Seine reopens to Paris swimmers after century-long ban (lemonde.fr | comments) After a ban lasting over a century, swimmers are now allowed to dive into the Seine River in Paris. This long-awaited reopening comes as part of a broader initiative to promote aquatic activities and rejuvenate the river, previously deemed unsafe due to pollution. The authorities have implemented safety measures to ensure a safe swimming experience.

• People living in houseboats along the river have been swimming in the Seine forever. And every summer, young people, possibly a little drunk, jump into the water as a dare from bridges near Notre-Dame. But yes, this is more mainstream and open to all so it's kind of big news.

Europe's first geostationary sounder satellite is launched (eumetsat.int | comments) The Meteosat Third Generation Sounder 1 satellite has been successfully launched, marking a significant advancement in Europe's meteorological capabilities. This geostationary satellite will enhance weather observation and forecasting, contributing to improved climate monitoring and environmental data analysis.

• I think this is the first time I have seen the .int tld used.
• Naive question: what’s the benefit here to scan downward (from the satellite position) over upward, from the ground?
• Canada should be next.

Why AO3 Was Down (reddit.com | comments) AO3 experienced an outage due to a significant increase in traffic, overwhelming its servers. The platform's team worked quickly to address the situation, but users reported accessibility issues during the downtime. The incident highlights the challenges of managing server capacity in response to sudden user demand.

• Hacker News helps me everyday break my information bubble. Archive Of Our Own is something that I wouldn't walk into when wandering through the internet
• For anyone who feels like looking up exactly what this bookmark was pointing to: I did, and very much wish I hadn't!

How to Network as an Introvert (aginfer.bearblog.dev | comments) Aishwarya Goel offers a practical guide for introverts on effective networking strategies. The post emphasizes techniques like dressing with conversation starters, preparing with news snippets, and controlling first impressions through body language. It aims to empower introverts to connect meaningfully without compromising their nature.

• what about just owning your own lack of confidence as a strength? performative confidence seems dishonest and a bit like cheating to me. also on that matter, why is confidence == good?
• Practical tips - thanks! I like the idea of baiting with a “whatzit” item

macOS Icon History (basicappleguy.com | comments) The article chronicles the evolution of macOS system icons over the past several decades, detailing design changes, aesthetic trends, and the technological advancements that have influenced iconography in Apple’s operating system.

• I find the 2025 versions to be a nicer looking than pre-2025 variants, so it’s overall an improvement. But I also find the 2014 to be usually a lot better (clearer and more obvious). So incrementally it’s an improvement, but historically still worse.

X-Clacks-Overhead (xclacksoverhead.org | comments) X-Clacks-Overhead is a custom HTTP transmission header inspired by Sir Terry Pratchett's works, particularly the 'Clacks' system in his Discworld series. It serves as a posthumous tribute to keep his legacy alive. The header can be used subtly by web authors to commemorate individuals without affecting site performance. Notable users include Mozilla and Debian.

• The thing that struck me about "GNU John Dearheart" was how it feels like it _really_ deeply captures hacker culture, like Pterry wasn't just referencing the culture, but that he really got it. Which is remarkable, because he gave me that impression about many, many topics. Such a loss.

The History of Electronic Music in 476 Tracks (1937–2001) (openculture.com | comments) This article provides an extensive overview of electronic music's evolution from 1937 to 2001, presented through a curated list of 476 tracks. It highlights pivotal moments, influential artists, and technological advancements that shaped the genre, offering insights into its cultural significance and development over decades.

• Sadly, no mention of Louis and Bebe Barron, who together created the first all-electronic soundtrack for the 1956 movie "Forbidden Planet". This was before the invention of the synthesizer a few years later. And, to this day, no one knows exactly how they created their music... (Almost no one, that is - it's my PhD topic ;-)

A 37-year-old wanting to learn computer science (initcoder.com | comments) A 37-year-old individual seeks to deeply learn computer science, showing previous self-taught web development experience. With a diverse career primarily as a teacher, they aim to invest the next decade mastering various technical skills, from API design to education app development, with family support.

