releases-ai-safety-tool-with-openai/ #ai #safety
ROOST Foundation releases AI safety tool with OpenAI
The ROOST Foundation delivered a landmark accomplishment in the Open Source ecosystem with the release of gpt-oss-safeguard, with OpenAI.
www.maffulli.net
November 26, 2025 at 2:39 AM
Let the neo-Luddites dream their dreams while @Mozilla.org and others stay busy shaping the future technology the way we want it to be: open and controlled by its users. https://www.maffulli.net/2025/11/24/internet-is-changing-browsers-need-to-adapt/ #ai #freedom #mozilla
Internet is changing, browsers need to adapt
The new Firefox AI panel has been met with vocal resistance from parts of the community. Change is hard, and the debate surrounding AI is proving to be highly polarizing. My message to the Mozilla leadership: Stay resolute. You are on the right side of history.
www.maffulli.net
November 25, 2025 at 3:21 PM
The Italian Data Protection Authority is swept by a huge scandal. The last act is the resignation of its administrative chief youtube.com/shorts/oqqFy... #privacy
Istitutional scandal within the Italian #privacy #dataprotection authority.
YouTube video by Simone Aliprandi
youtube.com
November 22, 2025 at 8:54 PM
A tag to distinguish content that was created by human or generated by AI. How hard can this be, right? How would you address it? Check the draft proposal at IETF.
datatracker.ietf.org/doc/draft-wi...
#AI #IETF #data #content
Classification and Tagging System for Digital Content to Preserve Clean Datasets for Machine Learning
This document specifies a classification and tagging system designed to identify and preserve the provenance of digital content (text, audio, video, and other media) to ensure the integrity of…
datatracker.ietf.org
November 20, 2025 at 7:23 PM
Internet was built to survive a nuclear attack: no single point of failure.The plan was perfect but ...Cloudflare is down and it took a large chunk of the internet with it.
November 18, 2025 at 2:32 PM
Reposted by Stefano Maffulli
Why AI writing is mid
How the current way of training language models destroys any voice (and hope of good writing).

www.interconnects.ai/p/why-ai-wri...
Why AI writing is mid
How the current way of training language models destroys any voice (and hope of good writing).
www.interconnects.ai
November 17, 2025 at 3:41 PM
Reposted by Stefano Maffulli
Let's rewrite this lede:
After more than a decade failing at regulating big US tech, the EU gives up.
www.nytimes.com/2025/11/17/t....
Europe Begins Rethinking Its Crackdown on Big Tech
www.nytimes.com
November 17, 2025 at 5:53 PM
Reposted by Stefano Maffulli
This week, our President, @camillefrancois.bsky.social, visited the Vatican for a discussion on protecting children in the digital age. We were very humbled to have been invited to participate in such an important meeting and at such a pivotal time.
November 14, 2025 at 7:30 PM
@paulk.bsky.social Keller @openfuture.bsky.social read the GEMA v. OpenAI judgment. In his view, the judgment reaffirms the TDM exceptions enable AI training—provided the training doesn’t lead to memorization. openfuture.eu/blog/g... #AI
GEMA v OpenAI: Memory is Fragile. Garbage Lasts Forever. – Open Future
GEMA's lawsuit against OpenAI has reignited the debate about AI training and copyright. This post examines the relationship between AI training and the TDM exceptions.
openfuture.eu
November 14, 2025 at 3:02 PM
Reposted by Stefano Maffulli
As age verification laws proliferate around the world, people have predictably turned to VPNs to protect their privacy. Instead of taking this as a sign that surveillance isn't popular, lawmakers have decided they might as well ban VPNs altogether. www.eff.org/deeplinks/2...
Lawmakers Want to Ban VPNs—And They Have No Idea What They're Doing
Remember when you thought age verification laws couldn't get any worse? Well, lawmakers in Wisconsin, Michigan, and beyond are about to blow you away.It's unfortunately no longer enough to force
www.eff.org
November 13, 2025 at 6:42 PM
I don't have the patience to debug Google Calendar #Android to understand why it's dropping my Open-Xchange calendar after it syncs to the phone. Pure rage for this blatant attack to the most elementary interoperability.
November 13, 2025 at 1:11 PM
Reposted by Stefano Maffulli
We are hiring a postdoc to build open-source tools for deep learning–based antibiotic discovery in @chzimmer.bsky.social lab at Institut Pasteur.
Part of the AI4AMR consortium with HZI and University of Würzburg.

Details: research.pasteur.fr/fr/job/open-...
Open-Source Developer for Deep Learning Antibiotic Discovery Platform - Research
Introduction Antimicrobial resistance (AMR) is one of the most urgent global health threats projected to cause up to 10 million deaths per year by 2050. The AI4AMR project brings together teams from I...
research.pasteur.fr
November 13, 2025 at 11:45 AM