• I wouldn't try to get into. I'm in my 30s and age discrimination is real. Nobody wants to hire a 30 something entry level engineer. They dont even want 30 something mid-level engineers. I know because I have a disability that seems to have capped me at these levels, I'm about to be fired, and the job prospects look abysmal.

Serving 200M requests per day with a CGI-bin (simonwillison.net | comments) Jake Gold evaluates the effectiveness of 1990s-era Common Gateway Interface (CGI) on modern hardware, demonstrating that with a Go and SQLite CGI program on an AMD 3700X, it can handle over 2400 requests per second, totaling 200 million requests per day. The analysis highlights advances in CPU technology that make CGI less inefficient than previously thought.

• Outside of nostalgia there's no engineering reason to do this - definitely not for performance. That same go program can easily go over 10k reqs/sec without having to spawn a process for each incoming request. CGI is insanely slow and insanely insecure.

Stop Hiding My Controls: Hidden Interface Controls Are Affecting Usability (acm.org | comments) Hidden interface controls hinder usability, a concern first broached by Douglas Engelbart in the 1960s. Engelbart's concepts of 'knowledge in the world' versus 'knowledge in the head' are examined, emphasizing that obscured controls can complicate user interaction and decrease efficiency in digital environments.

• My favorite is how Android shuffles buttons around just as I'm about to tap, so I wind up tapping the wrong one. I'm convinced advertisers will find a way to leverage that behavior in some new dark UI pattern.
• But the pervasive "form over function" design school disagrees with your desire for the UI to be useful, it has to look clean!

What 'Project Hail Mary' teaches us about the PlanetScale vs. Neon debate (alexoglou.com | comments) The ongoing debate between PlanetScale and Neon primarily revolves around performance, as highlighted by PlanetScale's CEO and various discussions on social media. The blog draws parallels to themes in 'Project Hail Mary,' particularly regarding scaling methodologies. Specific tweets and responses from the tech community illustrate the competitive landscape of database technologies.

• Tangentially related, is this book worth the hype? I don't read a lot of genre fiction, but don't like to miss out on the exceptional (just finished and loved Flowers for Algernon, as an example). Edit: Sounds like an enjoyable, low commitment book. Will give it a try, thanks for the feedback.

Local-first software: You own your data, in spite of the cloud (inkandswitch.com | comments) Local-first software advocates for user data ownership, countering the limitations of popular cloud applications like Google Docs. Highlighting principles such as offline access and security, the authors discuss the importance of collaborative yet user-controlled environments. They explore multi-user data structures called Conflict-free Replicated Data Types (CRDTs) as foundational technology for this new software paradigm.

• Anything with online dependencies will necessarily require ongoing upkeep and ongoing costs. If a system is not local-first (or ideally local-only), it’s not designed for long-term dependability. Connected appliances and cars have got to be the stupidest bit of engineering from a practical standpoint.

Problems the AI industry is not addressing adequately (thealgorithmicbridge.com | comments) Alberto Romero articulates a growing dissatisfaction with the AI industry's failure to address key issues such as reliability, accountability, and transparency. He emphasizes the need for ethical practices and urges stakeholders to prioritize fundamental problems over superficial enhancements, highlighting a loss of trust among supporters.

• Point 1. could just as easily be explained by all of the labs being very close, and wanting to jump ship to one that is closer, or that gives you a better deal.
• Are there some people here in HN believing in AGI "soonish" ?
• No-one authentically believes LLMs with whatever go-faster stripes are a path to AGI do they?

We're all CTO now (ideasasylum.com | comments) In the evolving tech landscape, every developer has transformed into a manager of AI coding agents, shifting focus from intensive coding to overseeing AI-generated outputs. While this can enhance productivity, it raises questions about job satisfaction and the dopamine rush associated with traditional coding challenges. As developers navigate this new role, they may struggle with maintaining their coding skills while adapting to management duties, posing a risk of dissatisfaction as the industry changes.

• If you're a solo developer building your next Salesforce killer, you will feel that dopamine rush every time AI helps you get closer to launch. Don't worry, there's still more coding problems after that one is solved by AI, because the "last 10%" is another 90% of work when it comes to polishing.

Optimizing Tool Selection for LLM Workflows with Differentiable Programming (viksit.substack.com | comments) The article details how utilizing local, learnable routers can optimize the selection of tools in large language model (LLM) workflows. It highlights the benefits of minimizing token overhead and reducing costs while also enhancing the structural organization of workflows through differentiable programming techniques with PyTorch and DSPy.

• I was experimenting with how local, learnable routers can reduce token overhead, and lower costs, and decided to publish a post about it. The main goal is to delegate tool calls via a PyTorch based learner and examples of how to integrate this into a DSPy pipeline. Feedback welcome!

Techno-Feudalism and the Rise of AGI: A Future Without Economic Rights? (arxiv.org | comments) The paper discusses the implications of Artificial General Intelligence (AGI) on economic and political structures, warning of a potential shift towards 'techno-feudalism.' It highlights AGI's dual role as both a worker and owner, advocating for new economic frameworks like universal AI dividends and equitable distribution of wealth to counteract rising inequality.

• Did the rise of fire, the wheel, the printing press, manufacturing, and microprocessors also give rise to futures without economic rights? I can download a dozen LLMs today and run them on my own machine. AI may well do the opposite, and democratize information and intelligence in currently unimaginable ways. It's far too early to say.

Cod Have Been Shrinking for Decades, Scientists Say They've Solved Mystery (smithsonianmag.com | comments) Investigating the longstanding issue of shrinking cod populations, scientists have identified climate change and overfishing as major contributors to this trend. Their research suggests changes in water temperature and chemistry directly affect cod growth, highlighting the urgent need for sustainable fishing practices to prevent further decline.

• My take away from this is that letting the small fish go under the premise that they are juveniles that will later grow to be bigger lets the adult midgets go, ruining the gene pool. I wonder if this finding will have any impact on conservation rules against taking small fish when fishing.

'Positive review only': Researchers hide AI prompts in papers (nikkei.com | comments) Research from 14 academic institutions across eight countries, including Japan and China, reveals that some researchers are embedding AI prompts in their papers to instruct reviewers towards favorable evaluations. This practice has sparked controversy around the integrity of peer review in scientific publishing.

• Just to be clear, these are hidden prompts put in papers by authors meant to be triggered only if a reviewer (unethically) uses AI to generate their review. I guess this is wrong, but I find it hard not to have some sympathy for the authors. Mostly, it seems like an indictment of the whole peer-review system.

The messy reality of SIMD (vector) functions (johnnysswlab.com | comments) SIMD (Single Instruction, Multiple Data) functions facilitate parallel processing by allowing a single operation to manipulate multiple data points simultaneously. This post elucidates their usage, advantages, and declaration methods, aiming to clarify misconceptions around their complexity and applicability in software development.

• I don't think the native C++, even when bundled with OMP, goes far enough. In my experience, ISPC and Google's Highway project lead to better results in practice - this mostly due to their dynamic dispatching features.

Baba Is Eval (fi-le.net | comments) The project 'Baba is Eval' aims to adapt the puzzle game 'Baba is You' for language models by enabling them to manipulate game states through automated interactions. This process involves fetching game data from its Lua scripts, implementing actions like movement and menu navigation, and addressing the unique challenges posed by the game's mechanics in a manner suitable for AI reasoning benchmarks.

• I would be way more interested in it playing niche community levels, because I suspect a huge reason it's able to solve these levels is because it was trained on a million Baba is You walkthroughs. Same with people using Pokemon as a way to test LLMs, it really just depends on how well it knows the game.

Heart attacks aren't as fatal as they used to be (vox.com | comments) The risk of death from heart attacks has dropped significantly, nearly 90% since 1970, primarily due to smoking reduction, statin medications targeting LDL cholesterol, CPR training, and advances in medical care. Today, over 90% of individuals over 65 hospitalized for heart attacks survive, reflecting remarkable progress in combating heart disease.

• Yeah cancer is the big killer nowadays. Survival rates for stage 4 cancer still poor after many decades of research. Worse yet, in many instances there are no obvious risk factors, such as people in their 30s or 40s who get colon cancer and were not eligible for screening .